CURRENT STATE OF KNOWLEDGE ON CLIMATE TRENDS AND VARIABILITY, AND DOWNSCALED CLIMATE CHANGE PROJECTIONS, FOR EASTERN AFRICA a summary report by The Climate System Analysis Group University of Cape Town South Africa Page |2 TABLE OF CONTENTS Regional Climate Zones and observed climate analysis ........................................................................................................ 8 Observed data ..................................................................................................................................................................... 8 Station observed data ...................................................................................................................................................... 8 Gridded observed data .................................................................................................................................................... 9 Re-analysis data sets ..................................................................................................................................................... 10 CSAG Observed Data.................................................................................................................................................... 11 Climate zones in eastern Africa ......................................................................................................................................... 11 Trend analysis ................................................................................................................................................................ 14 General country level observed climate ............................................................................................................................ 14 Priority landscape observed climate and trends ................................................................................................................ 20 Observed climate summary for region 1: Lamu Seascape............................................................................................ 23 Observed climate summary for region 2: Kwale – East Usambara Landscape ............................................................ 24 Observed climate summary for region 3: Udzungwa Landscape .................................................................................. 25 Observed climate summary for region 4: Rumaki and Matumbi Hill Sea and Landscape ............................................ 25 Observed climate summary for region 5: Greater Ruvuma Landscape ........................................................................ 28 Observed climate summary for region 6: Mtwara – Quirimbus Complex ...................................................................... 30 Observed climate summary for region 7: Primireas e Segundas Seascape ................................................................. 32 Observed climate summary for region 8: Delta of the Zambezi River and Marromeu Complex ................................... 32 Observed climate summary for region 9: Bazaruto Seascape ...................................................................................... 35 Observed ocean trends ......................................................................................................................................................... 35 Trends in Sea Surface Temperatures along the east african coast .............................................................................. 36 Trends in sea-level along the east african coast ........................................................................................................... 38 Trends in salinity and ph ................................................................................................................................................ 42 Climate variability and change .............................................................................................................................................. 44 Introduction ........................................................................................................................................................................ 44 The Seasonal Cycle .......................................................................................................................................................... 44 Inter-annual Variability ....................................................................................................................................................... 45 Climate Change ................................................................................................................................................................. 47 Climate Change in the CEAI region .................................................................................................................................. 49 Page |3 Climate Change and Natural Variability ............................................................................................................................ 50 Climate and ocean projections for priority landscapes ......................................................................................................... 51 Raw GCM Projections ....................................................................................................................................................... 51 An overview of downscaling .............................................................................................................................................. 54 Downscaling Methdology - SOMD .................................................................................................................................... 55 Summary projections for Priority place 1: Lamu Seascape ........................................................................................... 56 Summary projections for Priority place 2: Kwale – East Usambara Landscape ........................................................... 57 Summary projections for Priority place 3: Udzungwa Landscape ................................................................................. 60 Summary projections for Priority place 4: Rumaki and Matumbi Hill Sea and Landscape ........................................... 62 Summary projections for Priority place 5: Greater Ruvuma Landscape ....................................................................... 65 Summary projections for Priority place 6: Mtwara – Quirimbus Complex ..................................................................... 69 Summary projections for Priority place 7: Primireas e Segundas Seascape ................................................................ 71 Summary projections for Priority place 8: Delta of the Zambezi River and Marromeu Complex .................................. 73 Summary projections for Priority place 9: Bazaruto Seascape ..................................................................................... 77 Projected Ocean changes ..................................................................................................................................................... 77 Projected Sea Level rise .................................................................................................................................................... 77 Guidance on the use and application of downscaled climate projections ............................................................................ 78 Introduction ........................................................................................................................................................................ 78 Downscaling extreme events............................................................................................................................................. 78 Uncertainty and projections ............................................................................................................................................... 78 Confidence in projections .................................................................................................................................................. 78 Robust decision making .................................................................................................................................................... 79 Model filtering .................................................................................................................................................................... 79 Climate vulnerability assessment .......................................................................................................................................... 80 Undertaking a climate vulnerability assessment: .............................................................................................................. 82 Descriptive climate vulnerability assessment for priority landscapes: .................................................................................. 82 WWF Priority Place 1: Lamu Seascape (Kenya) ............................................................................................................... 82 Dry spell duration and drought: ...................................................................................................................................... 82 High temperatures: ........................................................................................................................................................ 82 Wind: .............................................................................................................................................................................. 82 Page |4 WWF Priority Place 2: Kwale landscape-East Usambaras Landscape (Kenya and Tanzania) ........................................ 83 Excessive / Intense Rainfall: .......................................................................................................................................... 83 Dry Spell Duration and Drought: .................................................................................................................................... 83 Wind: .............................................................................................................................................................................. 84 WWF Priority Place 3: Udzungwa Landscape (Tanzania) ................................................................................................ 84 Heavy / Intense Rainfall: ................................................................................................................................................ 84 Low Rainfall: .................................................................................................................................................................. 84 WWF Priority Place 4: RUMAKI and Matumbi Hill Sea and landscape (Tanzania) .......................................................... 85 Excessive Rainfall and Flooding: ................................................................................................................................... 85 Drought: ......................................................................................................................................................................... 85 Storm surges: ................................................................................................................................................................. 86 WWF Priority Place 5: Greater Ruvuma Landscape (Tanzania and Mozambique) .......................................................... 86 Excessive Rainfall and Flooding: ................................................................................................................................... 86 Wind: .............................................................................................................................................................................. 86 Lightning:........................................................................................................................................................................ 86 Drought: ......................................................................................................................................................................... 86 Fog/mist/low cloud: ........................................................................................................................................................ 87 WWF Priority Place 6: Mtwara-Quirimbas Seascape (Tanzania and Mozambique) ......................................................... 87 Wind, Hail and Storm events: ........................................................................................................................................ 87 Excessive / Intense Rainfall: .......................................................................................................................................... 87 Sea level changes: ......................................................................................................................................................... 88 Impacts of projected changes on priority landscapes ........................................................................................................... 88 WWF Priority Place 1: Lamu Seascape (Kenya) ............................................................................................................... 89 Dry Spell Duration: ......................................................................................................................................................... 89 High Temperatures: ....................................................................................................................................................... 89 WWF Priority Place 2: Kwale landscape-East Usambaras Landscape (Kenya and Tanzania) ........................................ 89 Excessive and Intense Rainfall: ..................................................................................................................................... 89 dry Spell duration: .......................................................................................................................................................... 89 High Temperatures / Heatwaves: .................................................................................................................................. 90 WWF Priority Place 3: Udzyngwa Landscape (Tanzania) ................................................................................................ 90 Page |5 Excessive and Intense Rainfall: ..................................................................................................................................... 90 Dry Spell Duration: ......................................................................................................................................................... 90 WWF Priority Place 4: RUMAKI and Matumbi Hill Sea and landscape (Tanzania) .......................................................... 90 Excessive and Intense Rainfall: ..................................................................................................................................... 90 Dry Spell Duration: ......................................................................................................................................................... 90 Storm surges: ................................................................................................................................................................. 91 WWF Priority Place 5: Greater Ruvuma Landscape (Tanzania and Mozambique) .......................................................... 91 Excessive and Intense Rainfall: ..................................................................................................................................... 91 Dry Spell Duration: ......................................................................................................................................................... 91 WWF Priority Place 6: Mtwara-Quirimbas Seascape (Tanzania and Mozambique) ......................................................... 91 Excessive and Intense Rainfall: ..................................................................................................................................... 92 Sea Level Rise: .............................................................................................................................................................. 92 Summary of climate vulnerabilities across all six priority regions: .................................................................................... 92 A methodological way forward:.......................................................................................................................................... 93 Step 1: Getting Started .................................................................................................................................................. 94 Step 2: Am I vulnerable to current climate? ................................................................................................................... 94 Step 3: How will I be affected by climate change? ........................................................................................................ 94 Step 4: Identifying, assessing and implementing adaptation options ............................................................................ 95 Step 5: Monitor and Review ........................................................................................................................................... 95 Observed data gaps .............................................................................................................................................................. 96 Opportunities and future activities ..................................................................................................................................... 96 Observed Data ............................................................................................................................................................... 96 Climate Change Projections .......................................................................................................................................... 97 Page |6 INTRODUCTION The WWF’s Coastal East Africa Initiative (CEAI) aims to add regional strategic focus to WWF’s work in Kenya, Tanzania and Mozambique. The Climate Change Adaptation Programme was initiated in early 2011 and aims to ensure that WWF programmes within the CEAI countries recognize and, where possible address, the impacts of global change on ecosystems and associated communities. The climate change programme has been tasked with the following outputs: develop a climate change adaptation strategy for WWF CEAI; provide foundation training on climate change adaptation to WWF management and programme staff in Kenya, Tanzania & Mozambique; conduct climate change vulnerability assessments in at least 3 priority landscapes or seascapes, and develop climate change adaptation strategies as appropriate; This report aims to contribute to the first of these outputs and consists of the following main sections 1. Regional climate zones and observed climate (climatology, trends and relevant extreme events) 2. Regional natural climate variability and climate change 3. Downscaled climate projections for priority landscapes 4. Guidance on the use and application of downscaled climate projections 5. Impacts of projected changes on priority landscapes The report is focussed on the nine priority landscapes identified by CEAI and identified in the map in Figure 1. It is important to note that this report is general in nature with only initial attempts to specialise or tailor information included. The scope of the report largely lies within the realms of a “state of knowledge” though the climate projections section does constitute new work tailored specifically to the CEAI landscapes. A specific aim of the report is to identify gaps, limitations and opportunities for further exploration and these are highlighted in the appropriate sections of the report as well as summarized in the concluding section on gaps and further work. Page |7 Figure 1: Map of CEAI priority landscapes Page |8 OBSERVED CLIMATE AND OCEAN AND OCEAN REGIONAL CLIMATE ZONES AND OBSERVED CLIMATE ANALYSIS A sound understanding of the existing regional climate is an essential foundation to any analysis of climate change or climate change impacts in a particular region. Regional climates are however complex and characterisation of climate can and should be sector specific and related to relevant aspects of climate that have an impact on a sector. For the purposes of this report we shall take an intentionally simple perspective on the climate as the context of the report is fairly general in nature. For these purposes therefore regional climate can be characterized by three basic parameters Seasonality or intra-annual variability (Winter/summer rainfall, bi-modal rainfall etc.) Inter-annual variability (variations from year to year and over longer periods such as decades) Climate trends (long term, monotonic changes over periods of 30 or more years) Depending on the context, each of these parameters can be described in more or less detail. For example, in an agricultural context, the variability in the date of the onset of rainfall would be of particular interest due to its profound effect on agricultural activities. However, in a water resource management context, seasonal total rainfall and the occurrence of multiple dry years in succession would be more relevant. However, before exploring specific details of the CEAI climate zones it is it is necessary to understand the primary means of understanding such climate zones, namely observed climate data. This report will largely be based on observed data and downscaled projections data derived partially from the original observed data. Observed data therefore forms a key foundation of the remainder of the report. OBSERVED DATA Any analysis of regional climate rests primarily on observational data. This section deals firstly with the broad context of observed datasets and associated caveats and complexities, and secondly with the observed data deployed in the development of this report. Station observed records, as the primary source of climate data, are addressed first. Following this is a section on gridded datasets, some of which are derived from