Martin Herold
1
, Yasumasa Hirata
2
, Patrick Van Laake
3
, Gregory Asner
Rosa María Román-Cuesta 6
4
, Victoria Heymell
5
,
1.
Wageningen University. Building 101. Droevendaalsesteeg 3, 6708 PB Wageningen. The
Netherlands. Tel. +31 317 481276; Fax: +31 317 419000. Martin.Herold@wur.nl
2.
Shikoku Research Center, FFPRI. Forestry and Forest Products Research Institute 2-915
Asakuranishi, Kochi, Kochi, 780-8077. Japan. Tel. +81-88-844-1121; Fax. +81-88-844-
1130. hirat09@affrc.go.jp
3.
UN REDD Vietnam Programme. 172 Ngoc Khanh, #805. Ba Dinh, Ha Noi Vietnam. patrick.van.laake@undp.org
4.
Carnegie Institution. 260 Panama Street. Stanford, CA 94305. USA. Tel.+1-462-1047 200 gasner@globalecology.stanford.edu
5.
FAO. Viale delle Terme di Caracalla 15. 00100 Rome, Italy. Tel. +39 06 570 54451 Fax:
+39 06 570 55137, victoria.heymell@fao.org
6.
UN-REDD Programme. FAO MRV team. Viale delle Terme di Caracalla 15. 00100 Rome.
Tel. +39 06 570 52044; Fax: +39 06 570 55137. rosa.roman@fao.org
1

Abstract
There are currently more than fifty definitions of forest degradation but none of them is accepted in the international negotiations as a univocal, operational and multipurpose definition (i.e. for the use in national-level reporting. While forest degradation is a broad topic, the review presented here is addressing the degradation issue from a climate change and forest carbon stock change perspective; in particular considering the current discussions on REDD+. The IPCC 4 th
Assessment Report sustained that the world’s degraded forests reached ca. 100 million of hectares per year. This represents almost 10 times more global area affected by degradation than by deforestation (i.e. ca.
100 million degraded ha.yr
-1
versus ca. 13 million deforested ha.yr
-1
during 2000-2005). For this reason, forest degradation rates must necessarily be reported together with deforestation rates to guarantee integrated and coherent climate mitigation actions. The REDD+ mechanism addresses the evident role of reducing deforestation and forest degradation as global climate mitigation tools, but it also considers the role of conservation, sustainable management of forests and the enhancement of forest carbon stocks. While the final rules for REDD+ are still under development, non-Annex I countries will have to evaluate their historic rates of deforestation and degradation to estimate their
Reference Emission Levels. There is not one method to monitor forest degradation that fits all circumstances and the methodological choice depends on a number of factors including the type of degradation, available data, capacities and resources and the potentials and limitations of various measurement and monitoring approaches. Current degradation rates can be measured through field based data (i.e. Multi-date national forest inventories and permanent sample plot data, commercial forestry datasets, proxy data from domestic markets, etc) and/or remote sensing data (i.e. direct mapping of canopy and forest structural changes or indirect mapping through modelling approaches), with the combination of them both providing the strongest alternative. Historic degradation assessments for non-Annex I countries frequently lack consistent historic field data, forcing countries to rely strongly on remote sensing approaches mixed with current field assessments of carbon stock changes. The current paper describes methodologies for assessing current and historical rates of forest degradation to support developing countries interested in implementing the REDD+ mechanism.
2

1.1
Definitions of forest degradation in relation to forest carbon stocks
There are currently more than fifty definitions of forest degradation (Lund 2009, Simula 2009) but none of them is accepted in the international negotiations
1
as a univocal, operational, multipurpose definition.
Forest degradation is generically defined as the reduced capacity of a forest to provide goods and services (FAO 2002). However, this definition is too broad to be operational. In the context of climate change, the IPCC (2003) developed a definition of forest degradation that focuses on human-induced changes in the carbon cycle in the long run:
“A direct human-induced long-term loss (persisting for X years or more) of at least Y% of forest carbon stocks [and forest values] since time T and not qualifying as deforestation or an elected activity under Article 3.4 of the Kyoto Protocol
2 ”.
In order to operationalise this definition, i.e. for the use in national-level reporting, it would be necessary to specify an area threshold, as well as time and carbon loss thresholds.
Forest degradation, from the point of view of climate change policy and the IPCC national estimation and reporting guidelines, refers to a loss of carbon stock within forests that remain forests (IPCC 2003). More specifically, degradation represents a human-induced negative impact on carbon stocks, with measured forest variables (i.e. canopy cover) remaining above the threshold for the definition of a forest. Moreover, to be distinguished from (sustainable) forestry activities, the decrease should be persistent. The IPCC 2003 definition faces several challenges if used for monitoring purposes: i) it lacks a clear definition of a temporal threshold considered as “long term”; ii) it lacks a suggestion or identification of minimum thresholds of carbon stock change associated with degradation to distinguish it from natural forest disturbances; and iii) it is challenged by the identification and isolation of human-induced degradation from other degradation factors, which may well be interlinked. The persistence could be evaluated by monitoring carbon stock changes either over time (i.e. a net decrease during a given period, e.g. 20 years) or along space (e.g. a net decrease over a large area where all the successional stages of a managed forest are present)
(GOFC-GOLD 2009).
Considering that, at national level, sustainable forest management may lead to national gross losses of carbon stocks (e.g. through harvesting) which are lower than (or equal to) national gross gains (in particular through forest growth), consequently a net decrease of forest carbon stocks at national level during a reporting period would be due to forest degradation within the country. Conversely, a net increase of forest carbon stocks at national level would correspond to forest enhancement.
Therefore, it is also possible that no specific definition is needed, and that any net emission will be reported simply as a net decrease of carbon stock in the category “Forest land remaining forest land” (GOFC-GOLD 2009) – a perspective that is also shared by an expert group convened by the
UNFCCC SBSTA
3
(UNFCCC 2008).
1 i.e. UN Convention on Biological Diversity (CBD), UN Convention to Combat Desertification (UNCCD), or UN
Framework Convention on Climate Change (UNFCCC).
2
Forest management, cropland management, grazing land management and revegetation.
3 SBSTA: Subsidiary Body for Scientific and Technical Advise
3

For the purpose of this paper, which is aims to provide a review of different assessment methods, the above mentioned assumption will be used. However, under other circumstances, a specific definition may be required. In this case, Simula (2009) summarizes the elements that an operational definition of forest degradation should provide: 1) identification of forest goods and services, 2) a spatial context of assessment, 3) a reference point, 4) coverage of both the process and state
(degradation/degraded forests), 5) relevant threshold values, 6) specification of reasons for degradation (human induced/natural), 7) an agreed set of variables, and 8) indicators to measure the change of a forest. Additional elements could be added or singled out, depending on the particular interests related to the purpose of the definition.
While forest degradation is broad topic, the review presented here is addressing the degradation issue from a climate change, carbon and REDD+ perspective. The authors have collated and critically reviewed case studies, articles, guidelines, manuals and other documents describing methodologies for assessing current and historical rates of forest degradation to support developing countries interested in implementing the REDD+ mechanism.
1.2
Main causes of forest degradation affecting changes in forest carbon stocks
Forest degradation can have any number of causes, dependent on resource condition, environmental factors, socio-economic and demographic pressure and “incidents” – e.g. pests, disease, fire, natural disasters. The understanding and separation of different degradation processes is important for the definition of suitable methods for measuring and monitoring. Various types of degradation will have different effects on the forest (carbon) and result in different types of indicators that can be used for monitoring degradation using in situ and remote methods (i.e. trees being removed, canopy damaged etc.).
For the purpose of this review the emphasis is on those forms of forest degradation that are caused by direct human impacts on the forests (i.e wood removal) or indirect human impacts (i.e. long term forest management that favours fire presence and impacts) on the forest. The reduction of forest degradation by human influenced causes is eligible under the REDD+ mechanism (4/CP.15
4
; Draft
Decision/CP.16
5 ).
1.2.1.
Extraction of forest products for subsistence
Privately or communally managed forests are often subject to extraction of forest products for immediate use by local households. Extractions are for such uses as fuelwood for cooking, collection of fruits, roots and other edible tree organs, collection of fodder for livestock, and harvesting of timber and thatch for construction. In more established and stable cultures and communities such extractions can be sustainable (e.g. tribal groups in Papua New Guinea and elsewhere, community managed forests in Nepal and India, the ejido system in Mexico), but in many other cases the increasing population of the last few decades has put so much pressure on the forest that the extraction is no longer sustainable.
The following sub-sections describe activities causing forest degradation and reductions in carbon stocks in forests that are not under other land uses; reflecting the requirements of the definition used for FAO’s global forest resources assessments. It should be noted that expansion of agriculture into forests (i.e. forest grazing, shifting cultivation, agroforestry) also cause losses in forest carbon stocks and should also be considered and monitored, and depending on the forest definition reported as either emissions on forest or non-forest land.
4 4/CP.15: http://unfccc.int/resource/docs/2009/cop15/eng/11a01.pdf#page=11
5 Draft Decision/CP. 16: http://unfccc.int/files/meetings/cop_16/application/pdf/cop16_lca.pdf
4