station datasets and/or satellite data. Finally, re-analysis model based datasets are discussed. STATION OBSERVED DATA These are normally daily or monthly time series records of observations of various parameters at particular locations. The primary parameters are rainfall and temperature though some stations also record parameters such as radiation, pressure and wind. These observing stations are managed and operated by countries meteorological services or through private and non-governmental organisations. Access to the data is either through direct requests to the managing organisation or through a number of sources of collated data which are listed below: Global Historic Climate Network (GHCN) Provides global network of quality controlled and cleaned daily and monthly station data records sourced from a number of observing organisations including in country meteorological services. Page |9 UN Food and Agriculture Organisation (FAO) The FAO collates and manages a number of sources of station observed data and makes this available as monthly mean values The biggest challenge to the use of station observed records is data quality. This is particularly acute in the African context where many observations are still, and have historically, been done through manual observation and paper based record systems. Errors are introduced at a number of points including errors in actual measurement, sensor degradation or lack of maintenance, recording errors, archiving errors, and finally, digitization errors. Errors manifest in a number of ways including extreme values, unseasonal values, constantly repeating values, inexplicable trends and many other ways. A great deal of effort goes into the data quality control and cleaning (removal of suspect values) process. The GHCN (see above) has implemented an extremely comprehensive, largely automated, quality control, correction and cleaning process. This makes the GHCN archive one of the most useful available. CSAG has re-implemented a large part of the GHCN process using modified parameters more relevant to the African context in an attempt to address problems that the GHCN globally relevant parameters do not address. This processing has been applied to other sources of data for South Africa. One result of this process is that a large number of stations do not meet the very basic requirements that CSAG requires in order to perform climate variability analysis and climate downscaling. The minimum requirement is simply at least 10 years of valid observed records, however this limit is extended to 30 years of valid data for any trend analysis to be undertaken. Ten years of data allows for basic analysis of climatological means though for such short observational periods decadal variability can strongly skew the results. For the purposes of statistical downscaling, as described later in this report, 10 years is deemed the minimum period from which it is possible to construct a robust statistical model. Unfortunately, many errors are almost impossible to identify in an automated way and are only detectable through visual inspection, a task of that is great size when one considers the many hundreds of stations available over Africa. Even visual inspection it is difficult to identify many problems. The result is that many errors do propagate through and are therefore unfortunately included in the statistical analysis and downscaling. GRIDDED OBSERVED DATA The second primary type of data is gridded data sets. These data sets are derived products produced from station observed data, satellite image data, or both. A number of such data sets are available that cover the globe completely or just the land masses. The most prevalent of these is listed below: Climate Research Unit (CRU) gridded monthly rainfall and temperature Gridded (2.5° resolution) monthly total rainfall and monthly mean temperature data sets derived from a variety of station data sets including GHCN (see above) and other private sources. Global Precipitation Climatology Centre (GPCC) Gridded products derived from global station data of various spatial resolutions and representing monthly rainfall and temperature. Global Precipitation Climatology Project (GPCP). Gridded products derived from both global station data and satellite derived data of various spatial resolutions and representing monthly rainfall and temperature. CPC Merged Analysis of Precipitation (CMAP) Similar to GPCP, a merged analysis of station observations and satellite derived data. Monthly and pentad global analysis. Tropical Rainfall Measuring Mission (TRMM) Rainfall estimate from polar orbiting satellites. Polar orbiting satellites hosting a suite of sensors including rainfall radar, microwave, visible and infrared scanners provide data to estimate rainfall rates in clouds. P a g e | 10 Famine Early Warning System (FEWS NET) Rainfall Estimate (RFE) merging method using geostationary (Meteosat 7) satellite imagery, SSMI microwave satellite data and WMO Global Telecommunication System (GTS) station observations. A number of important issues should be considered in using gridded observed products, particularly in the African context: Pure Interpolation based gridded products such as CRU used purely station observation records and interpolate station values to represent grid area averages. There are a number of sophisticated methods of doing this but all methods have to make certain assumptions about how well each station represents the area average for a grid cell. In areas of sparse station coverage this is particularly problematic. The problem is confounded by the tendency for stations to be located near urban areas or in accessible areas rather than mountainous areas. So for regions with complex topography and high mountains, which are often critical features in the climate and hydrological system, gridded rainfall are skewed towards available observations in low lying areas rather than in the mountain peaks. More modern methods attempt to address the above problem by including satellite derived data which has far more consistent spatial coverage. This is an important improvement. However there are again constraints and concerns. Satellite based estimates of rainfall are typically based on parameters such as cloud top heights and temperatures, cloud extents, satellite based precipitation radar and microwave sensors. All of these sensing methods have biases and problems. Additionally, sensors mounted on polar orbiting satellites are constantly moving around the earth and hence a particular spatial area is only revisited a few times a day, depending on the product. While much work has been done in improving satellite derived precipitation estimates, results still show that performance is best in areas of extensive convection and poorest in areas of high topography, orographic1 rainfall and stratiform2 rainfall. Merged station and satellite products suffer from problems with homogeneity or step jumps where different satellite products or stations come or go through the time period for a particular grid point. Products such as GPCP should only be used for trend or variability analysis with some caution as described in the relevant documentation and academic papers. Another aspect of satellite products is the relatively short observational period due to the recent development of satellite sensing platforms. While the earliest satellites were introduced in the late 1970s, the more recent satellites with specific rain rate sensing capabilities were only introduced in the late 1990s (TRMM launched in 1997). While this will clearly improve in the future, for climate variability and change studies the satellite record is often simply too short and dependence on surface observations is still heavy. Finally, satellite derived products are largely focussed on rainfall and do not address the issue of surface air temperature which is difficult to detect through satellite measurements. While rainfall is probably the most important variable in the context of societal and ecological impacts, temperature is also an important variable in many situations. Station based observations therefore still play a critically important role as both primary sources of data as well as a means to calibrate and validate satellite based products. RE-ANALYSIS DATA SETS Another increasingly important source of pseudo observed data is that of atmospheric re-analyses. A re-analysis is produced by a numerical model that is constrained in some way by observations. In some senses this produces a sophisticated interpolation of observations where the interpolated values are informed not just by the observations but by atmospheric physical processes such as advection, convection, radiation and evaporation. The result is potentially a more realistic interpolation. A number of global scale re-analysis have been produced such as the National Center for Environmental Predication (NCEP) versions 1 and 2. However, these re-analysis should be used with caution as they the 1 2 Orographic rainfall is rainfall that is produced by the uplift of moist air as it is forced over mountainous topography Stratiform rainfall is produced by low level cloud resulting from general large scale uplift rather than localized convective uplift P a g e | 11 degree to which the accurately reflect observed climate varies greatly depending on the location and the variable. Typically they do a fairly good job at non-surface variables such as winds and pressure fields but still struggle to capture surface variables such as rainfall. They are also limited in resolution with NCEP being at approximately 200km spatial resolution. The recently released CFSR re-analysis fields are at a much higher 50km resolution and show significant promise but should still be used with caution and only after proper validation of the variable in question. CSAG OBSERVED DATA Due to the scarcity of station observations contributing to the gridded products over Africa as well as the short record length of satellite derived observed products, CSAG has selected the option of relying on station scale observed records for the climate analysis and climate downscaling components of this report. It is tempting to resort to gridded interpolated products such as CRU to fill in the gaps in the station network but such products consist of monthly records only which limits the detail of the analysis that can be done. Such interpolated products are also based on very limited station networks and as such, while they may provide data for the full area, the data quality and representivity is inherently limited. CSAGs station archive for Africa consists largely of a combination of GHCN records and WMO sourced records. All of the station records are archived on a daily timescale allowing for detailed analysis of daily statistics so critical to many climate impacts studies. While monthly mean records are often available, analysis of climate impacts is severely limited when using such data as many climate extremes and impacts occur at a sub-monthly time scale. CSAG applies a rigorous cleaning process that extends the standard GHCN cleaning process by using parameters more specific to the African context. This allows for the identification and removal of more suspect values. The resultant station locations for the study region are identified in figure x [map of station locations] It must be noted that the lack of observed data is a severe constraint on robust climate analysis in the region and access to additional observed records would greatly enhance opportunities to explore regional climate dynamics, trends and future projections. CLIMATE ZONES IN EASTERN AFRICA The region under analysis in this report covers three countries, Kenya, Tanzania and Mozambique, bounded to the east by the Indian Ocean and to the west by major water bodies including Lakes Nyasa, Tanganyika and Victoria. . The individual priority landscapes and seascapes of particular interest to WWF are identified in figure 1 above. From a climate perspective the region covers the classic climate zones of tropical and sub-tropical. However regional climate drivers, such topography and water bodies (oceans and lakes) impose a complexity to these climate zones. A number of methods can be used to identify regional climate zones. One example of a global delineation of climate zones is the recently updated Koppen Geiger climate zones3 as illustrated in Error! Reference source not found. (map f Köppen Geiger zones for the area). These zones are, however, fairly generic as the methodology is designed to be applied on a global scale. The Köppen Geiger classification is based on mean annual temperature and precipitation as well as the seasonality of precipitation. The original objective was to produce climate zones that reflect the natural vegetation zones observed across the globe under the assumption that vegetation zones are a good proxy of climate zones. A full description of the methodology can be found in the referenced paper. The zones are divided into the following main climates: 3 A: Equatorial B: Arid C: Warm temperate Kottek, M., J. Grieser, C. Beck, B. Rudolf, and F. Rubel, 2006: World Map of the Köppen-Geiger climate classification updated. Meteorol. Z., 15, 259-263. DOI: 10.1127/0941-2948/2006/0130. P a g e | 12 D: Snow E: Polar Each major zone is annotated with the following rainfall characteristics: W: Desert S: Steppe f: Fully humid s: Summer dry w: Winter dry m: Monsoonal and the following temperature characteristics: h: Hot arid k: Cold arid a: Hot summer b: Warm summer c: Cool summer d: Extremely continental Most of the CEAI priority sites fall within the As/w regions which describes a tropical dry summer or winter depending on the location being in the northern hemisphere or southern hemisphere. P a g e | 13 Figure 2: Map of Köppen-Geiger climate zones for the CEAI region. The coarseness of the Köppen Geiger classifications means that a lot of the local detail for particular locations is lost and hence its utility for a regional climate analysis is minimal. Additionally, the climate zones have been developed using historical observed climate (CRU and GPCC) which have poor station coverage in coastal east Africa and hence the delineation of the zones is unreliable. Finally, in the context of historical changes or projected future changes, there is little indication that such climate zones will represent areas of homogenous change. Firstly, the scarcity of reliable observing stations within the region makes it difficult to identify generalized climate zones. While it is tempting to utilise a gridded product such as CMAP, the lack of observing stations contributing to CMAP over P a g e | 14 this area raises serious doubts about the accuracy and hence utility of such a data source. This is highlighted in the associated CMAP documentation paper 4 as well as a number of other sources. Secondly, some of the priority landscapes themselves are fairly extensive in size, particularly region 5 (Greater Ruvuma Landscape) and would likely intersect more than one climate zone making any analysis of the climate within a priority landscape complex if restricted to analysis within distinct climate zones. The remainder of this report will however avoid the use of generalised climate. The report will concentrate rather on the particular observing station locations that are available and attempt to infer, where justified, the more general regional climate. However in some cases the stations available are not suitable as they do not represent the regional climate well and this will be noted as a data limitation. TREND ANALYSIS Trend analysis is a fairly complex undertaking in the context of observed climate data. Data quality and length of a record are significant obstacles to producing any kind of statistically significant trend analysis for a station location. It is also difficult, without detailed analysis, to identify trends that are indeed long term monotonic trends rather than an artefact of some other source of inter-annual variability that may be operating on a decadal or longer time period. The recent paper by Lyon and De Witt, (Geophysical Research Letters, 2012) identifies a recent decrease in rainfall in the east African region. However the area analysed is large and further exploration of the source data (GPCP and GPCC) shows that local regional scales the presented decline is not present. Therefore while such work is of course critical to our understanding of the regional climate dynamics, local scale rainfall is a function of both the large scale dynamics and regional scale complexities. A further concern is that products such as GPCP should not typically be used for trend analysis due to the problems of homogeneity or step jumps as described in the section above on observed data. For the purposes of this report we undertake only to do the simplest form of trend analysis which involves fitting a linear regression to the monthly time series for observed station records. Trends are however calculated on each month of the year independently in order to disaggregate the seasonality of trends as many long term trends do not reflect in annual means but rather in specific months of the year. This is particularly important of trends in seasonality where the timing of the seasons is changing over time. The analysis included below allows such shifts to be identified. Such monthly trend plots are produced for stations within the priority landscapes. Time series plots of annual mean rainfall are also presented for a number of areas across the three countries. Trends have not been calculated for these locations as they are sourced from the GPCP data and hence fitting statistical trends is not recommended. The time series plots do however highlight the complexity of rainfall trend analysis where determining a significant trend in a time series of large interannual and, more problematically, decadal variability, is extremely difficult. The results presented should be taken as indicative of possible historical trends but due to the reasons described above, any action or decisions based on such trends should first consider a more detailed analysis. GENERAL COUNTRY LEVEL OBSERVED CLIMATE Figure 3 below shows the general rainfall climatology over the three countries of Mozambique, Tanzania and Kenya for sub-periods of the year (Dec-Jan, Mar-May, Jun-Aug, Sep-Nov). The data is sourced from GPCP in order to provide complete coverage though it must be recognised that the station density in the region is low and varies dramatically over time. The data is plotted on the native 2.5° latitude/longitude grid in order to illustrate the spatial representivity. More information is available in the official GPCP documentation. 4 Xie, P., and P. A. Arkin, 1997: Global Precipitation: A 17-Year Monthly Analysis Based on Gauge Observations, Satellite Estimates, and Numerical Model Outputs. Bull. Amer. Meteor. Soc., 78, 2539--2558. P a g e | 15 Figure 3: Observed monthly mean rainfall (mm/month) for four sub-seasons for the period 1979-2010 (Source GPCP) While the GPCP time series data should not generally be used for trend analysis due to the possibility of inhomogeneities, timeseries have been extracted for a number of locations throughout the CEAI countries. The locations of the timeseries are highlighted in the map below (Figure 4) and the time series plots themselves follow P a g e | 16 Figure 4: GPCP Time series locations P a g e | 17 Figure 5: GPCP annual mean rainfall for area 1 Figure 6: GPCP annual mean rainfall for area 2 P a g e | 18 Figure 7: GPCP annual mean rainfall for area 3 Figure 8: GPCP annual mean rainfall for area 4 P a g e | 19 Figure 9: GPCP annual mean rainfall for area 5 Figure 10: GPCP annual mean rainfall for area 6 P a g e | 20 Figure 11: GPCP annual mean rainfall for area 7 The above plots show that the historical record over annual average rainfall is complex and varies significantly over the CEAI countries. They also highlight the difficulty in determining long term trends from such time series. Statistical measures of trends should be used cautiously with such time series as they often produce over-confident parameters that are miss-interpreted. One marked feature that appears in the time series for Kenya is the high rainfall for 1997 which was a strong El Nino year and resulted in heavy rainfall in tropical East Africa and in many areas was a causative factor in outbreaks of diseases such as rift valley fever. PRIORITY LANDSCAPE OBSERVED CLIMATE AND TRENDS Given the above discussion on observed data the following sections will explore the observational records relevant to each priority landscape (Figure 1) and briefly describe the climate characteristics identified previously. These are seasonality, inter-annual variability, and long term trends. The climate of the study areas is explored using historic climate data from weather stations located within each of the study areas. The station data was obtained from the Global Historic Climate Network (GHCN) daily dataset 5 as well as from a dataset obtained directly from the World Meteorological Organisation (WMO). Of the two datasets, the GHCN data is generally of higher quality and was therefore the preferred dataset, however the WMO dataset contains a number of stations not included in the GHCND dataset and was therefore also included. All station records within this broad study area were assessed. But due to the sparsity of stations and the low quality and completeness of their records, only 12 stations were found to provide suitable data for the study area. Of these, 6 provide daily rainfall and maximum and minimum temperature, while the others only provide daily precipitation. The period of coverage differs between stations with only four stations having sufficiently long and complete records required for robust trends to be calculated (minimum 30 years of valid data). A map of the station location is presented in figure 4 with a brief summary of the station attributes presented in Table 1. 5 http://www.ncdc.noaa.gov/oa/climate/ghcn-daily P a g e | 21 All the station records were subjected to a suite of automatic quality assurance checks, based on those developed by the National Climatic Data Centre6, The prevalence of errors in some of some of these records was so extensive that the automatic cleaning was ineffective in identifying and removing all the errors and therefore the interannual variability of some of these records may be exaggerated. Two of the most common errors found within these station records include: zero rainfall being recorded instead of the required undefined value during periods where the station was not active/reporting. Spikes / dips in the temperature records are prevalent especially within the Tanzanian stations, where the temperature varies by up to 25°C from one day to the next. Figure 12: Map showing the station locations for each of the 9 regions. Region 1: Lamu (A) Region 2: Mombasa (B) and Tanga (C). Region 3: Iringa (D). Region 4: Dar Es Salaam (E). Region 5: Songea (F) and Lichinga (G). Region 6: Mtwara (H). Region 7: Lumbo (I. Region 8: Quelimane (J and Beira (K) Region 9: Vilanculos (L). The historic climate of a region is traditionally described by the mean value of variables such as precipitation and temperature, but it should also include information on the variability, extremes and trends in these variables. Secondary variables or monthly statics of these raw variables is often useful in better characterizing the climate of a region. Region Station ID Lat Lon Altitude variables Period 1 Lamu 63772 -2.27 40.83 30 ppt 1979 - 2008 2 Mombasa 63820 -4.03 39.62 55 ppt, tmax, tmin 1957 - 2003 2 Tanga 63844 -5.08 39.07 35 ppt 1979 - 2000 3 Iringa 63887 -7.63 35.77 1428 ppt 1979 - 2009 4 Dar es Salaam 63894 -6.87 39.20 55 ppt, tmax tmin 1958 - 2000 6 http://journals.ametsoc.org/doi/pdf/10.1175/2010JAMC2375.1 P a g e | 22 5 Songea 63962 -10.68 35.58 1067 ppt, tmax, tmin 1979 - 2000 5 Lichinga 67217 -13.30 35.23 1365 ppt, tmax tmin 1951 - 2000 6 Mtwara 63971 -10.27 40.18 113 ppt 1979 - 2009 7 Lumbo 67241 -15.03 40.67 11 ppt 1979 - 2008 8 Quelimane 67283 -17.88 36.88 16 ppt, tmax, tmin 1951 - 2003 8 Beira 67297 -19.80 34.90 16 ppt, tmax, tmin 1951 - 2000 9 Vilanculos 67315 -20.00 35.32 21 ppt 1979 - 2008 Table 1: Brief summary of the station observation records for each of the 9 study regions. ID = WMO station identifier number, Lat = Latitude of the station (decimal degrees), Lon = Longitude of station (decimal degrees), ppt = precipitation (mm/day), tmax = maximum daily temperature, tmin = minimum daily temperature. Explanation of observed climate plots Rainfall: Wide bars indicate the median monthly rainfall for the climate period. Narrow bars indicate the 10 th and 90th percentile range of monthly rainfall for each month during the climate period and hence indicate the magnitude of interannual variability. Temperature: Blue envelope indicates the 10th to 90th percentile range of monthly mean daily min/max temperature and hence the magnitude of the inter-annual variability. Dashed line indicates the median monthly mean daily min/max temperature. Trends: Wide bars indicate the decadal trend for each month during the indicted climate period (magnitude of the trend per decade). Dry spell duration: Dry spell duration is a measure of the average length of all dry spells that fall within or partly within a particular month. The methodology allows for a representation of dry spell durations only a monthly basis while still accommodating the occurrence of dry spells longer than one month. If a dry spell does extend beyond a single month then it will contribute proportionally to each months mean dry spell duration. P a g e | 23 OBSERVED CLIMATE SUMMARY FOR REGION 1: LAMU SEASCAPE The observed climate for the Lamu Seascape is represented by the weather station at Lamu. This station provides precipitation data for the period 1979 – 2000. Figure 13: Annual cycle of monthly rainfall (mm) for Lamu station. Figure 14: Annual cycle of monthly mean dry spell duration for Lamu station. Figure 15: Annual cycle of monthly rain days > 90th percentile for Lamu (90th percentile = 25.2 mm). The Lamu station has a tropical climate which is modulated by its location on the coast. It experiences bimodal rains with the long rains occurring from April - June and short rains from November to December. The dry spell duration is short throughout the year with the exception of the main dry season (January – March) where long durations between rain events are evident during some years, however this interannual variability may be exaggerated due to errors within the data. The long rains experience more frequent and intense rainfall than the short rains, with heavy (> 10mm) and extreme rainfall (> 90th percentile) being experienced most often during May. P a g e | 24 OBSERVED CLIMATE SUMMARY FOR REGION 2: KWALE – EAST USAMBARA LANDSCAPE The observed climate and future projected climate change for the Kwale – East Usambara Landscape are represented by the weather stations at Monbasa and Tanga. The Mombasa station provides daily minimum and maximum temperature and precipitation for the period 1957-2003, while the Tanga station only provides daily precipitation data (1979-2000). Large gaps in the records and the low quality of the records make these records unsuitable for trend analysis to be performed for this region. Figure 16: Annual cycle of monthly rainfall (mm) for Mombasa (left) and Tanga (right) stations. Figure 17: Annual cycle of monthly rain days > 90th percentile for Mombasa (90th percentile = 25.7 mm) (left) and Tanga (90th percentile = 24.0 mm)(right). Figure 18: Annual cycle of monthly mean maximum daily temperatures (deg C) for Mombasa (left) and TANGA (right) stations. P a g e | 25 Figure 19: Annual cycle days/month exceeding 32 deg C for Mombasa (left) and TANGA (right) stations. The Mombasa and Tanga stations have a tropical climate which is modulated by their proximity to the coast. They experience bimodal rains centered on April - June (‘long’ rains) and October to December (‘short’ rains) with a core dry season from January to March. Light rainfall occurs from March – December, but heavy and extreme rainfall is restricted to the long and short rains. Temperatures are warm with low diurnal and inter-seasonal variability. The warmest temperatures occur during March where more than half of the days exceed 32°C and coolest temperatures occur around August. OBSERVED CLIMATE SUMMARY FOR REGION 3: UDZUNGWA LANDSCAPE The observed climate and future projected climate change for the Udzungwa Landscape is represented by the weather station at Iringa. This station provides precipitation data for the period 1960 – 1990. Figure 20: Annual cycle of monthly rainfall (mm) for Iringa station. The Iringa station is found at high elevation in central Tanzania. Even though it is located within the tropics, the rainfall record shows a single rainfall season between December and April instead of the typical bimodal cycle. OBSERVED CLIMATE SUMMARY FOR REGION 4: RUMAKI AND MATUMBI HILL SEA AND LANDSCAPE The observed climate and future projected climate change for the Rumaki and Matumbi Hill Sea and Landscape is represented by the weather station at Dar es Salaam, since it is the closest station to the study area. This station provides daily minimum and maximum temperature and precipitation data for the period 1958 – 2000. The quality of the record deteriates after 1990 and therefore trends were only undertaken for the period 1958-1990. P a g e | 26 Figure 21: Annual cycle of monthly mean dry spell duration for Dar Es Salaam station. Figure 22: Decadal trend of monthly rainfall (mm/decade) for Dar Es Salaam station. Figure 23: Annual cycle of monthly mean dry spell duration for Dar Es Salaam station. Figure 24: Decadal trend of monthly rainfall (mm/decade) for Dar Es Salaam station. P a g e | 27 Figure 25: Annual cycle of monthly mean maximum daily temperatures (degC) for Dar Es Salaam station. Figure 26: Decadal trend of monthly mean maximum daily temperatures (degC/decade) for Dar Es Salaam station. Figure 27: Annual cycle of monthly mean minimum daily temperatures (degC/decade) for Dar Es Salaam station. Figure 28: Decadal trend of monthly mean minimum daily temperatures (degC) for Dar Es Salaam station. Dar es Salaam features a tropical climate which is modulated by its location on the coast. It experiences bimodal rains with the long rains occurring from March – May and the short rains from November – December. Dry spells between rain events is generally short since light rain occurs throughout the year, but heavy and extreme rainfall are restricted to the P a g e | 28 two peak rain seasons. The rainfall trends are complex, but do suggestion that the frequency and intensity of rainfall is generally decreasing, while the tails of the rainy seasons are getting wetter. Temperatures are warm with low diurnal and interseasonal variability. The warmest temperatures occur during February – March, where on average half the days exceed 32°C, while the coolest temperatures occur in July and August. Generally negative trends are found in the maximum temperature and the number of days exceeding 32°C with the largest trends occurring during the peak of the two rain seasons. A positive trend is found in minimum temperature throughout the year but is strongest during the first half when minimum temperatures are higher. OBSERVED CLIMATE SUMMARY FOR REGION 5: GREATER RUVUMA LANDSCAPE The Greater Ruvuma Landscape covers a large geographic and does not represent a unified climatic zone. The observed climate and future projected climate change for this region is represented by the weather stations at Lichinga and Songea, since these are the only two stations which met the data quality standards. These stations are located in the more mountainous terrain in the far west of the study area and therefore do not provide the most representative climate for the whole landscape. The Songea station provides daily data for the period 1979-2000, while the Lichinga station record extends from 1951-1990 and meets the data quality requirement for trend analysis to be performed. Figure 29: Annual cycle of monthly rainfall (mm) for Lichinga (left) and Songea (right) stations. Figure 30: Decadal trend of monthly rainfall (mm/decade) for Lichinga station. P a g e | 29 Figure 31: Annual cycle of monthly mean dry spell duration for Lichinga (left) and Songea (right) stations. Figure 32: Decadal trend of monthly mean dry spell duration (days/decade) for Lichinga station. Figure 33: Annual cycle of monthly mean maximum daily temperatures (degC) for Lichinga (left) and Songea (right) stations. Figure 34: Decadal trend of monthly mean maximum daily temperatures (degC/decade) for Lichinga station. Figure 35: Annual cycle of monthly mean minimum daily temperatures (degC/decade) for Lichinga (left) and Songea (right) stations. P a g e | 30 Figure 36: Annual cycle of monthly mean minimum daily temperatures (degC) for Lichinga station. The two stations within the Greater Ruvuma Landscape are found within the tropics however their climates are modulated by their location far inland and at higher attitude than the rest of the study area. The stations experience a single rainy season during summer from November to April and a clear dry season during the winter months. Rainfall is higher and more frequent in the Lichinga station than the Songea. Rainfall trends are presented for Lichinga and suggest an expansion of the rainy season with positive trends in March, October and November for both frequency and intensity of rainfall. There is also a positive trend in the dry spell duration from June to September. The temperatures of these stations are lower than the other stations and probably the rest of the Greater Ruvuma Landscape, with very few days exceeding the 32° C threshold.. Maximum daytime temperatures are found at the beginning of summer (October –November) before they peak of the rainy season, while the lowest maximum temperatures are found in July. Highest minimum temperatures are found throughout summer (November – April, but decrease during the core of the winter. Trend analysis of the Lichinga station record shows a negative trend in maximum and minimum temperatures during October and November which may be associated with the increase trend in rainfall a during this time. OBSERVED CLIMATE SUMMARY FOR REGION 6: MTWARA – QUIRIMBUS COMPLEX The observed climate and future projected climate change for the Mtwara – Quirimbus Complex is represented by the weather station at Mtwara. This station provides precipitation data for the period 1957 – 2000, however the quality of the record deteriorates after 1990 and therefore trends were only undertaken for the period 1957-1990. Figure 37: Annual cycle of monthly rainfall (mm) for Mtwara station. P a g e | 31 Figure 38: Decadal trend of monthly rainfall (mm/decade) for Mtwara station. Figure 39: Annual cycle of monthly rain days > 90th percentile for Mtwara (90th percentile = 29.2 mm) (left) and Beira (90th percentile = 38.0 mm)(right). Figure 40: Annual cycle of monthly mean dry spell duration for Mtwara station. Figure 41: Decadal trend of monthly mean dry spell duration (days/decade) for Mtwara and Beira (right) stations. P a g e | 32 The Mtwara station is located on the coast and experiences a tropical climate with summer rainfall occurring between December – April. Rainfall frequency and intensity is high throughout the summer, with extreme rainfall being most common during January. The rainfall trends are quite complex with months within the peak rainy season showing trends of alternating sign. The trend in dry spell duration shows a positive trend during winter associated with the decrease in the frequency of rain events during this time. Of all months, February shows the strongest trend (negative) in both the frequency and intensity of rainfall. OBSERVED CLIMATE SUMMARY FOR REGION 7: PRIMIREAS E SEGUNDAS SEASCAPE The observed climate and future projected climate change for the Primireas e Segundas Seascape is represented by the weather station at Lumbo. This station provides precipitation data for the period 1979-2000. Figure 42: Annual cycle of monthly rainfall (mm) for Lumbo station. Figure 43: Annual cycle of monthly rain days > 90th percentile for Lumbo (90th percentile = 28.0 mm). The Lumbo station is located on the coast just to the north of the study area. The site experiences predominantly summer rainfall from November to April with maximum monthly rainfall averaging around 120 mm/month. Heavy and extreme rainfall is restricted to the summer months and dry spell duration is longest between September – November. OBSERVED CLIMATE SUMMARY FOR REGION 8: DELTA OF THE ZAMBEZI RIVER AND MARROMEU COMPLEX The observed climate and future projected climate change for the Delta of the Zambezi River and Marromeu Complex is represented by the weather stations at Quelimane and Beira. Both station provide all three daily variables for the period 1951-1990 and are of sufficiently high quality that trend analysis was possible. P a g e | 33 Figure 44: Annual cycle of monthly rainfall (mm) for Quelimane (left) and Beira (right) stations. Figure 45: Decadal trend of monthly rainfall (mm/decade) for Quelimane (left) and Beira (right) stations. Figure 46: Annual cycle of monthly rain days > 90th percentile for Quelimane (90th percentile = 32.4 mm) (left) and Beira (90th percentile = 38.0 mm)(right). Figure 47: Decadal trend of monthly rain days > 90th percentile (days/decade) for Quelimane (90th percentile = 32.4 mm) (left) and Beira (90th percentile = 38.0 mm)(right). P a g e | 34 The Quelimane and Beira stations are located along the Mozambique coast within the summer rainfall region. Total monthly rainfall is high from December to March averaging around 200 mm/month. Rainfall is most frequent during summer, but light rainfall occurs throughout the year with a minimum during spring. Extreme rainfall is restricted to the core summer season. Rainfall trends are complex and differ between the two locations. Figure 48: Annual cycle of monthly mean maximum daily temperatures (degC) for Quelimane (left) and Beira (right) stations. Figure 49: Decadal trend of monthly mean maximum daily temperatures (degC/decade) for Quelimane (left) and Beira (right) stations. Figure 50: Annual cycle days/month exceeding 32 deg C for Quelimane (left) and Beira (right) stations. P a g e | 35 Figure 51: Decadal trend in days/month exceeding 32 deg C for Quelimane (left) and Beira (right) stations. Temperatures are warm over these locations, though Beira is slightly cooler than Quelimane. A clear seasonal cycle is evident with warmest daytime temperatures occurring during summer and coolest temperatures in June and July. Maximum temperatures exceeding 32°C occur more often for Quelimane than Beira with up to half of the days during January and February reaching this threshold. Minimum temperatures are mild with little interannual variability. Temperature trends for these two locations also show significant differences between locations. Quelimane shows negative trends in all months for both maximum and minimum temperature. While Beira shows generally positive trends. OBSERVED CLIMATE SUMMARY FOR REGION 9: BAZARUTO SEASCAPE The observed climate and future projected climate change for the Bazaruto Seascape is reporesented by the weather station record at Vilanculos. This station provides precipitation data for the period 1979 – 2000. Figure 52: Annual cycle of monthly rainfall (mm) for Vilanculos station. The Vilanculos station is located on the coast of Mozambique in the subtropics. The monthly total rainfall is low in relation to the other regions, however the interannual variability is very high suggesting that there may be errors remaining in the data. The rainy season extends from November to April, with maximum frequency and intensity occurring during February and March. OBSERVED OCEAN TRENDS Historically, observations of ocean parameters such as temperature, salinity and sea-level, has been restricted to coastal measurements and ship borne measurements. This has restricted the spatial coverage of observations. In particular, ship based measurements are not continuous but take place at a specific place and time as the ship travels along its path. Repeat measurements for specific locations are therefore intermittent and rare. Coastal observations due provide long time period observations but coastal waters can be strongly influences by the coastline itself, river outflows, sedimentation etc. Satellite observations of the ocean have therefore become a critical source of data for ocean research in the past two decades. Early sensors were largely restricted to measurements of ocean surface temperature but more recent satellites deploying multi-spectral sensors, x-ray, and radar sensors, are able to determine or provide proxy information about many other parameters. Surface winds, surface height, bio-chemistry and waves are just a few of these parameters. Sub-surface measurements have been performed using buoys and systems such as ARGO which consists of a large number of drifting buoys that periodically sink to a depth of up to 2km before ascending and transmitting measured data back to a series of data centres through a satellite data link. This section of the report will attempt to summarise some of the generally available information about the observed ocean relevant to the CEAI countries drawing significantly from Chapter 5 of the IPCC Fourth Assessment Report. P a g e | 36 TRENDS IN SEA SURFACE TEMPERATURES ALONG THE EAST AFRICAN COAST The Reynolds Optimal Interpolated (ROI) Sea Surface Temperature (SST) dataset7 provides a monthly time averaged, gridded, nearly global, dataset of ocean surface temperatures. This dataset has formed the backbone of much SST research. According to results presented in the IPCC AR4, SSTs in the Indian Ocean have generally increased at a rate of as high as 0.2°C/decade in the last decade. Regional trends in SST are however more complex due to shifts in ocean circulation as well as coastal effects such as upwelling. The plots below are extracted from the ROI dataset for areas along the coasts of Mozambique, Tanzania and Kenya respectively. West Indian Ocean SSTs are strongly linked to ENSO and the plots below exhibit some of the strong ENSO events including the 1997 event. All the plots show a steady increase in temperatures throughout the 30 year period. Figure 53: SST timeseries from 1981 – 2010 for the coast of Kenya 7 Reynolds, R.W., N.A. Rayner, T.M. Smith, D.C. Stokes, and W. Wang, 2002: An improved in situ and satellite SST analysis for climate. J. Climate, 15, 1609-1625. P a g e | 37 Figure 54: SST timeseries from 1981 – 2010 for the coast of Tanzania Figure 55: SST timeseries from 1981 – 2010 for the north coast of Mozambique P a g e | 38 Figure 56: SST timeseries from 1981 – 2010 for the southern coast of Mozambique TRENDS IN SEA-LEVEL ALONG THE EAST AFRICAN COAST Tide gauge sourced observations take place along the coastline as gauges are typically installed at harbours. These observed the relative level of the sea-level at high frequency for the purpose of monitoring tidal levels and extreme events such as storm surges. A critical issue with tide gauges is the calibration and correction for vertical movement of continental plates which in some areas is significant. A tide gauge can register an increase in sea level as a result of either a real increase in absolute sea-level, or a relative increase due to subsidence of the associated continental land mass. The second source is ocean bottom pressure gauges which measure the pressure of the column of water above the gauge. These can be installed in the middle of large ocean areas providing useful insight into sea levels across basins. They also require careful calibration and correction in order to be used for long term sea-level rise trend analysis. Finally, more recent satellite derived data is obtained from satellite based sensors that measure the distance between the satellite and the ocean surface with accuracy of a few millimetres and at high spatial resolution. This provides a unique perspective of ocean levels across all the major ocean basins and has revealed a great deal of detail about ocean circulation and dynamics. The observational record is fairly short however and so the use of satellite derived sea-level rise trends is fairly new. It is important to note that while global scale or even basin scale sea-level rise is largely a response to increase ocean temperatures as well as the contribution of melting continental ice, regional scale sea-levels are much more complex. Regional ocean currents strongly influence local scale sea-levels and shifts in regional ocean circulation patterns, either over long time periods due to climate change, or over years or decades due to decadal variability, are significant. There are many places in the world where local sea-levels appear to be dropping even when the basin scale sea-level is rising. This makes understanding local scale sea-level trends difficult and applying future projections of sea-level rise complex without much further analysis of the regional dynamics. P a g e | 39 A number of sea-level reconstructions have attempted to merge all sources of sea-level data (tide gauges, bottom pressure sensors, and satellite) to produce a longer time period, global coverage, moderate resolution time series of sealevel. These reconstructions have highlighted the complexity and region detail of global sea-level change over time. Figure 57 below identified the western Indian Ocean as one of a number of regions where observed sea-level has actually fallen over the last several decades. P a g e | 40 Figure 57: (a) Geographic distribution of short-term linear trends in mean sea level (mm yr–1) for 1993 to 2003 based on TOPEX/Poseidon satellite altimetry (updated from Cazenave and Nerem, 2004) and (b) geographic distribution of linear trends in thermal expansion (mm yr–1) for 1993 to 2003 (based on temperature data down to 700 m from Ishii et al., 2006). P a g e | 41 The Permanent Service for Mean Sea Level (PMSL) 8 archives records of sea-level measurements from a number of sources across the globe. For the CEAI countries two stations are available that have long records. These records are for Mombasa and Zanzibar. Figure X below shows the monthly averaged sea-level records for Mombasa while figure X shows the same for Zanzibar. Figure 58: Observed monthly sea level – Mombasa (source www.pmsl.org) Figure 59: Observed monthly sea level - Zanzibar (source www.pmsl.org) While the variability of these observations is high, visually we can see that while Zanzibar may be experiencing an increase in seal level, Mombasa appears to be experiencing a decrease in sea-level in line with basin scale observations (see Figure 57 above). Unfortunately little other publically accessible data is available further other locations along the coast though there is anecdotal evidence that other observations are available and access to these observations would provide a clearer picture of the east African coastal sea level trends. 