1.2.2.
Extraction of forest products for local markets
Most developing countries have seen rapid urbanization in recent decades and this has created a market for forest-based products which, in some cases, has resulted in forest degradation.
Particularly the production of charcoal has led to forest degradation, in dry forest ecosystems such as the miombo of southern Africa. Other products that are harvested include timber, bamboo and rattan for construction and furniture making and minor products such as raffia, vines and leaves for making household utensils and rope.
Serving local and national markets for forest-based products has been facilitated by the expansion of road infrastructure in most developing countries. Better infrastructure has decreased the cost and expanded the options of transportation of forest products. Since the cost of forest products – e.g. charcoal – is often far lower than “urban” alternatives – e.g. kerosene – the cost of transportation is easily recouped.
1.2.3.
Industrial extraction of forest products
International scrutiny of harvesting operations in developing countries has led to the development of alternative harvesting schemes to replace the total removal of commercially interesting tree species of specimens above a minimal girth and with little or no consideration for the remaining stock during harvesting. While the management is fostering regeneration of the forest – this is an explicit requirement under all international certification schemes for selective harvesting – the forest will have noticeably lower carbon stocks for many years.
1.2.4.
Natural disturbances such as wildfire
All forms of excessive forest product extraction leading to degradation, as described in this section, impact the resilience of the forest to withstand external impacts, such as fire, pests and drought.
These impacts can be positive, although most are negative. Most impacts are driven by changes in the local hydrology: through extraction of trees or tree products more solar radiation is transferred to the soil which leads to drying of the soil and ultimately stress for the trees. For forests out of human influence, natural fires and degradation due to insect outbreaks are not reported under the
Convention (i.e. remote Siberian boreal forests ignited by lightning). However, very few developing countries have non-human influenced forest, so fires as well as insect outbreaks and any other forest disturbances, must be reported and will reduce a country’s REDD+ emission gains.
Different degradation processes are usually active within the same country. Some may affect large areas, some not, and it is common that they are not equally distributed among the country’s territory. Thus, forest degradation activities are often focused in specific areas and this should be considered in national measurement and monitoring efforts.
1.3
Forest degradation as key source category in the context of REDD+
Disturbances that lead to degradation such as forest fires, pests (insects and diseases) and climatic events including drought, wind, snow, ice, and floods have been reported to roughly affect 100 million of hectares globally per year (FAO 2006a, in the IPCC 4 th Assessment Report (Nabuurs et al.
2007)). Globally, this value represents almost 10 times more area affected by forest degradation than by deforestation (i.e. 12.9 million ha.yr
-1
(2000-2005), FAO (2006b); MEA (2005), indicating the scale and importance of global forest disturbances that lead to degradation. While these values are a compilation of areas affected by forest disturbances around the world, tropical regions are well known for large scale disturbances that lead to forest degradation: fire activity has been repeatedly reported to affect the tropic and subtropical region more than other latitudes (Dwyer et al.
1999,
5