8 http://www.pmsl.org P a g e | 42 TRENDS IN SALINITY AND PH As mentioned above, observations of salinity and pH are typically sourced from research cruise observations that occur intermittently and typically cover different areas of the ocean. However attempts have been made to describe recent changes in both salinity and pH and these are summarised in the IPPC AR4. Key figures from this report are extracted here. Figure X below shows various time series of ocean CO 2 concentrations and ocean pH levels for a number of ocean basins showing a general increase in dissolved CO 2 concentrations and a decrease in pH levels across the major ocean basins over the past 15 years. Figure X below shows an estimate of the anthropogenically driven uptake of carbon in the ocean highlighting how different ocean basins and regions have taken up CO2 from the atmosphere. CO2 uptake occurs the easiest in areas of high wind speed where strong surface mixing of both the ocean and atmosphere facilitate the rapid absorption of atmospheric CO2. CO2 uptake is strongly linked to ocean pH as can be seen in figure X above. This data should not however be used as a proxy for local measurements of ocean acidification. Figure 60: Changes in surface oceanic pCO2 (left; in μatm) and pH (right) from three time series stations: Blue: European Station for Timeseries in the Indian Ocean (ESTOC, 29°N, 15°W; Gonzalez-Dávila et al., 2003); green: Hawaii Ocean Time-Series (HOT, 23°N, 158°W; Dore et al., 2003); red: Bermuda Atlantic Time-series Study (BATS, 31/32°N, 64°W; Bates et al., 2002; Gruber et al., 2002). Values of pCO2 and pH were calculated from DIC and alkalinity at HOT and BATS; pH was directly measured at ESTOC and pCO2 was calculated from pH and alkalinity. The mean seasonal cycle was removed from all data. The thick black line is smoothed and does not contain variability less than 0.5 years period. P a g e | 43 Figure 61: Figure 5.10. Column inventory of anthropogenic carbon (mol m–2) as of 1994 from Sabine et al. (2004b). Anthropogenic carbon is estimated indirectly by correcting the measured DIC for the contributions of organic matter decomposition and dissolution of carbonate minerals, and taking into account the DIC concentration the water had in the pre-industrial ocean when it was last in contact with the atmosphere. The global inventory of anthropogenic carbon taken up by the ocean between 1750 and 1994 is estimated to be 118 ± 19 GtC. P a g e | 44 CLIMATE VARIABILITY AND CHANGE INTRODUCTION Regional climate is always a function of a number of climate drivers. At the local scale, local scale climate drivers are important. These include constants such as regional topography and water bodies, and more dynamic drivers such as soil moisture and vegetation. More distant drivers include regional climate systems such as the major high and low pressure systems and their positional shifts and changes in magnitudes. Regional sea surface temperatures (SSTs) are also important drivers as the regional ocean is often a critical source of moisture, even for inland locations. Ocean basin scale SST patterns significantly impact basin scale circulation patterns which can have important impacts on regional climates. Finally, global scale drivers such as the El Nino Southern Oscillation (ENSO) have strong controls over the climate of many areas of the world. In the east African context, all of the above sources of variability are important. East Africa is characterized by significant topography in many areas, large inland lake systems, and proximity to the Indian Ocean. Indian Ocean climate is strongly controlled by SST patterns such as the Indian Ocean Dipole (IOD) as well as the influence of ENSO. In the context of the WWF CEAI regions, most of the priority landscapes identified fall within the tropical or equatorial climate with some influence of sub-tropical climates in the most southern landscapes of Mozambique. THE SEASONAL CYCLE The most dominant and well known form of climate variability is the seasonal cycle. This is generally driven by shifts and changes in the regional pressure systems. In east Africa the most critical seasonal shift is the north south movement of the Inter-tropical Convergence Zone (ITCZ). The ITCZ is a line of strong moisture convergence driven strong solar heating and advection of moisture towards the equator from the sub-tropics (see more info below). The ITCZ extends as a belt from east to west across central Africa. During some parts of the year the ITCZ can split into two belts. The ITCZ migrates north and south through the year with its northward position occurring during austral winter and its southern position during austral summer. Critically, the ITCZ brings strong convective rains to most parts of central tropical Africa including many areas of the east African region. Inter-tropical Convergence Zone - ITCZ The ITCZ is the dominant source of intra-annual variability in tropical Africa including large parts of the east African region. The ITCZ occurs most coherently over the tropical oceans but also extends over several tropical continental areas including Africa. The ITCZ is a roughly continuous zone of strong low level moisture convergence, uplift and convection and is associated with rainfall across the tropics. The ITCZ is a central feature of the Hadley circulation. The Hadley cell describes the vertical circulation cell that forms due to strong uplift around the equator and the corresponding zone of subsiding air in the subtropics. The zone of subsiding air forms the large sub-tropical high pressure cells that are found around the globe. Air is convected to high levels in the equatorial regions due to strong surface heating and latent release. Convection causes precipitation which removes moisture from the air. This air then moves poleward at higher levels and progressively cools. As it cools it begins to descend around the latitude of the sub-tropics. This descending air produces surface high pressures and very stable, dry, clear sky conditions. The Hadley circulation therefore plays a critical role in the development of the strong rainfall gradient from the wet tropics through to the much drier sub-tropics. P a g e | 45 While the ITCZ is generally observed as a continuous and homogenous belt of convection over the oceans, the influence of the land surface and topography over the continents often results in splits and breaks in the continuity of convergence zone. The positioning and strength of the ITCZ at various times of the years is also governed not only by the seasonal cycle but also by regional dynamics such as moisture transport, land surface dynamics and large scale tele-connections. In areas close to the equator a bi-modal season is experienced. These areas typically experience two rainy seasons during the summer with a short dry season between. This is a result of the movement of the ITCZ. During spring the ITCZ will migrate south and bring early summer rains to the region. These rains are often smaller and shorter in duration and hence are named the “short rains”. The ITCZ will then migrate even further south, as the peak of summer approaches, resulting in a short period of drier conditions for areas near the equator. As autumn approaches the ITCZ begins to migrate northwards again and once again brings a late summer rainfall. These second rains are longer and are commonly known as the “long rains”. The result is a bi-modal rainfall seasonality. For areas further from the equator this does not occur as the ITCZ does not move south of these areas during the peak of summer. The result is a more typical uni-modal, single summer rainfall season. The bi-modal seasons are often relied upon by society as they allow for two planting and harvesting seasons during the year. The failure of one of the seasons or the lack of a short dry season can have large impacts on agriculture in a region. Another important aspect of the intra-annual seasonality in the east African region is the interaction between temperature and rainfall. The typical cycle is that as spring arrives temperatures begin to increase with highest temperatures occurring during late spring. The arrival of the rains typically results in a reduction in the daily maximum temperatures both due to the increase cloud cover but also due to latent heat of evaporation from a moist surface. Mid summer temperatures are therefore typically cooler than late spring. As the rains subside temperatures sometimes will increase but typically the rains will continue into autumn by which time temperatures will start to decrease. Spring is therefore a critical time in this area as increasing temperatures after the dry season quickly evaporate what water or soil moisture might exist. Late arrival of the rains can have a significant impact on the regional ecology, agriculture and society. Besides the seasonal cycle, a significant further source of intra-annual variability is the Madden-Julian Oscillation (MJO) which operates on a period of 30-60 days. The MJO consists of a series of eastward migrating centers of both enhanced and suppressed convection that travel along the equator across both the Indian Ocean and the Pacific. The MJO has been shown to have strong influences on the timing of rainfall onset in east Africa as well as the magnitude of the late rainfall season (March – May) 9. INTER-ANNUAL VARIABILITY While the intra-annual seasonality is the most dominant and obvious expression of climate variability, inter-annual variability (variations between years) is of the cause of the most significant climate events in this region including severe droughts, late onset of rains and floods. Inter-annual variability is particularly high in this region with the variation in annual total rainfall being of the order of 100% of the median rainfall. Drought years are fairly common, as are years of heavy rainfall and flooding. Current understanding of inter-annual variability in the east African region suggests that ENSO is a very important driver while the more local Indian Ocean dynamics including the IOD may also play a significant role. Figure [x] illustrates the correlations between east African rainfall (areas 2 and 3) and both ENSO and IOD. Pohl, B. and Camberlin, P. (2006), Influence of the Madden–Julian Oscillation on East African rainfall: II. March–May season extremes and interannual variability. Quarterly Journal of the Royal Meteorological Society, 132: 2541–2558. 9 P a g e | 46 More detail: Indian Ocean Dipole The Indian Ocean Dipole (IOD) describes an ENSO like SST gradient across the Indian Ocean and was first described by Saji (1999). The IOD operates in a two phases. During the positive phase, SSTs in the western Indian Ocean are warmer with corresponding cooler waters in the east. This seems to result in drier conditions in the adjacent western land masses of northern Australia and Indonesia and wetter conditions in east Africa. The negative phase is reversed with cooler waters off African and warmer waters in the eastern ocean resulting in drier conditions in east Africa and wetter conditions in the east. Of much ongoing debate is the possible linkage between IOD and ENSO. While initial studies (Saji etc…) proposed the IOD as operating independently from ENSO, later work has challenged this (refs) and suggest instead a lagged relationship between ENSO and an IOD response with the IOD responding to ENSO over a 5 to 7 month lag period. There is much ongoing research exploring IOD and IOD ENSO linkages as well as the complexity of IOD impacts in the east African region 10 11 The role of ENSO in the east African climate has been the subject of much research 12 13 14 15 16. East African rainfall response has been shown to be fairly consistently wetter during the ENSO warm phase (El Nino) for Tanzania and Kenya and drier further south in Mozambique. The response is almost predominantly experienced during the early short rains (October – December) rather than the later rains or the later part of the rainfall season (March - May). The opposite response occurs during the cold phase (La Nina) with drier conditions in Tanzania and Kenya and wetter conditions in southern Mozambique. Of critical interest however is the consistency of the regional response. For east Africa the direction of the response (wetter or drier) seems to be fairly consistent but the magnitude of the response does not seem to be linearly related to the magnitude of the ENSO warming or cooling. In particular it has been noted that the strong correlation between east African rainfall and both ENSO and IOD broke down almost completely for a ten year period between 1983 and 1993 and has since at least partially restored17. Additionally, local scale responses can be quite heterogeneous with not all regions experiencing the same relationship with either ENSO or IOD. These aspects all have important consequences for seasonal forecasting. ENSO SST patterns are now routinely forecast several months in advance with significant reliability [cite sources] however with the large uncertainty around the regional response comes large uncertainty and lack of confidence in seasonal forecasts derived from these SST forecasts whether dynamic or statistical. D. Manatsa, C. H. Matarira, G. Mukwada, Relative impacts of ENSO and Indian Ocean dipole/zonal mode on east SADC rainfall, International Journal of Climatology, 2011, 31, 4 11 Richard Washington, Anthony Preston, Extreme wet years over southern Africa: Role of Indian Ocean sea surface temperatures, Journal of Geophysical Research, 2006, 111, D15 12 Indeje, M., Semazzi, F.H.M. & Ogallo, L.J. ENSO signals in East African rainfall seasons. International Journal of Climatology 20, 19-46 (2000). 10 13 Evolution of ENSO-Related Rainfall Anomalies in East Asia, Renguang Wu, Zeng-Zhen Hu, Ben P Kirtman in Journal of Climate (2003) 14 NICHOLSON, S. E. and KIM, J. (1997), THE RELATIONSHIP OF THE EL NIÑO–SOUTHERN OSCILLATION TO AFRICAN RAINFALL. International Journal of Climatology, 17: 117–135. 15 Agnes L. Kijazi, C. J. C. Reason, Intra-seasonal variability over the northeastern highlands of Tanzania, International Journal of Climatology, 2012, 32, 2 16 Alessandra Giannini, Michela Biasutti, Isaac M. Held, Adam H. Sobel, A global perspective on African climate, Climatic Change, 2008, 90, 4, 359 17 Clark, Christina Oelfke, Peter J. Webster, Julia E. Cole, 2003: Interdecadal Variability of the Relationship between the Indian Ocean Zonal Mode and East African Coastal Rainfall Anomalies. J. Climate, 16, 548–554. P a g e | 47 More detail: El Nino Southern Oscillation The ENSO phenomenon was first identified in 1923 by Sir Gilbert Walker through an analysis of pressure oscillations over many years across the Pacific Ocean basin. However it wasn’t until 1969 that the dynamics of the relationship between the observed pressure gradients and the period occurrence of warm waters off the coast of Chile, locally known as La Nina, were established Bjerknes 1979 (WMR) The Southern Oscillation therefore describes the period occurrence of a warm pool of water in the eastern Pacific Ocean. This is the so called “warm” phase of the oscillation. The “cool” phase is observed as a general cooling of SSTs across the whole Pacific basin. The dominant effect of ENSO is in the global tropics and the established mechanism is through a longitudinal shift in the positioning of the global, tropical, walker circulations. For east Africa, and ENSO warm event therefore manifests as an increase in rainfall as the locus of convection linked to the Indian Ocean walker cell moves west over the continent. Conversely, a cool event results in a shift in the locus of uplift to the east resulting in a drying over east Africa. It must be noted that the opposite response occurs further south in southern Mozambique and South Africa where mid-latitude dynamics come into play. ENSO also impacts the sub-tropics and mid-latitudes through a moderation of the Hadley circulation and resultant impacts on sub-tropical convergence zones 18. However ENSO impacts in the sub-tropics and mid-latitudes are far less well understood than tropical impacts and appear to be far less consistent, not only in magnitude of response, but also in the sign. One final aspect of regional climate variability in tropical Africa is that of persistence of trends and inter-seasonal persistence. There are a number of suggestions that in moist tropical climates the moist land surface itself can become a climate driver both as a source of moisture as well as moderating the land surface through uptake of latent heat of evaporation. While such local feedbacks are complex and difficult to diagnose, statistical evidence suggests that wet and dry periods tend to persist longer than would be expected given external drivers alone. This suggests that local soil moisture can become a local source of moisture persisting regional rainfall while conversely a lack of moisture can cause a persistence of a dry period. Whether this persistence occurs only intra-seasonally or inter-seasonally as well remains an open question 19 CLIMATE CHANGE Climate variability often refers to only those sources of variability that are cyclical or periodic in some way while climate change refers variations that are not obviously periodic but are instead long term trends of a more or less constant sign. Long term climate changes have occurred throughout the observable climate history as if evidenced by a number of proxy indicators such as ice cores. As such long term climate changes are a natural part of the climate system. Drivers of such changes include variations in solar output, volcanic eruptions, orbital variations and complex ice albedo feedbacks. Paleo-climate science is the study of such climate variations and is an area of much ongoing study. Of more topical concern however is anthropogenic climate change or climate change resulting from human activities. Anthropogenic warming was first predicted near the end of the 19th century as the radiative properties of various gases including CO2 were being explored. Even at that point it was surmised that human activities would produce a warmer climate. 18 Cook, Kerry H., 2000: The South Indian Convergence Zone and Interannual Rainfall Variability over Southern Africa. J. Climate, 13, 3789–3804. 19 Nicholson, S. & Entekhabi, D. The quasi-periodic behavior of rainfall variability in Africa and its relationship to the southern oscillation. Archives for Meteorology Geophysics and Bioclimatology 34, 311-348 (1986). P a g e | 48 Much of the focus of climate change work and discussion around climate change has focussed on green house gases (GHG). These are a class of gases that interact strongly with solar radiation, in particularly, long wave or infrared radiation that is emitted from the warm surface of the earth or oceans. These gases are often rated by the radiative forcing. Radiative forcing is an indication of the degree to which a gas absorbs long wave radiation, more specifically though it indicates the warming (or cooling) effect a gas has on the earth’s temperature. Radiative forcing is generally expressed as the change in energy balance at the top of the atmosphere in units of W/m 2. While CO2 is the GHG most discussed, other such gases include methane (NH4) and water vapour (H20). CO2 is clearly a naturally occurring gas and is the result of respiration in both plants and animals. However there has been a clear and rapid increase in the concentration of CO 2 in the atmosphere (Figure 62 belowError! Reference source not found.) and this has been attributed to human activities. NH4 is also a naturally occurring gas and is generally the result of decomposition of organic matter. It is quite likely however that the production of NH4 is being accelerated through warming of areas such as the arctic tundra allowing permanently frozen ground to begin decomposition processes. NH4 actually has a higher specific radiative forcing than CO 2 . Figure 62: Atmospheric concentrations of important long-lived greenhouse gases over the last 2,000 years. Increases since about 1750 are attributed to human activities in the industrial era. Concentration units are parts per million (ppm) or parts per billion (ppb), indicating the number of molecules of the greenhouse gas per million or billion air molecules, respectively, in an atmospheric sample. (source: IPCC Fourth Assessment Report, 2007) More recently, much attention has also begun to focus on the role of high clouds and aerosols as sources of negative radiative forcing. The presence of these in the atmosphere results in a cooling which counters the warming of GHGs. Aerosols are sourced from two main sources, volcanic eruptions and human activities. While the current consensus science concludes that GHGs still have a strongly dominant effect and hence are producing a climate warming, the role of high clouds and aerosols as sources of cooling are under intense scrutiny. Figure 63 below indicates the relatively radiative forcing of each of the GHGs as well as aerosols. This figure also indicates the uncertainty around the radiative forcing and as such highlights our scientist’s as yet incomplete understanding of this process. However, there is extremely strong consensus and mounting empirical evidence that the P a g e | 49 climate is warming faster than it has even warmed before and this forms the basis of the intense concern over the potential or even currently occurring impacts of such changes. Figure 63: Summary of the principle components of radiative forcing of climate change (source IPPC, Fourth Assessment Report, 2007). Error bars indicate an estimate of the uncertainty of the radiative forcing for each term. CLIMATE CHANGE IN THE CEAI REGION While there is strong evidence and scientific consensus that global temperatures are increasing, the critical questions around climate change are those of regional responses and impacts. Certainly different regions of the world are already experiencing the impacts of climate change in different ways whether through changes in rainfall, temperatures, sea levels or winds. However many of these observed changes are difficult to associate directly with global warming as other sources of variability, such as those described above, can also influence the regional climate over periods of decades or longer. The challenge therefore is to project the likely impacts of climate change at the regional scale while being fully cognisant of the large uncertainties and often relatively poor understanding of regional climate variability. The great majority of the area under scrutiny in this report falls within the tropical rainfall climate regime. One of the dominant effects of global warming is an increase in the water vapour content of the atmosphere potentially resulting in a more energetic hydrological cycle. This expresses itself as more and heavier rainfall. Indeed, for most of the east African region the latest GCM based projections largely agree on wetter conditions for the area. This is however a simplistic view of the regional impacts. While a higher level of moisture in the atmosphere could indeed reasonably produce wetter conditions in the tropics, other regional dynamics could also be affected. Another important response to a warmer climate is strengthening of the Hadley circulation that drives the development of the sub-tropical high pressure belts. Stronger and more extensive high pressure systems such as the South Indian High P a g e | 50 could shift mid-latitude storm tracks further south, impacting sub-tropical and mid-latitude climates and also modifying the development and north-south migration of the ITCZ over Africa. The possibility therefore exists for later onset of rainfall in tropical regions, more frequent failure of the rainy seasons, and drier conditions further from the equator in areas more dominated by the sub-tropical high pressure systems. Some of these changes have already been anecdotally reported for the east African region and many of the latest dynamic GCM projections indicate similar responses are likely under future climate conditions. GCM projections of future climate at the regional scale are however difficult to apply robustly and as a result various climate downscaling methods are often applied to develop more regionally relevant information. These will be addressed in the following section on regional projections for the CEAI area. CLIMATE CHANGE AND NATURAL VARIABILITY Another area of great interest is the interaction between global warming and natural sources of climate variability such as ENSO and the IOD. Research in this area is largely based on GCM experiments though there are some attempts to understand recent observed changes in ENSO and IOD in the context of a warming climate20. There has been some modelling work that has suggested that climate warming will result in a shift towards a more constant ENSO warm phase in the future potentially resulting in more permanent warm phase impacts globally. In the east African context this would result in wetter conditions as more normal along with the potential for more frequent extreme wet periods or floods. As regards the IOD, there has been less work in this regard though once again some suggestions that a more permanent warm western ocean state could emerge in the future again potentially increasing rainfall in the region. It must however be noted that many modern GCMs are incapable of capturing the inter-annual variability of ENSO in a realistic way with some model unable to even capture the basic ENSO signal at all. This seriously limits any confidence in projections of ENSO dynamics in the future. Much work is however continuing in this area, due to the critical relevance to any understanding of future regional climate impacts. Of course of great importance is the relative magnitude of natural variability versus climate change. As has already been highlighted, natural variability on the inter-annual and decadal time periods is large. There is much debate around the importance or relevance of climate change within the context of large climate variability. The critical aspect in this debate is actually the occurrence of extreme events and how climate change changes the frequency of exceedence of critical thresholds. 20 Diaz, H. F., Hoerling, M. P. and Eischeid, J. K. (2001), ENSO variability, teleconnections and climate change. International Journal of Climatology, 21: 1845–1862. doi: 10.1002/joc.631 P a g e | 51 CLIMATE AND OCEAN PROJECTIONS CLIMATE AND OCEAN PROJECTIONS FOR PRIORITY LANDSCAPES This section describes climate and ocean projections for the CEAI countries with a specific detailed focus on the CEAI priority landscapes. A number of sources of data have been included. For the general country level climate projections, CMIP3 Global Climate Model projections have been provided. For the priority landscape projections, statistically downscaled station scale projections have been used. A description of the statistical downscaling methodology is included in the relevant section below. The final component of projections is a brief overview of projected sea-level rise and sea surface temperatures. This section is very general and fairly limited in scope due to a lack of specific regional data on these topics. RAW GCM PROJECTIONS Figure 49 and 50 below present the future anomies in precipitation and mean surface temperature projected by 10 GMCs for the mid 21st Century (the 2046-2065 minus 1961-2000 seasonal climatologies). From the multi-model set of GCMs the lower 25th percentile, the median and the upper 75th percentile are used to indicate the envelope of change projected by the different models. The spread of change projected by the different models is clearly evident by comparing the results for the lower and upper percentile. However, these differences do not substantively alter the climate change messages. Results are presented as seasonal anomalies: December - February (DJF), March - May (MAM), June – August (JJA) and September – November (SON). The projected change in temperature by the mid 21 st Century is quite consistent between models. Maximum warming is projected over southern African interior, with the smallest change located over the ocean. The austral spring (September – November) is projected to warm more than the other seasons while austral summer shows the smallest increase in temperature. The projected change in monthly total precipitation is less consistent than temperature, with models disagreeing on both the amplitude and sign of change over some areas. With this in mind, a few messages can be outlined: Precipitation over the northern parts is projected to increase into the future, while the southern parts show generally decreasing precipitation. Seasonally, there is a shift in the location of projected drying and wetting over the region with the increase in precipitation following the location of the ITCS. During austral summer most of the models project increased precipitation over the study area, with only the 25th percentile showing drying over the far south of the study area. During March to May the region shifts north and generally no change projected during the dry season (austral winter). No change or a slight drying is projected over most of the study area during September – November, with the exception of central Kenya where the models project increased precipitation. These results suggest that the study area may be located within the spatial boundary between regions experiencing wetting vs. drying change and the placement of this boundary is not consistent between models. P a g e | 52 Figure 64: Projected change in average surface air temperature (ºC) from 10 GCMs of the CMIP-3 archive. Anomalies are the difference between 2046-2065 and 1961-2000, based on the SRES A2 emissions scenario. The columns are for each 3 month season of the year (DJF, MAM, JJA, SON). The upper row shows the 75th percentile of the model range, the middle row is the median, and the lower row the 25th percentile. P a g e | 53 Figure 65: Projected change in average monthly total precipitation (mm/month) from 10 GCMs of the CMIP-3 archive. Anomalies are the difference between 2046-2065 and 1961-2000, based on the SRES A2 emissions scenario. The columns are for each 3 month season of the year (DJF, MAM, JJA, SON). The upper row shows the 75th percentile of the model range, the middle row is the median, and the lower row the 25th percentile. P a g e | 54 AN OVERVIEW OF DOWNSCALING Downscaling is the process of producing climate or other projections at a finer spatial scale than GCM projections. Two main categories of downscaling exist, namely dynamic downscaling and statistical/empirical downscaling Dynamical downscaling deploys regional climate models. These models are very similar to global climate models in that they attempt to simulate the fundamental atmospheric physics. The regional models are however configured to simulate the atmosphere over a limited spatial area but at much higher spatial resolution (typically 50km or less). It is for this reason that they are sometimes referred to as “Limited Area Models”. Regional models require information about the state of the climate system beyond their spatial borders and for this GCM model data is used. The expression is for this is “forcing”. Regional models are forced at their boundaries by a GCM. As a result, the regional model largely reproduces the same climate dynamics as the GCM at the large scale while being able resolve some of the finer scale detail within the limited spatial area. Examples of such regional models are PRECIS (Hadley Center), WRF (NCAR) and RegCM (ICTP). Statistical or empirical downscaling is quite different. These methods are based on the underlying assumption that the weather or climate at the local scale is somehow related to the weather or climate at the large scale. Therefore it is possible to statistically model this relationship by using past empirical observed data. Various statistical methods are used ranging from weather generators to artificial neural networks and self organising maps (see the SOMD method below) There is much debate in the literature around the advantages and disadvantages of dynamical versus statistical downscaling. Dynamical downscaling would seem to be the superior method because of its basis in the underlying physical processes. This is particularly raised in the context of climate change where empirical models based on historical climate may, arguably, be invalid for future climate. Dynamical models are not fundamentally tied to historical observations in any way and are free to simulate future climate dynamics that may not have been experienced in the past. This issue is often referred to as “stationarity”. However, dynamical models, like their GCM parents, suffer from systematic and non-systematic biases and errors, particularly in their resolving of precipitation processes. Simulations of regional climate by different regional models, forced by identical boundary conditions, can often lead to markedly different representations of the regional rainfall climatology and, more critically, seasonal rainfall characteristics. 21 These biases, while gradually being minimized through model improvement, still remain a serious obstacle to the application of regional models to climate change impacts analysis. Statistical models generally produce far more consistent results and so are often considered more robust. The greatest criticism is around the issue of stationarity and extremes. The issue of stationarity is likely less serious than it may seem. The reason for this is because analysis of the driving GCM models circulation fields suggests that much of the climate change signal at the regional scale is not through a shift of the regional climate into a completely new state or configuration, but rather a change in the frequency of occurrence of particular climate states. For example, in South Africa, much of the change in rainfall seems to be linked to changes in the frequency of Tropical Temperate Troughs which are the major rain producing system in the region. Statistical models are adept at capturing the local scale impacts of such changes. The issue of extreme events links to that of stationarity to some extent. The challenge of extreme events in statistical models is that extreme events are, by definition, rare and hence they are not well represented or captured in many statistical modelling methods. Added to that is the problem that future extreme events may well be as a result of atmospheric states beyond the limits of events that have already been observed. In these cases statistical models are unable to capture such impacts and will under-represent the occurrence or risk of extreme events. 21 On RCM-based projections of change in southern African summer climate, Geophysical research letters [0094-8276] Tadross, M yr:2005 vol:32 P a g e | 55 DOWNSCALING METHDOLOGY - SOMD The statistical downscaling methodology used to produce the projections below is called Self Organising Map Downscaling (SOMD) as it based on a data clustering method called self organising maps developed by Kohenen method is fully documented in the paper by Hewitson and Crane (2006) 23. 22. The Briefly, the method ingests re-analysis (pseudo observed) climate fields such as winds, humidity and geopotential height for a synoptic scale window surrounding a particular observing station. These fields are then classified through the SOM method into a number of arch-type observed climate states. For each arch-type state, a statistical distribution of observing daily station response (rainfall or temperature) is compiled in the form of a cumulative distribution function (CDF) of daily response values. To construct a future projection time series, future synoptic climate states are constructed from a particular GCM output (20th Century or some future period and emissions scenario), for the same spatial window as in the previous step. Each of these states is then mapped to one of the arch-type states previously identified. Once the arch-type state is identified, the associated CDF is randomly sampled in order to obtain the station scale response at the daily time scale. The resulting daily time series forms the basis for the future projection which is then aggregated into monthly anomaly statistics and other useful parameters as presented in the following section. The advantage of such a method is that it does not attempt to draw information from the GCM precipitation or temperature fields at the grid scale which are known to be of limited value. Rather, information is extracted from the relatively robust large scale atmospheric state of wind, moisture and pressure fields. The paper referenced describes how such a method helps to reduce the uncertainty of GCM based climate projections. A limitation of the method, highlighted again in the later section on interpretation, is that the method is grounded in the classification of observed climate states and hence cannot provide information about responses to future potentially unobserved climate states. However, analysis of GCM fields does suggest that much of the impact of climate change on regional climates consists of changes in the frequency of occurrence of currently observed climate states rather a transition into a completely new sequence of climate states. Only in extreme circumstances in areas of steep climatic gradients would it be likely that completely new climate states will be regularly observed in the future. However there are implications regarding the representation of extreme events which is discussed later. Two different SRES 24 emission scenarios are presented in the following projections. The A2 scenario describes a world where individual nations remain fairly independent, there is continuous increased world population and more regionally focussed economics. This scenario represents a moderately high emissions scenario. The B1 scenario describes a more ecologically and environmentally aware world with global integration but a focus on local solutions and clean technologies. The B1 scenario represents the lower end of the emissions scenarios. The use of both A2 and B1 is an attempt to capture some of the uncertainty introduced by different emissions scenarios without dealing with the full spectrum of all 40 SRES scenarios or even the full six scenario families described by SRES. Ten of the CMIP3 25 archive GCMs are used to produce a multi-model suite of projections for each station location. These projections are presented as summary envelopes where the envelope represents the range between the 10 th and 90th percentile of the GCM projections for each month for a particular parameter. The multi-model median projection is also 22 The self-organizing map. Kohonen, Teuvo, Neurocomputing: An International Journal, Vol 21(1-3), Oct 1998, 1-6. 23 Hewitson, B. C. and Crane, R. G. (2006), Consensus between GCM climate change projections with empirical downscaling: precipitation downscaling over South Africa. International Journal of Climatology, 26: 1315–1337 24 Special Report on Emissions Scenarios, IPCC, 2000 - Nebojsa Nakicenovic and Rob Swart (Eds.) Cambridge University Press, UK. pp 570 25 http://www-pcmdi.llnl.gov/ipcc/about_ipcc.php P a g e | 56 represented. The rationale for summarizing the projections in this way is described in more detail in the following section on interpretation of climate projections. Explanation of climate projection plots – Precipitation Top panel - climatologies: Grey bars indicate the 10th to 90th percentile range of the control period multi-model climatologies (1961-2000). Red bars indicate the same but for the future period multi-model projections (2046-2065). Bottom panel – anomalies: Wide bars indicate the 10th and 90th percentile range of the future change (future minus the control) with the median change marked as a solid black line. SUMMARY PROJECTIONS FOR PRIORITY PLACE 1: LAMU SEASCAPE The future projected climate change for the Lamu Seascape has been downscaled to the weather station at Lamu. This station provides precipitation data for the period 1979 – 2000. Figure 66: Change in monthly total rainfall for Lamu station under SRES A2 (left) and SRES B1 (right) emission scenario. The most robust projected change in precipitation is found between June and August where monthly total rainfall is projected to decrease. Models generally project an increase in rain day frequency from November through to April, with most positive change in extreme rainfall occurring in October and November, with a decrease between June to September for all types of rainfall frequency. A decrease in the dry spell duration during the dry season (January – April) is shown for most models suggesting that the period between rain events may increase into the future. P a g e | 57 SUMMARY PROJECTIONS FOR PRIORITY PLACE 2: KWALE – EAST USAMBARA LANDSCAPE The future projected climate change for the Kwale – East Usambara Landscape have been downscaled to the weather stations at Mombasa and Tanga. The Mombasa station provides daily minimum and maximum temperature and precipitation for the period 1957-2003, while the Tanga station only provides daily precipitation data (1979-2000). Figure 67: Change in monthly total rainfall for Mombasa stations under SRES A2 (left) and SRES B1 (right) emission scenario. Figure 68: Change in monthly mean dry spell duration for Mombasa stations under SRES A2 (left) and SRES B1 (right) emission scenario. P a g e | 58 Figure 69: Change in monthly rain day frequency > 0.5 Mombasa stations under SRES A2 (left) and SRES B1 (right) emission scenario. Figure 70: Change in monthly mean maximum daily temperature (deg C) for Mombasa station under SRES A2 (left) and SRES B1 (right) emission scenario. P a g e | 59 Figure 71: Change in monthly mean maximum daily temperature (deg C) for Tanga station under SRES A2 (left) and SRES B1 (right) emission scenario. Figure 72: Change in monthly days exceeding 32 deg C for Mombasa under SRES A2 (left) and SRES B1 (right) emission scenario. The most robust message is for a decrease in rainfall during the short dry season and for a small but robust increase in rainfall from November – February. Dry spell duration is projected to decrease during main dry season though the confidence is quite low. Rainfall frequency is generally projected to decrease between June – September, with a general increase during the rest of the year. Both minimum and maximum temperatures are projected to increase into the future by roughly 2° C under the SRES A2 scenario and by 1.5° C under the SRESB1 scenario. The number of days on which the temperature exceeds 32° C is projected to increase significantly so that most days between December and March exceed this threshold. The largest change is projected to occur during May and November. P a g e | 60 SUMMARY PROJECTIONS FOR PRIORITY PLACE 3: UDZUNGWA LANDSCAPE The future projected climate for the Udzungwa Landscape has been downscaled to the weather station at Iringa. Figure 73: Change in monthly rain day frequency > 0.5 mm Iringa station under SRES A2 (left) and SRES B1 (right) emission scenario. Figure 74: Change in monthly rain day > 10mm frequency for Iringa station under SRES A2 (left) and SRES B1 (right) emission scenario. P a g e | 61 Figure 75: Change in monthly mean dry spell duration for Iringa