Giglio et al.
2006) and severe storms and wind blows are also well known large scale degradation factors in tropical South America (Negrón-Juárez et al.
2010). For this reason, in the context of
REDD+, forest degradation rates must necessarily be reported together with deforestation rates to guarantee integrated and coherent climate mitigation actions.
Forest degradation often has different driving forces than deforestation. The emission levels of degradation are lower than for deforestation (per unit area); but cumulative and secondary effects result in significant carbon emission and degradation is often a precursor to deforestation.
Addressing deforestation does not automatically reduce rates of degradation. Failing to include degradation in a REDD+ agreement could leave considerable amounts of forest-based emissions unaccounted for (Murdiyarso et al.
2009). In case reducing deforestation measures are taken, monitoring forest degradation is important to avoid displacement of emissions from reduced deforestation. The evident role of reducing deforestation and forest degradation as global mitigation tools led to the petition by the Coalition for Rainforest Nations in Montreal 2005, at COP11, to reinforce Article 2 of the Kyoto Protocol regarding the protection and enhancement of sinks and reservoirs of greenhouse gases not controlled by the Montreal Protocol.
As a result of this petition, in December 2007, COP13 in Bali adopted 2 decisions:
1. The Bali Action Plan Decision 1/CP.13,
Where the Intergovernmental Panel on Climate Change de cided to address:
“Policy approaches and positive incentives on issues relating to reducing emissions from deforestation and degradation in developing countries; and the role of conservation, sustainable management of forests and enhancement of forest carbon stocks in developing countries”
2.
Reducing emissions from deforestation in developing countries: approaches to stimulate action
Decision 2/CP.13. This decision provides a mandate for several elements and actions by Parties relating to reducing emissions from deforestation and forest degradation and the role of conservation, sustainable management of forests and enhancement of forest carbon stocks in developing countries.
A methodological decision followed in COP15, Copenhagen 2009:
3.
Methodological guidance for activities relating to reducing emissions from deforestation and forest degradation and the role of conservation, sustainable management of forests and enhancement of forest carbon stocks in developing countries 4/CP.15.
This decision requests developing country Parties to take certain guidance into account for the 5
REDD+ activities, in particular those relating to measurement and reporting:
“To establish, according to national circumstances and capabilities, robust and transparent national forest monitoring systems and, if appropriate, sub-national systems as part of national monitoring systems that:
(i) Use a combination of remote sensing and ground-based forest carbon inventory approaches for estimating, as appropriate, anthropogenic forest-related greenhouse gas emissions by sources and removals by sinks, forest carbon stocks and forest area changes;
6

(ii) Provide estimates that are transparent, consistent, as far as possible accurate, and that reduce uncertainties, taking into account national capabilities and capacities;
(iii) Are transparent and their results are available and suitable for review as agreed by the Conference of the Parties.”
IPCC guidance
As stated in paragraph 6 of Decision 2/CP.13, as the basis for reporting greenhouse gas emissions from deforestation and forest degradation under the Convention, non-Annex I Parties are encouraged to use the most recently adopted or encouraged reporting guidelines, which include the application of the Good Practice Guidance for Land Use, Land-Use Change and Forestry (GPG-
LULUCF, IPCC 2003, 2006). This guidance will promote the elaboration of country forest reports on emissions by sources and absorptions by sinks that are transparently measured, consistently calculated over time, comparable with other countries methodologies, and as complete and accurate as possible. Unlike previous Guidelines, the LULUCF 2003 Good Practice Guidelines is “land use based”, meaning that countries need to report degradation under the category: “Forest land use that remain forest land use”.
To estimate the emissions associated with forest degradation, countries need to evaluate two aspects
(IPCC 2003):
1) Changes in forest carbon stocks due to the degradation processes per unit area (Emission Factors
= EF). How much carbon is lost from the forests and released to the atmosphere due to the degradation process (i.e. commonly measured through forest field sampling and through repeated forest inventories) (reported as MgC.ha
-1
-yr
-1
). Emission factors should be calculated for each of the
5 forest pools requested under the UNFCCC: aboveground, belowground, deadwood, litter, and soil organic carbon.
2) Area of forest land areas that remains forest land effected (Activity Data = AD); ideally for different disturbances or degradation types. How much and where forest area is or undergoing degradation changes? (i.e. statistics calculated through forest inventories or through remote sensing)
(reported in ha).
The estimation of the carbon emissions from forests on the national level caused to degradation will be done by multiplying the EF by the AD and sum it up for all forest and degradation types.
A few extra aspects are required to estimate emissions associated with forest degradation:
1) Selection of the main carbon pools to measure and monitor forest degradation: the IPCC (2003) defines five carbon pools to be measured and monitored: aboveground biomass, below-ground biomass, litter, dead wood and soil organic carbon. Key stock sources must be selected. In the tropics, the most generalized approach is to monitor only above-ground biomass, even though soil stocks in peatlands also require attention and can contain more carbon stock than the AGB. In this working paper we will mainly focus on methodologies to monitor changes in aboveground forest biomass.
2) Reporting Tiers: the IPCC (2003) provides three tiers for estimating emissions, with increasing levels of data requirements and analytical complexity and increasing accuracy:
7