station under SRES A2 (left) and SRES B1 (right) emission scenario. The downscaled results suggest a slight increase in frequency and intensity of rainfall during the mid to late rainy season (January – March) and a clear increase in the average number of days between rain events during the dry season (May – October). P a g e | 62 SUMMARY PROJECTIONS FOR PRIORITY PLACE 4: RUMAKI AND MATUMBI HILL SEA AND LANDSCAPE The future projected climate change for the Rumaki and Matumbi Hill Sea and Landscape has been downscaled to the weather station at Dar es Salaam, since it is the closest station to the study area. Figure 76: Change in monthly total rainfall for Dar Es Salaam station under SRES A2 (left) and SRES B1 (right) emission scenario. Figure 77: Change in monthly mean dry spell duration for Dar Es Salaam station under SRES A2 (left) and SRES B1 (right) emission scenario. The downscaled results suggest a decrease in rainfall during the main dry season (June – October) and an increase during the rest of the year. Rain day frequency shows the same general tendency and is associated with an increase in dry spell duration during the dry season. P a g e | 63 Figure 78: Change in monthly mean maximum daily temperature (deg C) for Dar Es Salaam station under SRES A2 (left) and SRES B1 (right) emission scenario. Figure 79: Change in monthly mean minimum daily temperature (deg C) for Dar Es Salaam station under SRES A2 (left) and SRES B1 (right) emission scenario. P a g e | 64 Figure 80: Change in monthly days exceeding 32 deg C (days) for Dar Es Salaam Both minimum and maximum temperatures are projected to increase into the future by roughly 2° C under the SRES A2 scenario and by 1.5° C under the SRESB1 scenario. The number of days on which the temperature exceeds 32° C is projected to increase significantly with the largest change projected to occur during October which might be linked to the delay in the onset of the short rains. P a g e | 65 SUMMARY PROJECTIONS FOR PRIORITY PLACE 5: GREATER RUVUMA LANDSCAPE The Greater Ruvuma Landscape covers a large geographic and does not represent a unified climatic zone. The future projected climate change for this region has been downscaled to the weather stations at Lichinga and Songea, since these are the only two stations which met the data quality standards. These stations are located in the more mountainous terrain in the far west of the study area and therefore do not provide the most representative climate for the whole landscape. Figure 81: Change in monthly total rainfall for Lichinga stations under SRES A2 (left) and SRES B1 (right) emission scenario. Figure 82: Change in monthly total rainfall for Songea stations under SRES A2 (left) and SRES B1 (right) emission scenario. P a g e | 66 Figure 83: Change in monthly rain day frequency > 0.5 mm Lichinga stations under SRES A2 (left) and SRES B1 (right) emission scenario. Figure 84: Change in monthly rain day frequency > 0.5 mm for Songea stations under SRES A2 (left) and SRES B1 (right) emission scenario. The downscaling results suggest a decrease in the start of the rainy season (October – December). A decrease in the frequency of rain days is projected during the shoulder seasons which is supported by the projected lengthening of the number of days between rain events during winter and spring (especially for the Songea station under the SRESA2 scenario). P a g e | 67 Figure 85: Change in monthly mean maximum daily temperature (deg C) for Lichinga station under SRES A2 (left) and SRES B1 (right) emission scenario. Figure 86: Change in monthly mean maximum daily temperature (deg C) for Songea station under SRES A2 (left) and SRES B1 (right) emission scenario. P a g e | 68 Figure 87: Change in monthly days exceeding 32 deg C for Songea station under SRES A2 (left) and SRES B1 (right) emission scenario. Maximum daily temperatures are projected to increase by between 2 and 2.5°C with the largest warming occurring during the warmest months (October and November). Days where the maximum temperature exceeds 32° C are rare in the current climate but are projected to increase by up to 8 days / month in Songea under the SRESA2 scenario. P a g e | 69 SUMMARY PROJECTIONS FOR PRIORITY PLACE 6: MTWARA – QUIRIMBUS COMPLEX The future projected climate change for projections for the Mtwara – Quirimbus Complex have been downscaled to the weather station at Mtwara. Figure 88: Change in monthly rain day frequency > 0.5 mm Mtwara station under SRES A2 (left) and SRES B1 (right) emission scenario. Figure 89: Change in monthly rain day > 10mm frequency for Mtwara station under SRES A2 (left) and SRES B1 (right) emission scenario. P a g e | 70 Figure 90: Change in monthly mean dry spell duration for Mtwara station under SRES A2 (left) and SRES B1 (right) emission scenario. The downscaling results suggest a decrease in the rainfall frequency and intensity during the start and end of the rainy season, and an increase in the core of the season (January – February). This is associated with an increase in the dry spell length during the dry season (May – October). P a g e | 71 SUMMARY PROJECTIONS FOR PRIORITY PLACE 7: PRIMIREAS E SEGUNDAS SEASCAPE The future projected climate change for the Primireas e Segundas Seascape has been downscaled to the weather station at Lumbo. Figure 91: Change in monthly rain day frequency > 0.5 mm Lumbo station under SRES A2 (left) and SRES B1 (right) emission scenario. Figure 92: Change in monthly rain day > 10mm frequency for Lumbo station under SRES A2 (left) and SRES B1 (right) emission scenario. P a g e | 72 Figure 93: Change in monthly mean dry spell duration for Lumbo station under SRES A2 (left) and SRES B1 (right) emission scenario. The downscaling results suggest a decrease in the rainfall frequency and intensity at the start of the rainy season and an increase during the core of the season (January – March). The length of time between rain events is projected to decrease during the first half of the dry season but increase toward the end of the season (October). P a g e | 73 SUMMARY PROJECTIONS FOR PRIORITY PLACE 8: DELTA OF THE ZAMBEZI RIVER AND MARROMEU COMPLEX The future projected climate change for the Delta of the Zambezi River and Marromeu Complex has been downscaled to the weather stations at Quelimane and Beira. Figure 94: Change in monthly total rainfall for Quelimane stations under SRES A2 (left) and SRES B1 (right) emission scenario. Figure 95: Change in monthly total rainfall for Beira stations under SRES A2 (left) and SRES B1 (right) emission scenario. P a g e | 74 Figure 96: Change in monthly rain day frequency > 0.5 mm Quelimane stations under SRES A2 (left) and SRES B1 (right) emission scenario. Figure 97: Change in monthly rain day frequency > 0.5 mm for Beira stations under SRES A2 (left) and SRES B1 (right) emission scenario. The downscaling results for this region are quite messy, with only a suggestion of a decrease in rainfall frequency and intensity during October and November and surprisingly to an increase during July. The frequency of light rain events is projected to increase during May to July. P a g e | 75 Figure 98: Change in monthly mean maximum daily temperature (deg C) for Quelimane station under SRES A2 (left) and SRES B1 (right) emission scenario. Figure 99: Change in monthly mean minimum daily temperature (deg C) for Quelimane station under SRES A2 (left) and SRES B1 (right) emission scenario. P a g e | 76 Figure 100: Change in monthly days exceeding 32 deg C for Quelimane station under SRES A2 (left) and SRES B1 (right) emission scenario. Both minimum and maximum temperatures are projected to increase into the future by roughly 2° C under the SRES A2 scenario and by 1.5° C under the SRESB1 scenario. The number of days on which the temperature exceeds 32° C is projected to increase significantly especially from January to March. P a g e | 77 SUMMARY PROJECTIONS FOR PRIORITY PLACE 9: BAZARUTO SEASCAPE The future projected climate change for the Bazaruto Seascape has been downscaled to the weather station record at Vilanculos. Figure 101: Change in monthly total rainfall for Vilanculos station under SRES A2 (left) and SRES B1 (right) emission scenario. The downscaling results for this region are quite messy with models disagreeing on the sign of change for all variables and months of the year. This disagreement prevents any robust climate change messages to be formed. PROJECTED OCEAN CHANGES Data on projected ocean parameters such as sea-level, sea surface temperatures, salinity and pH are far scarcer and less certain that projections of climate parameters. This is because of the difficult in simulating the complex water balance of the planet which includes important components such as land and sea ice. In addition the complexity of the carbon cycle and its representation in climate models produces very high levels of uncertainty around the uptake of CO2 by the oceans and hence ocean pH. Given the uncertainty around these parameters it would seem pragmatic to use current trends as an estimate of future trends, at least for the near to medium term of 10 to 20 years. PROJECTED SEA LEVEL RISE Projections of sea level rise are for continued increases in global mean sea-level with important regional variations as described in the earlier section on observed sea-level rise. Projections range from between 0.2m to 0.5m increase by the end of the 21st century. Regional expressions or variations are extremely uncertain. P a g e | 78 GUIDANCE ON THE USE AND APPLICATION OF DOWNSCALED CLIMATE PROJECTIONS INTRODUCTION Downscaled climate projections offer potentially more relevant information regarding the range of possible future climate states for a given location than a GCM could provide. However there are a number of important concepts and caveats that must be considered when using such projections. DOWNSCALING EXTREME EVENTS Statistical downscaling generally, and in particular the SOMD method used in this report, is based on observed records. The statistical model is constructed based on observations of historical local station responses to large scale climate states. For rainfall, this means that the model, when driven by GCM projected future climate states, cannot produce daily rainfall values outside of the range of the observed records. This only impacts the analysis of extreme events but is nevertheless important to consider. The recommended method of exploring extreme events is therefore to consider the frequency of exceedence of some threshold that does fall within the observed range. A common method is to calculate the frequency of exceedence of the 95% percentile of the observed distribution. The downscaling model is far more capable of robustly projecting such exceedence statistics than projecting actual absolute extreme daily rainfall values. Downscaled temperature projections are dealt with differently in that the SOMD method directly applies the GCMs projected mean temperature increase to the downscaled projections. This is necessary as circulation state fields alone will not produce climate warming in the statistical model. This means that the SOMD model can produce daily temperatures outside of the range of observations. However, in reality, the same caveats apply to analysis of extreme values and hence the use of threshold exceedence is recommended. Of course, longer term extreme events such as very wet seasons or droughts are not affected by such model constraints as these events are generally a result of series of daily events that do fall within the range of observations. UNCERTAINTY AND PROJECTIONS Perhaps the biggest challenge to the application and interpretation of climate projections, whether GCM output or downscaled, is the issue of uncertainty. From the projections presented above it is clear that for many variables there is large disagreement between the various projections derived from different GCMs. The disagreement is presented as an envelope or range of projections. In this report the range is reported as the range between the 10 th percentile and the 90th percentile of the suite of projections for each month of the year. For 10 models this will exclude the two models on each extreme. Such a range of projections can often present serious problems for impacts analysis, particularly if the range crosses the zero line indicating the possibility of both a positive or negative change. A common approach is to only consider the median of the envelope. This approach is difficult to defend as it does not capture uncertainty in any way but rather just ignores it. Likewise, considering the extreme models is also less robust these models are often presenting changes that are not physically robust due to model biases or errors. Considering the range between the 10 th and 90th percentile of the multi-model projections excludes extreme models but does not disregard the reality of uncertainty. CONFIDENCE IN PROJECTIONS Inevitably, following the question of uncertainty is one of confidence. The envelope range of projections presented in this report attempts to represent the range of possible future climate states given our uncertainties. However, in reality, this envelop represents the range of projections given uncertainty in the climate models. However there are a large range of uncertainties feeding into future projections and model uncertainty is only one component of this. A full discussion on P a g e | 79 uncertainty, likelihood and confidence is beyond the scope of this report. However, the current general approach is that a very wide spread in the projections suggests that there is greater uncertainty in the projections. However, we cannot conclude from a wide spread of projections that the extreme projections are not possible. Hence the approach of intersecting vulnerability and projection envelopes described in the preceding section. ROBUST DECISION MAKI NG One fairly robust approach to the use of multi-model projection envelopes, as provided in this report, is to consider the decision making process as the starting point. If we consider each identified vulnerability as describing the context of an adaptation decision then it is possible to explore that decision given a projected range of future climate states. For example, if a vulnerability to reduced water supply has been identified then that would predicate a number of possible adaptation measures such as building a dam or reducing water usage. Within each of those possible measures there is a decision point or threshold where the decision to invest in an adaption measure would be taken. If such a threshold can be described, either quantitatively or even just qualitatively, then the position of the threshold within the range of projected changes can be identified. Clearly, if the decision point doesn’t fall within the range of projections then there is little justification for a particular adaptation measure. However, if the decision point falls well within the projected envelope of changes that that measure should be considered. For marginal cases the decision is less clear and further exploration and risk analysis would need to be done to inform the decision. MODEL FILTERING A question that is commonly asked is whether it is possible to select a subset of GCMs that are more accurate or reliable for a particular region. This is an area of much ongoing academic debate. The critical issue in model selection is the metric by which models should be selected. Typically models are rated based on their ability to represent the 20th century climate. However such a metric assumes a certain measure such as representation of seasonality or annual total rainfall which may not be relevant to a particular impacts sector. However, an over-arching issue is that model selection assumes that a model that represents the 20th century climate well will best project a future climate state given GHG forcing. It is argued that a GCM can be tuned to represent the 20th century well but that that tuning constrains or limits the models ability to respond to increased radiative forcing (GHG increases) and hence projected changes could be constrained. It is quite likely that a model that does not fully capture the 20 th climate nuances might nevertheless better capture a physically robust response to radiative forcing. The approach taken in this report is that all models projections should be considered as equally likely and hence all projections are included in the calculation of the projection envelopes. Once exception is mentioned in the projections section where the CSIRO mk3.5 model is excluded as it fails completely to produce rainfall for some of the east African region. This is an extreme enough problem and has such a large impact on the projection envelopes that is considered defensible to remove this model. P a g e | 80 VULNERABILITY AND IMPACTS CLIMATE VULNERABILITY ASSESSMENT Before looking to the future climate projections it is necessary to assess current vulnerability to climate variability. Knowing ones vulnerability to climate variability will provide a base from which to assess future vulnerability and hence adaptation options. This climate vulnerability assessment forms part of the wider climate adaptation process, a schematic of which is outlined below. Figure 102: The structure of the adaptation process (UK Climate Impacts Programme Adaptation Wizard) In this chapter we undertake a preliminary assessment of the vulnerability of a selection of the priority landscapes and then follow on to assess their future vulnerability based on future projections of climate. The objective of this exercise was a rapid overview of impacts of climate variability in the region. The vulnerability information was collected in a template that was distributed to practitioners and partners in each area and was therefore based on a very small sample of vulnerabilities from each priority area. Hence, there was no sectoral focus on vulnerabilities and the vulnerabilities reported in this document are skewed towards impacts on people and communities rather than the natural environment. A qualitative significance of the impact has been ascribed for each of the impacts per priority landscape. This information was drawn from the subjective opinion of the participants who completed the template and has no numerical basis. It is included here as an indicator of the hierarchy ascribed to events by the people in the impacted communities. It must be noted that climate vulnerability is just one of a spectrum of vulnerabilities that may be experienced by a community and planning decisions should not be undertaken based on the climate vulnerability alone. It is recommended that a more comprehensive assessment is undertaken if this information is to be used for planning purposes. A methodology for this is presented in the concluding section. P a g e | 81 What vulnerability is and why knowing ones vulnerability is important: The IPCC definition of vulnerability to climate is “The degree to which a system is susceptible to, or unable to cope with, adverse effects of climate change, including climate variability and extremes. Vulnerability is a function of the character, magnitude, and rate of climate variation to which a system is exposed, its sensitivity, and its adaptive capacity” 26. In essence vulnerability to climate has three dimensions; exposure, sensitivity and adaptive capacity27. The magnitude to which one is vulnerable is variable but almost everything and everyone is vulnerable to global change. Knowing where ones vulnerabilities to climate lie is crucial to knowing where to focus efforts in mitigating future risks. Climate change means that we cannot plan for business as usual anymore. For organisations to account for both the opportunities and threats arising from climate change, the changing climate needs to be taken into account in planning mechanisms. Traditionally climate assessments have tended to undergo a top-down approach, which start with scenarios of future global change in order to assess the future impact to the subject of assessment28. This is regarded as sub-optimal because top-down approaches usually only get half way to an assessment of whom or what is most at risk from climate change and generally provide first order impacts only. Therefore, it is recommended that organisations who wish to undertake an adaptation assessment undergo a bottom-up approach beginning with an assessment of the current system of interest and the factors which influence its vulnerability to current weather and climate. It makes logical sense to base adaptation to climate change on the premise that if one is not adapted to the current climate then it is highly unlikely one will be adapted to the future climate. Hence, a vulnerability assessment should start by assessing ones experiential vulnerability to current climate. This has the advantage that decision-makers are generally dealing with the sorts of information that they are familiar with, so early progress can be made without the need to collate complex and specialist climate and environmental information. By noting the experiential impacts of a climate/weather event, it allows the user to compile a comprehensive database of impacts. Working from experience allows the database to capture direct and indirect impacts of an event that may not have been thought of in a theoretical exercise. It also allows the user to associate costs to an event and assess how well the event was dealt with by the organisation. This, in turn, provides an indication of the underlying capacity of the organisation to cope with further events of the same magnitude. Each event may impact different people, places and sectors very differently so performing a vulnerability assessment has many advantages29. It provides information on where ones vulnerability lies and provides an opportunity to document how an event had varying impacts on different people, sectors and places and how each responded differently to the event. It can lead to an understanding of ‘critical thresholds’ and sensitivities either physically or socially defined which can be used to understand how risks might change under different climate and socio-economic scenarios for the future. In addition cataloguing the consequences of past climate events tends to focus efforts on climate events rather than trends. It allows one to focus in on the most high risk climate variables and time frames of concern when looking to the future Given the inherent uncertainties in planning for the future, understanding the strengths and weaknesses of current socio-economic as well as biophysical structures is useful and tangible information to inform adaptation planning. 26 IPCC, 2001: Climate Change 2001: Synthesis Report. A Contribution of Working Groups I, II, and III to the Third Assessment Report of the Integovernmental Panel on Climate Change [Watson, R.T. and the Core Writing Team (eds.)]. Cambridge University Press, Cambridge, United Kingdom, and New York, NY, USA, 398 pp. 27 Kasperson, J. X., Kasperson, R. E., Turner, B. L., II, Schiller, A. & Hsieh., W.-H. (2003) The Human Dimensions of Global Environmental Change, eds. Diekmann, A., Dietz, T., Jaeger, C. & Rosa, E. S. (MIT Press, Cambridge, MA). 28 Brown, A., Gawith, M., Lonsdale, K. & Pringle, P. (2011) Managing adaptation: linking theory and practice. UK Climate Impacts Programme, Oxford, UK. 29 UKCIP, 2009. A local climate impacts profile: how to do an LCLIP. UKCIP, Oxford. P a g e | 82 If senior management buy in needs to be obtained to take action, a vulnerability assessment can be used as evidence for action, especially as costs of previous events can be incorporated into the case for action It drives preparedness for taking action in the face of a changing climate, while both assessing the risk of threats and taking advantage of the opportunities that may arise from climate change. UNDERTAKING A CLIMATE VULNERABILITY ASSESSMENT: The first step to assessing climate vulnerability is to look back at past weather events and note the event chain reaction that took place around the event. Typically this would occur by noting the ‘event’, the ‘impact’ of the event and the ‘consequence’ of the event. For each event this would require additionally noting the source of information, the date of the event, the location of the event and any other information that is thought important to supplement the events description. For the purposes of this study, a desk based assessment took place whereby WWF practitioners and local partners rapidly gathered and supplied whatever information was readily available to CSAG in a template form. Below, we have summarised the information received and the original versions are attached in the annex. DESCRIPTIVE CLIMATE VULNERABILITY ASSESSMENT FOR PRIORITY LANDSCAPES: WWF PRIORITY PLACE 1: LAMU SEASCAPE (KENYA) Impacts that have been experienced in the Lamu seascape are a result of episodes of drought, high temperatures and strong winds. DRY SPELL DURATION AND DROUGHT: Periods of drought in the region have had the impact that the crops have failed. This has had a series of knock-on effects such as an increase in poverty, a reduction in farming activities and in some cases immigration due to the need to seek income from elsewhere. HIGH TEMPERATURES: Increases in temperature and humidity in the area have led to changes in biodiversity because the sea grass has died leading to a decrease in breeding grounds for fish. The fish habitats and breeding grounds have also been covered by sand. Both of these impacts led to a reduction in fish stocks. WIND: Strong winds have resulted in lowered fisheries productivity because fishing closer inshore has increased. Climate Variable Impact Consequence Opportunities Crop failure Drought/Dry Spells Reduced farming activities Increased poverty Crop failure, immigration, high cost of living Qualitative significance of impact None identified Medium to High P a g e | 83 High Temperatures/ Increase in Heat and Humidity Change in biodiversity Low or no harvest High amounts of dead sea grass washed to the beaches in many areas Fish habitats and breeding sites covered by sand Increased fishing inshore Strong Winds Low fisheries productivity None identified Medium None identified Medium Table: Summary of future climate vulnerability in the Lamu Seascape WWF PRIORITY PLACE 2: KW ALE LANDSCAPE-EAST USAMBARAS LANDSCAPE (KENYA AND TANZANIA) As indicated in the vulnerability assessment, the Kwale landscape is vulnerable to a variety of climate/weather impacts. From documentation of previous events it has been impacted by periods of high temperatures/heat waves, excessive rainfall/flooding, wind and drought. EXCESSIVE / INTENSE RAINFALL: In the case of excessive / intense rainfall, the result has been impacts such as damage to buildings, subsidence, damage to vegetation, changes in biodiversity and damage to infrastructure such as roads, railways and communication networks. These impacts led to severe consequences for the communities affected. For instance, flooding led to the destruction of homes and properties and, in some cases, even loss of life. Transport systems were disrupted due to obstructed roads or damage to railway lines. This usually has knock-on impacts on economic productivity and mobility of the communities. However, there were potential opportunities associated with periods of high rainfall such as increased agricultural production which, in turn, led to increased honey production. DRY SPELL DURATION AND DROUGHT: In contrast to periods of excess rain, too little rain has also been known to have damaging affects within the Kwale area. Drought led to damage to vegetation, fires, pests and power failure, while high temperatures and droughts led to reduced water supply/quality, crop failure and changes in biodiversity. Climate Variable Impact Consequence Opportunities Excessive/Intense Subsidence Damage to vegetation Damage to properties Loss of life Increase in agricultural and honey production Qualitative significance of impact High P a g e | 84 Rainfall Dry spell duration & High Temperatures Changes in Biodiversity Damage to Infrastructure Disruption of transport Damage to vegetation Fires Pests Power failure Reduced water supply/quality Crop failure Changes in biodiversity Famine and malnutrition Agricultural losses Reduced livestock Power rationing Emigration Reduced biodiversity (plant and wildlife loss) Reduced honey production None identified Medium to High A major consequence of these impacts is famine and high levels of malnutrition in the region. This was a result of damaged vegetation and crops due to lack of water, reduced honey production and a reduced number of livestock. Power supply was interrupted and rationed as water supply used in generation of power was reduced. These factors caused a change in lifestyle as communities emigrated away from their home areas in search of employment elsewhere. In terms of the wider ecology, periods of drought caused fires which burnt forests and diminished biodiversity. This was compounded by wildlife and plant deaths due to lack of water. WIND: Excessive wind in the Kwale region, particularly in the April – June 2006 period led to damaged vegetation with the result that there was no honey production at all. Table: Summary of future climate vulnerability in the Kwale landscape WWF PRIORITY PLACE 3: UDZUNGWA LANDSCAPE (TANZANIA) Heavy/intense rainfall as well as periods of low rainfall has impacted on the Udzungwa landscape. HEAVY / INTENSE RAINFALL: The impacts on the Udzungwa landscape as a result of heavy/intense rainfall include a loss of properties, crops and infrastructure. Many houses were entirely destroyed, leading to village members in the Kilombero district having to rely on relatives, local communities and, in the Uchindile village, the district disaster committee. Currently there are still people living in tents in the Kilosa district and a high degree of poverty still exists as a consequence. Property loss also affected communication infrastructures in the Kilosa district, including roads and railways in the area. Heavy rainfall impacted farms and crops in the area. Farmers’ rice seedlings were washed by the floods, as a consequence increased poverty and famine occurred. LOW RAINFALL: On the other end of the spectrum, low rainfall impacted on the level of the Mtera dam in the Kilolo district and, as a consequence, affected the supply of electricity, resulting in further challenges for the local communities. P a g e | 85 Climate Variable Impact Consequence Opportunities Rice seedlings were washed away by floods Loss of properties, crops and infrastructure Increasing poverty of the villages due to losses occurred Homelessness Loss of crops Communication infrastructure damaged Less Water in the Mtera dam Low supply of electricity Heavy/intense rainfall Low Rainfall Qualitative significance of impact None identified Medium to High Table: Summary of future climate vulnerability in the Udzungwa landscape WWF PRIORITY PLACE 4: RUMAKI AND MATUMBI HILL SEA AND LANDSCAPE (TANZANIA) The Rumaki and Matumbi Hill Sea and landscape have been affected by excessive rainfall and flooding, storm events and drought. EXCESSIVE RAINFALL AND FLOODING: Heavy rainfall and subsequent flooding have severely impacted the region through damage to buildings and residences, human mortality and migration, flooding crop farms, mangroves, coconuts and fisheries in RUMAKI and Kilwa. The consequences of such impacts include population displacements caused by flood damage to buildings and residences, resulting in villages being prepared to migrate. Heavy rainfall resulted in a cholera outbreak, local village residents therefore had to rely on government emergency food and medical aids. Famine resulted from the damage caused to crops. It was noted that blockages to certain river tributaries resulted in a re-direction of water flow, causing some agricultural fields to dry with salt water whilst others became flooded. Flooding by fresh water resulted in the death of an extensive area of mangrove trees and the logging of coconut fields where about 10 acres of coconut trees died. Fisheries biodiversity was affected by increased volume of fresh water, as marine organisms moved towards the deep sea for usual metabolism. Soft soils carried by river tributaries during floods resulted in high deposition upon coral reefs and therefore death of the reefs. DROUGHT: In the Tanzanian coastal regions, Cassava seedlings dried off during drought years leading to large crop failure. Severe consequences of such crop failure included famine and mortality, as well as displacement of people in an attempt to find food for their families. Climate Variable Impact Consequence Opportunities Qualitative significance of impact P a g e | 86 Excessive Rainfall and Flooding Damage to buildings, crops, mangroves, coconuts, fisheries, residences and corals Crop failure Drought Storms Damage to residences, crops, coconuts, mangroves and vessels Famine Population displacement Reduced marine fisheries Mortality Death of corals Famine and mortality Displaced population Displaced population Beach erosion None identified High None identified High None identified Medium STORM SURGES: Storms impacted RUMAKI through a large number of consequences. Storm surges such as Tsunami and increased sea wave strength lead to the destruction of trees, crops and coconut fields. Mangrove and tree destruction had a knock on effect in that the land covered by such flora became increasingly barren, lessening the buffering role it played for the islands (land surfaces), as a result such islands had a high submerging risk. Increased waves strength cause large amounts of beach erosion, affecting the village communities in these areas. Table: Summary of future climate vulnerability in the RUMAKI and Matumbi Hill Sea and landscape WWF PRIORITY PLACE 5: GREATER RUVUMA LANDSCAPE (TANZANIA AND MOZAMBIQUE) The Tunduru district has been affected by a large number of events including Excessive rainfall/flooding, drought, strong winds, lightning events and fog/mist/low cloud. Information gathered for this region was sparse. EXCESSIVE RAINFALL AND FLOODING: Excessive rainfall largely affected crops in the area. The consequence of such impacts led to the destruction of a significant (15-25 Hectares) area of paddies. WIND: Windstorms damaged and destroyed a large number of houses in the Tunduru district. This lead to subsidence of such houses thereby making the houses uninhabitable and unsafe. LIGHTNING: A lightning and thunder event impacted the area through damage and destruction to houses. DROUGHT: P a g e | 87 Many households were negatively affected by drought as famine resulted and people went hungry. The drought of January 2006 was particularly severe resulting in 14561 households affected. Another drought event occurred in March 2008 when 2576 households were affected. The consequences of such an impacts lead to a change in lifestyle as the community was forced to act out of the ordinary to sustain themselves. FOG/MIST/LOW CLOUD: Fog/mist and low cloud led to an increase in the number of pests. Significant hectares of crops were destroyed as a consequence of the increased number of pests. Climate Variable Impact Consequence Opportunities Significant hectares of paddy were destroyed Large number of houses were unroofed/ destroyed Large number of houses were destroyed One resident was injured Households were affected Significant hectares of crops were destroyed Crop failure Excessive rainfall/flooding Wind Lightning and thunder Damage to buildings/subsiden ce Damage to buildings/subsiden ce Drought Fog/mist/low cloud Changes in lifestyle Increase in the number of pests Qualitative significance of impact None identified High None identified Medium to High None identified High None identified High None identified Medium to High Table: Summary of future climate vulnerability in the greater Ruvuma Landscape WWF PRIORITY PLACE 6: MTWARA-QUIRIMBAS SEASCAPE (TANZANIA AND MOZAMBIQUE) WIND, HAIL AND STORM EVENTS: The Quirimbas National Park (QNP) is particularly vulnerable to climate/weather related events. The vulnerability assessment revealed significant impacts from storm events where strong winds and heavy rains caused damage to properties and infrastructure. These impacts are also evident on the coast where invasion of sea water caused damage to buildings and subsidence. Agricultural production was affected by storms and hail resulting in lower yields. Wind, hail and storm events led to damage to vegetation, floods, landslides and damage to infrastructure in the Cabo Delgado province. Strong winds damaged property by blowing the roofs off homes and schools and, when combined with a heavy rainfall event, destroyed bridges and roads. Hail is particularly damaging to crops and destroyed crops such as vegetables and fruits. EXCESSIVE / INTENSE RAINFALL: P a g e | 88 A high significance impact within the study region is that of extreme rainfall. This can be characterized as either an intense rainfall period occurring over a short period of time or a prolonged rainfall event where excessive rainfall takes place. Both incidences resulted in flooding and landslides. These impacts led to severe consequences for the communities affected. For instance, flooding led to the destruction of homes and properties. Transport systems were disrupted due to obstructed roads or damage to railway lines. This usually has knock-on impacts on economic productivity and mobility of the communities. Flooding led to consequences on the health and hygiene of a community as sanitation systems were damaged and mobility was reduced. SEA LEVEL CHANGES: The islands of Matemo and Quirimba and coastal areas such as Pemba have shown particular vulnerability to changes in sea level. This occurred after the Indonesian tsunami and on occasions of increases in the level of high tide. These events are not necessarily related to climate but do provide an indication of the impact that an increasing sea level may have on the area so it is important to account for it in the vulnerability assessment. Encroachment of sea water has had the impact of damage to buildings and subsidence in the coastal region. As a consequence, houses in the coastal region were destroyed and families displaced. Agricultural losses were recorded with lower than average yields. The tourism industry was also affected with coastal erosion protection walls that had been built by the hotels destroyed by the encroaching water. The following section will explore the dynamics of the regional climate more generally looking at sources of intra-annual and inter-annual variability and their regional impacts as well as the potential regional climate impacts of anthropogenic climate change. Thereafter will follow a detailed analysis of downscaled climate change projections including projections, where possible, of the variables identified above. Climate Variable Excessive/Intense Rainfall Impact Consequence Opportunities Damage to property Damage to Infrastructure Landslides Flooding Sea Level Rise Damage to buildings Subsidence Entrainment of land by erosion Damage to crops Property and Infrastructure damage Health impacts as sanitation systems fail Communities homeless Damage to crops Tourism affected Qualitative significance of impact None identified Medium to High None identified High Table: Summary of future climate vulnerability in Mtwara IMPACTS OF PROJECTED CHANGES ON PRIORITY LANDSCAPES This section draws on the projections of climate from each of the relevant climate zones to infer future potential impacts, given the information provided on current vulnerability in section 1. Sufficient climate information is available for region 16 to infer a robust assessment of future vulnerability. This section uses general trend information from section 1 and future projection information from section 3 of this report. For a more detailed analysis and interpretation of the observed data and climate projections data per emission scenario, please see these sections. P a g e | 89 WWF PRIORITY PLACE 1: LAMU SEASCAPE (KENYA) DRY SPELL DURATION: The observed trend in dry spell duration shows short dry spell durations throughout the year other than in the dry season (January - March) where there has been a wide range of observed dry spell durations. The general trend in the future projections is towards decreasing dry spell length by 2050s, however there are some months where the uncertainty range does expand into area of increasing dry spell length. On the whole, the projections suggest that dry spell duration may not be a significant consideration for the Lamu Seascape so the impacts associated with drought may not worsen into the future. HIGH TEMPERATURES: There is no observed or future climatological information on temperature or wind for this station. However, the general trend under climate change is for temperatures increases. Impacts from high temperatures in this region are exhibited in general changes in biodiversity. Explicitly this has been seen in the loss of fish habitats both through sand covering breeding sites and sea grass dying off. WWF PRIORITY PLACE 2: KW ALE LANDSCAPE-EAST USAMBARAS LANDSCAPE (KENYA AND TANZANIA) EXCESSIVE AND INTENSE RAINFALL: Both the observed trends and future projections (2050s) of monthly total rainfall show a general trend towards increasing rainfall during the wet season months (November to February) and decreasing rainfall during the dry season months (June – September). Similarly the frequency of rain days is generally projected to increase in the wet season months and decrease during the dry season months. The number of intense rainfall days (rain days above the 90 th percentile of the observed record) also shows a trend towards a general increase, particularly in the wet season months, by the 2050s. Impacts from excessive and intense rainfall events were ranked, in the vulnerability assessment, as one of the highest significance impacts due to the high level of disruption this causes the community. The future projections suggest an increase in these impacts, particularly in the wet season months. These impacts include damage to buildings, subsidence, damage to vegetation, changes in biodiversity and damage to infrastructure. The consequence of these impacts is potentially very damaging with the most prevalent being flooding and damage to properties leading to potential loss of life. This is coupled with potential wide-spread damage to infrastructure. DRY SPELL DURATION: For the Mombasa station the dry spell duration has shown a trend towards shorter dry spell lengths, apart from the months of February and March. The future projections show a similar trend with dry spell lengths generally projected to become shorter apart from some of the dry season months where dry spell lengths may increase, although the uncertainty ranges in the projections straddle both the positive and negative indices. Even though the duration of dry spells may become shorter there is no indication whether there will be more/less dry spell events so one cannot rule out the possibility of future vulnerability to dry spells. Future dry spells may lead to droughts which will have a number of impacts such as; damage to vegetation, fires, pests, power failure, reduced water supply/quality, crop failure and changes in biodiversity. These impacts lead to famine, P a g e | 90 potential agricultural losses, reduced livestock, power shortages, biodiversity impacts and potential major changes in the lifestyles of the community such as emigration to other regions. HIGH TEMPERATURES / HEATW AVES: Maximum temperature, minimum temperature and days > 32 degrees are all projected to increase in the future projections. Therefore impacts due to high temperature conditions are projected to become worse. These impacts include crop failure, reduced water supply/quality and changes in biodiversity with the consequence that the community experiences famine conditions and loss of wildlife and plants. WWF PRIORITY PLACE 3: UDZYNGWA LANDSCAPE (TANZANIA) EXCESSIVE AND INTENSE RAINFALL: Rainfall is projected to increase in amount and intensity in the late rainy season (January – March) by the 2050s under both the SRES A2 and B1 scenarios. This trend is seen in the projected change in monthly rainfall totals, the change in rain day frequency and the projected change in rain days above 22.5mm (the 90 th percentile of the observed record). Impacts from an increase in amount and intensity of rainfall are already being felt in the region. These impacts are projected to become worse under the future climate. These impacts may include properties being destroyed or subsiding resulting in people in communities becoming homeless and the need for disaster relief efforts. Crops may be lost as a result of flooding, such as rice seedlings which have been known to be washed away by floods in the past. Communication infrastructure is also affected by flooding, such as roads and railways. These impacts all lead to increasing poverty in the community due to the losses incurred. DRY SPELL DURATION: The projections of dry spell duration indicate a significant increase in dry spell duration from May – October by the 2050s. This increase in dry spell duration may lead to more