• Tier 1 uses default values for forest biomass and forest biomass mean annual increment (MAI) which are obtained from the IPCC Emission Factor Data Base (EFDB), corresponding to broad continental forest types (e.g. African tropical rainforest). Tier 1 also uses simplified assumptions to calculate emissions.
• Tier 2 uses country-specific data (i.e. collected within the national boundary), and by resolving forest biomass at finer scales through the delineation of more detailed strata. For degradation, in the absence of repeated measures from a representative inventory, Tier 2 uses the gain-loss method using locally-derived data on mean annual increment.
• Tier 3 uses actual inventories with repeated measures of permanent plots to directly measure changes in forest biomass and/or uses well parameterized models in combination with plot data.
Tier 3 often focuses on measurements of trees only, and uses region/forest specific default data and modelling for the other pools. The Tier 3 approach requires long-term commitments of resources and personnel, generally involving the establishment of a permanent organization to house the programme.
3) Methods to estimate carbon stock changes (emission factors): There are two fundamentally different, but equally valid default approaches to estimate carbon stock changes, under the IPCC:
1) The stock-based or stock-difference approach
2) The process-based or gain-loss approach.
These approaches can be used to estimate stock changes in any carbon pool, although their applicability to soil carbon stocks is limited. The stock-based approach estimates the difference in carbon stocks in a particular pool at two points in time. This method can be used when carbon stocks in relevant pools have been measured and estimated over time, such as in national forest inventories. The process-based or gain-loss approach estimates the net balance of additions to and removals from a carbon pool. This type of method is used when annual data such as biomass growth rates and wood harvests are available. In reality, a mix of the stock difference and gain-loss approaches can be used (GOFC-GOLD 2009).
Most non-Annex I countries are only now starting to develop their national forest inventories and cannot count on two or more measurements in time to apply the Stock-Difference method at a national level. Most non-Annex I countries will therefore have to rely on the Gain-Loss method to calculate their emission factors.
1.4
Identifying and promoting the use of effective and cost efficient methodologies and tools to monitor forest degradation
The latest request by the Climate Convention to measure, monitor, and report human-induced forest greenhouse gas emissions and removals opens new methodological challenges. Developing countries will have to identify and promote the use of effective and cost efficient methodologies and tools to monitor forests and related degradation in terms of changes in forest carbon stocks and sequestration rates in “forests remaining forests”, with an initial focus on historical periods, consistency over time, and the need to establish capacities to continue monitoring in future periods.
This will be particularly important for forest carbon rich countries such as those in tropical regions, where forest degradation can easily be identified as a key source category. A key source category is
“an emission or sink category that is prioritised within the national inventory system because its estimate has a significant influence on a country’s total inventory of direct greenhouse gases in terms of the absolute level of emissions, the trend in emissions, or both” (IPCC 2003). Key source
8

categories should be estimated using higher tiers if sufficient resources are available. Higher tiers refer to lower level of uncertainty associated with the data and therefore higher accuracies.
The key category concept is important since it helps focusing country monitoring efforts to estimate the most relevant components of the GHG budget. Since direct emissions from degradation are smaller than those from deforestation (per unit area/emission factor), a country may have the flexibility to put more efforts in monitoring deforestation and can consider which types of degradation are significant on the national level and should be measured using high tiers. These considerations help to make the monitoring efforts more efficient in terms of monitoring cost versus measured carbon stock changes (per hectare).
There is not one method to monitor forest degradation. The choice of different approaches depends on a number of factors including the type of degradation, available data, capacities and resources and the potentials and limitations of various measurement and monitoring approaches.
Methodological challenges associated with measuring forest degradation are diverse, among others:
1) Consideration of temporal degradation thresholds and spatial scales. The effect of forest degradation on forest carbon stocks depends on time. To avoid mixing the effects of short-term carbon stock reductions using sustainable forest management practices with long-term effects of unsustainable practices leading to forest degradation, a temporal threshold can be selected for each ecosystem type. Temporal thresholds of degradation that guarantee “long-term” disturbance depend on ecosystem resilience and the intensity of the disturbance, and are complex to establish. Spatially, there is a general dominant perception that a forest stand is the basic unit of decisionmaking in conserving or enhancing forest carbon. However, forest management decisions are based on land planning which concerns larger forest management units
(for example, watershed, landscapes and forest concessions). Minimum mapping units also have key methodological implications.
2) Integration of field and satellite datasets: Monitoring changes in carbon stocks due to forest degradation relies heavily on field surveys but can benefit from integration of remotely sensed data with site-specific biophysical field attributes. Which biophysical parameters to measure and which time thresholds are appropriate for relating both data approaches are key issues to consider. Moreover, it is best to calibrate remote sensing variables with local forest biomass instead of using regional means. However, limited historical data exists for forest degradation.
3) Spatial impact and intensity: Different forest degradation processes activities are often focused in specific areas within a country and this should be considered in national measurement and monitoring efforts to track the most important activities and impacts and to use the available monitoring resources most efficiently.
4) Identification of key forest carbon stocks affected by degradation: Methods for calculating carbon stock changes vary for each relevant carbon pool (above ground biomass, below ground biomass, litter, dead wood and soil organic carbon), as well as for emissions of non-CO
2
greenhouse gases. In boreal forests such as Siberian Taigas or in tropical mountain regions, degradation of carbon rich peatlands will require different methodologies from above ground measurements of forest degradation in tropical evergreen rainforests.
9

1.5
Current challenges to measuring historical forest degradation
Historical forest degradation is a critical element in quantifying a country’s potential reduction in emissions and, therefore, its potential future benefits through a potential post-Kyoto mechanism.
Ex-ante evaluations of forest degradation are required to estimate a country´s reference emissions level, which will condition its future access to carbon crediting. The estimation of historical forest degradation faces however, further complications besides the previously mentioned general methodological challenges:
1) Lack of data : many countries, in particular those in tropical regions lack historical data on forest degradation and its impacts on forest carbon stocks. Historical national data are often limited to satellite archives while remote sensing itself has limitations in detecting degradation activities.
2) Insufficient capacity : while many developing countries have some level of experiences and data for monitoring commercial forestry activities, these capacities are often not sufficient to implement a national survey to assess historical deforestation and forest degradation.
3) Temporal considerations : there is currently no agreement regarding the temporal threshold associated with “long-term carbon stock loss”. If the forest degradation process in the field implies a cumulative long-term gradual carbon stock loss, methodological approaches such as direct measurements through remote sensing and fieldwork could still be usefully applied.
However, if forest carbon stock losses occur in shorter time periods, field validation and remote sensing measurements are challenged by the quick recovery of the forest, jeopardizing its measurement and monitoring. Moreover, in areas with high cloud persistence satellite data might be unavailable to track short-term degradation disturbances.
4) Integration of different data sources : Historical datasets on forest degradation are rare. The integration of remotely sensed data with site-specific biophysical field attributes for past assessments and other sources (i.e. forest management data) is challenging. The magnitude of historical forest degradation may have to be estimated through indirect approaches: modelling and/or other indirect methods to minimize the risk of overestimating avoided emissions under REDD+.
5) Inconsistencies when linking historic and present degradation datasets and methodologies:
Data limitations for historical periods, in particular those of consistent multi-date field data or the availability of high-quality remote sensing data, are often less prominent in current or future monitoring efforts and assessments are expected to be more accurate and verifiable.
Nevertheless, the need for consistency between historical, current and future estimates is important and should be a primary target of related surveys.
To estimate the emissions attributed to forest degradation, countries need to evaluate changes in forest carbon stocks due to the degradation processes (Emission Factors = EF), and changes in forest land areas that remain forest land (Activity Data = AD); ideally for different disturbances or degradation types, since different degradation processes imply different levels of carbon loss.
Since Parties need to offer country specific data with uncertainties (Tier 2 reporting), the estimation of country specific EF factors relies heavily on field sampling, frequently through National Forest
Inventories (NFI), while the estimation of annual estimates of AD is more operatively done through national wall-to-wall remote sensing approaches.
10