occasions where there is a low level of water in the Mtera dam which will lead to a low supply of electricity at the country level. WWF PRIORITY PLACE 4: RUMAKI AND MATUMBI HILL SEA AND LANDSCAPE (TANZANIA) EXCESSIVE AND INTENSE RAINFALL: This area experiences bimodal rains with the long rains occurring from March – May and the short rains from November – December. The rainfall trends suggest that the frequency and intensity of rainfall is generally decreasing, while the tails of the rainy seasons are getting wetter. The climate projection results suggest a decrease in rainfall by 2050s during the main dry season (June – October) and an increase during the rest of the year. Rain day frequency and intensity shows the same general tendency towards an increase in the wet season months. In the past, heavy rainfall / flooding events have impacted the region through damage to buildings and properties, loss of life, migration out of the region, crop losses, killing of mangrove trees by fresh water and fishery losses. These impacts led to population displacements, health issues, famine and biodiversity loss. As the rainfall is projected to increase in amount and intensity during the rainy seasons in the future (by the 2050s), these impacts and their associated consequences are projected to worsen into the future. DRY SPELL DURATION: P a g e | 91 The observed trends in dry spell duration exhibit a decrease in dry spell duration for the months of October to January (nominally the rainy season months) and an increase in dry spell duration during the months of May – July (nominally the dry season months). A similar trend is shown in the climate projections for the 2050s which suggest a decrease in rainfall during the main dry season (June – October) and an increase in dry spell duration during the same months. Previous dry spells in the area have led to drought which has caused crop failure, resulting in famine and mortality in the region. The climate projections for the 2050s suggest that drought events may increase so these types of impacts may increase into the future. STORM SURGES: There are no explicit projections for storm surge however, from the trend information wind speeds have been on the increase in the area and the annual trend in sea level rise has been 3-4mm/year. These both indicate that storm surges may become more frequent under future climate change. Storm surges have previously impacted the area and the projections of future climate suggest that these impacts will worsen into the future. These impacts include destruction of trees, crops (including coconut fields) and mangrove trees. This means that the land has already became increasingly barren, lessening the buffering zone for the islands. Added to this increasing wave strengths may continue to cause large amounts of beach erosion, further affecting the village communities in these areas. WWF PRIORITY PLACE 5: GREATER RUVUMA LANDSCAPE (TANZANIA AND MOZAMBIQUE) This region has had previous impacts from events of excessive rainfall, strong winds, lightning, drought and fog/mist. These are all significant events but we can only comment here on rainfall and drought as only these two variables have available climate projections. EXCESSIVE AND INTENSE RAINFALL: In terms of amount of rainfall, the observed records exhibit a trend of an increase in rainfall on the borders of the rainy season. The future projections for SRES emission scenario A2 show a small projected increase in rainfall in January – March (at the end of the rainy season) and decreases are shown for the beginning of the rainy season. The SRES emission scenario B1 does not show a clear trend for projected rainfall amounts. The intensity of rainfall in the observed record has shown an increasing trend in November, January and March and a decreasing trend in February. In the projections there is a projected change towards more intense rainfall at the end of the rainy season and less intense rainfall at the beginning of the rainy season. Increasing excessive rainfall / flooding events under the future climate will impact the community. Flooding has already led to crop failures with extensive swaths of cropland destroyed. If intense rainfall increases in the future then these impacts are projected to get worse. DRY SPELL DURATION: Both the observed and projected future climate shows a trend towards an increase in dry spell duration in the dry season. In the vulnerability assessment drought was indicated as a significant event which has been known to result in famine, hunger and ultimately changes in lifestyle as the community is forced to act out of the ordinary to sustain themselves. If dry spell duration increases in the future then these impacts are projected to worsen. WWF PRIORITY PLACE 6: MTWARA-QUIRIMBAS SEASCAPE (TANZANIA AND MOZAMBIQUE) From the vulnerability assessment it can be seen that this region is currently particularly affected by strong wind events. Unfortunately wind is a climate variable that has a large uncertainty in the climate model so very little can be said about P a g e | 92 future winds for the region. The wind events identified in the vulnerability assessment are most often associated with a storm event and are accompanied by extreme rainfall. Therefore, projections for future rainfall have been used to infer potential vulnerabilities for future storm events. EXCESSIVE AND INTENSE RAINFALL: From the observed information there appears to be no clear trend in the change of rainfall or heavy rainfall events. However, when looking to the projections information for the 2050s, there is a slight projected increase in monthly total rainfall in rainy season (Jan – March) with a slight decrease projected for dry season (May – November). This is accompanied by a projected increase in dry spell duration in the dry season. There are also more frequent heavy rainfall events (change in monthly rain days above the 90th percentile of observed record) projected for the rainy season and a decrease in heavy rainfall events projected for the dry season months. Excess rainfall can lead to impacts such as landslides, floods and damage to infrastructure. These impacts can cause entrainment of land by erosion, potential damage to crops, property and infrastructure and damage to sanitation systems with the consequence that health and hygiene are affected in the communities. These impacts are a reflection of the current vulnerability of the community to extreme rainfall events. However, if events become more extreme in the future, additional impacts could be experienced. SEA LEVEL RISE: The trend for sea level shows an increase of between 4-6mm/year from 1992-2010. If this trend continues it will mean that the coastal area exhibits worsening impacts from sea level rise in the future. An indication of future potential impacts can be gleaned from the current vulnerability assessment. These impacts may include damage to buildings and subsidence. As a consequence of this, families may be rendered homeless as their homes are destroyed, there may be agricultural losses and tourism may be affected as the coastal erosion protection barriers continue to fail. Over time tourism developments may be forced to implement additional or more effective protection barriers against the erosive forces of sea level rise. SUMMARY OF CLIMATE VULNERABILITIES ACROSS ALL SIX PRIORITY REGIONS: Climate Variable Intense and excessive rainfall Wind, hail and storms and lightning Impact Consequence Landslides Damage to properties Subsidence Damage to vegetation Changes in Biodiversity Damage to Infrastructure Crop losses Damage to fisheries Damage to vegetation Damage to infrastructure and properties Floods Landslides Low fisheries productivity Damage to fishing vessels Homelessness Deaths Disruption of transport Health issues Increase in agricultural and honey production Increased poverty Death of coral Famine Reduced marine fisheries Homelessness Disruption of transport Agricultural losses Reduced honey production More fishing inshore P a g e | 93 High temperatures and dry spells Damage to vegetation Fires Pests Power failure Reduced water supply/quality Crop failure Changes in Biodiversity Sea level and surge Damage to buildings Subsidence Increase in pests Famine and malnutrition Death Agricultural losses Reduced livestock Power rationing Emigration Reduced biodiversity (plant and wildlife loss) Reduced honey production Homelessness and families displaced Agricultural losses Crops destroyed Fog and mist Table: Summary vulnerability assessment A METHODOLOGICAL WAY FORWARD: Determining ones current and future vulnerability to climate, as compiled in this report, is an important first step in developing an adaptation plan for a region. There are many approaches to developing an adaptation plan but the generic process is much the same. We have chosen to present the UK Climate Impacts Programme Adaptation Wizard to illustrate a methodological approach to implementing adaptation. However, before embarking on comprehensive adaptation planning it is important to further research the details of each step in the process. Figure 103: The structure of the adaptation process (UK Climate Impacts Programme Adaptation Wizard) P a g e | 94 STEP 1: GETTING STARTED The getting started step is an important component that should not be overlooked because it sets down the groundwork for the framework. Issues that should be answered in this step include ones particular motivation for undertaking an adaptation assessment, what one is hoping to achieve and what difficulties one might face along the way. This will enable one to assemble an appropriate team and gain buy-in to the process which in turn will promote a successful implementation process. STEP 2: AM I VULNERABLE TO CURRENT CLIMATE? The vulnerability analysis for six of the nine priority regions has been started within this project. However, it is important to note that the vulnerability analysis undertaken for the purposes of this study was a preliminary desk-based assessment of the vulnerability of each region and was not an in depth comprehensive study. So the vulnerabilities / impacts outlined in this section are merely an indication of the types of impacts that may occur in the future, not a comprehensive impacts analysis. Preliminary vulnerability assessments were only received for six of the nine priority areas so it is not robust to comment on the vulnerability of the other priority areas within the study. However, it can be said that the regions studied showed very similar vulnerabilities. Therefore, it is reasonable to assume that some, if not all, of these impacts will be exhibited in other regions. Vulnerability analysis is, to a large extent, location specific so it should be noted that there will be impacts at other stations within regions that have not been drawn out in this study. Naturally, coastal regions will exhibit additional impacts because of the added dynamic of the marine environment. These impacts, particularly in the light of future sea level projections, require further investigation. This study has not drawn out any information on critical thresholds, i.e. climatological conditions that, if exceeded, will cause unacceptable consequences. When looking at future vulnerability and adaptation options this information is critical for making informed decisions. A changing climate could mean that a critical threshold will be exceeded more frequently than in the past, and adaptation measures will need to be taken to manage risks at an acceptable level, taking into account individual risk attitudes. An additional base for informing an adaptation assessment is how previous impacts were dealt with and the costs that were incurred. This information provides an indication of where there is scope for institutional capacity building and preparedness, but also what the cost-benefit of instituting adaptation actions may be. STEP 3: HOW WILL I BE AFFECTED BY CLIMATE CHANGE? This project delivers projections of climate that can be used as a basis for completing this step. Based on the preliminary vulnerability information received through the desk based assessment, a commentary on the potential impacts of climate change for each of the priority areas is provided within this report. However, as previously noted, this is based on information from one station point in each priority region so this exercise should be expanded to more stations within each region to increase its comprehensiveness. There is also additional information that should be considered within this step, such as the risk that the climate impacts present and whether these are more or less important than other socio-economic risks. The risk of climate change can seem like the most important risk to plan for but, on further analysis, often there are other overriding factors that negate action on climate change. This needs to be ascertained before progressing with climate change adaptation initiatives. In order to progress to the next step is it is necessary to flesh out the climate change risks and rank them, in an exercise, according to their likelihood of occurring versus the magnitude of the consequence. As this is a subjective exercise it is important to note the thinking behind the risk rankings so that the work can continually be reviewed. This exercise should be undertaken for multiple future time periods to assess the potential urgency of each risk. The result allows one to P a g e | 95 identify which risks require priority action and also ensures a manageable number of risks to address in the next step of the process. Figure 104: Illustration of risk ranking exercise STEP 4: IDENTIFYING, ASSESSING AND IMPLEMENTING ADAPTATION OPTIONS Having identified a number of priority risks in the previous step, now a range of adaptation options can be brainstormed to mitigate against each risk. This should only be undertaken if one feels there is enough information at hand to embark on this process. It is encouraged to brainstorm a number of relevant adaptation options and then evaluate each one against a set of selection criteria to choose the most appropriate measure to manage the climate risk. The selection criteria may include areas such as; effectiveness, efficiency, equity, flexibility, sustainability, practicality, legitimacy, urgency, cost, robustness etc. In addition, there are several factors that should be considered when undertaking an adaptation option selection. These include determining the urgency of action, what level of adaptation is required and if there are ways of integrating the adaptation option into the mainstream activities already underway. A crucial element to all these considerations is the level of risk that can be accepted, and alongside that, what the consequences may be if either over or under-adaptation resulted. Once the adaptation options have been chosen an implementation plan can be formulated and the adaptation actions implemented. The implementation plan will be organization specific and will depend on particular organizational procedures. STEP 5: MONITOR AND REVIEW The final step in the adaptation process is to continually monitor and review the adaptation strategy. It would be unwise to assume that once the plan is implemented that the job is done. Adaptation is an iterative process and requires close monitoring as to its effectiveness and whether adjustment is required. What is known about the past, present and future is rapidly changing and adaptation options need to be continually assessed against a dynamic world. P a g e | 96 GAPS AND FUTURE WORK The report has highlighted a number of gaps, issues and challenges with climate analysis and projection in the CEAI countries and priority landscape regions. This section will attempt to summarise these issues as well as point forward to potential future work or activities that could assist in closing gaps and/or overcoming issues and challenges. OBSERVED DATA GAPS One of the most obvious and apparent gaps in the analysis is that of observed data. All three countries have a severe lack of quality, long period, observed station records. Tanzania and Mozambique in particular have very poor coverage of station data. Kenya has a higher coverage but suffers from poor data quality and inconsistent coverage that is common across Africa. The limitations imposed by the lack of observed station data are important. Poor data quality as well as short observed time periods precludes drawing robust conclusions from such datasets. Alternative sources of observed data such as satellite records have been discussed in the report and highlighted as a potentially important means of overcoming the data gaps. However the limitations of satellite data were also highlighted including the short record length, difficulty in capturing certain types of rainfall, and challenges in calibration due to the lack of station records. Certainly into the future satellite derived observations will play an increasing important role in climate analysis however the need for ground truthing and validation will continue and hence the need for continued maintenance and growth of the observing station network is of great importance Ocean observed data is another area where there appears to be a distinct lack of easily accessible data. While the PMSL project referenced in the report does maintain and distribute tide gauge data for a number of locations, only two locations seem to be available for the east African coastline. The TOPEX/POSEIDON satellite altimetry data is available but requires significant processing and analysis in order to explore regional sea-level rise questions. Ocean chemistry such as salinity and pH data is available but often in difficult to use data formats that require significant effort to unpack and explore. OPPORTUNITIES AND FUTURE ACTIVITIES OBSERVED DATA A number of important opportunities exist in the area of observed data: Accessing and making available private datasets It is understood that there are a large number of observed datasets (weather station and ocean) that are managed and held by country meteorological offices, environment agencies, research organisations etc. These datasets are a valuable resources and yet are often difficult to obtain unless through personal networks or connections with key people in the organisations. In addition, many met services do not release archived data records as they consider them to have commercial value. This places a severe restriction on the science and analysis that can be done in a particular country. The GHCN provides a critical service in managing weather station records and have expressed interest in gaining access to further records from private or restricted sources. Data cleaning In addition to data access, data cleaning and quality control is of paramount importance. In order for robust statistics, analysis, and modelling to be performed using observed records, the records need to pass through comprehensive tests for quality and homogeneity. This is particularly true for trend analysis as sensor P a g e | 97 degradation, equipment changes, and station location changes can radically alter then sign of time series trends if not detected and accounted for. Trend analysis and attribution It is apparent through this report that trend analysis of historical observed climate is complex, particularly for high variability variables such as precipitation. Data quality, complexities of merged/hybrid data sources, and sensitivity of trend statistics produce difficulties in interpreting computed trends. Added to this is the challenge of attributing observed trends to anthropogenic or long term climate change as opposed to natural cycles of variability. Simplistic and naive approaches to these questions can produce very misleading results. Further work is required in order to determine the regional climate trends in the CEAI countries that takes into account all of these issues. CLIMATE CHANGE PROJECTIONS As regards the availability and development of climate change projections at relevant temporal and spatial scales, a number of gaps exist as well as opportunities: Regional Climate Modeling and CORDEX While dynamical downscaling or regional climate modelling has not been used in this report for the reasons described, dynamical modelling is still a valuable and necessary tool to be used to unpack regional climate dynamics and explore future climate dynamics for a region. The gap in this field is the lack of model simulations for different regions. Typically research activities, due to lack of resources or methodological constraints, identify a single regional model and apply that to their research question. While this can provide very useful insights and answers, the inherent uncertainty associated with the use of a single model within unknown future climate bias and error make such simulations unsuitable for climate impacts work. The COrdinate Regional Downscaling Experiment is an attempt to fill this gap. CORDEX is a World Climate Research Programme (WCRP) mandated activity spanning a large number of research groups across the globe. The objective is to coordinate a set of controlled downscaling (both dynamical and statistical) experiments for specific regions across the world. The experiments are controlled in that the target resolution and domains are tightly specified and the output variables, statistics and storage formats are all identical. This allows for very rich exploration of the performance and uncertainty associated with different downscaling methods. CORDEX is, however, still a future opportunity as currently only historical climate simulations have been done. A few sample future climate simulations have recently been completed but these results remain in the domain of research rather than application. CSAG at UCT is the coordinating partner for the African domain experiments and is intensely involved in this activity. CSAG will also become the African data dissemination partner for CORDEX Africa data when suitable impacts relevant data is available Understanding of regional dynamics and feedbacks Much work still remains around regional scale dynamics and feedbacks between the climate systems and the land surface, soil moisture, ocean, lakes and vegetation. In many areas of Africa, including east Africa, research suggests that local climate forcing factors and feedbacks can strongly moderate regional manifestations of larger scale climate change. Regional climate models as well as other modelling such as hydrology and vegetation can help to unpack these systems but accurate and continuous monitoring and data collection is also foundational to support such research.