There exists, of course, a clear need to merge remote sensing to support field data collection, and field validation is needed to ground-truth remote sensing approaches. To the other extreme are those initiatives where remote sensing is being tested to estimate local EF (i.e. LIDAR and Radar technologies as a way to derive forest structure and above ground biomass), and field sampling is used as a way to estimate AD. These approaches are, however, not exempt from complications and limitations for reporting under the UNFCCC. Hence, EF derived from Remote Sensing rarely include uncertainty estimations that relate directly to forest biomass variability and their estimates do not include any other pool outside the above-ground biomass. Conversely, AD reporting through field data such as National Forest Inventories can only offer adequate levels of uncertainty if there is a sufficient number of ground sampling units (e.g. several dozen thousand per country)
6
(Stach et al.
2009). Moreover, NFI would not allow for annual AD reporting – needed for REDD+ reporting, since they are only re-measured over longer time periods (i.e. 5-10 year recensus).
Table 2 shows a summary of the relevancy of different forest degradation approaches (i.e. field data collection versus remote sensing) developed by Acharya and Dangi (2009) for Nepal.
6 This is the case for Annex I countries but frequently not the case for the National Forest Inventories of non-Annex I countries.
11

Table 2: Relevancy of different forest degradation assessment approaches in Nepal (based on aerial photographs 1:12000 to 1:60000) and Landsat TM image experiences) Source: Acharya and Dangi (2009)
2.1 Field observations and surveys to assess Emission Factors: changes in forest carbon stocks
A critical step to estimate forest degradation is a well designed and implemented sampling scheme to collect carbon stock data on the ground to measure carbon stock changes due to degradation.
Frequent field methods to evaluate carbon stock changes include (GOFC-GOLD 2009):

Inventory based approaches (national, sub-national) (i.e. Mexico);

Data from targeted field surveys (including interviews), research and permanent sample plots (i.e. India; Russia);

Commercial forestry data (i.e. logging concessions and harvest estimates) (i.e. Democratic
Republic of Congo);

Proxy data from domestic markets (charcoal, subsistence).
12

In the case of most Annex I countries, the collection of forest data through periodic forest inventories since the 1980s allows them to estimate the EF associated with historical and current forest degradation processes. For most non-Annex I countries, however, these long-term forest datasets are almost non-existent, or are focused on specific field assessments for commercial timber. In these cases, the time variable has to be substituted by space (e.g. evaluating net carbon stock decreases over a large area where all the successional stages of managed and unmanaged forests are present) (GOFC-GOLD 2009). This latter approach would consider the carbon stocks of the unmanaged forests as the reference value and would estimate the EF of the degraded forests by comparison. Some non-Annex I countries have some information available based on ground field data collection.
2.1.1
Sampling strategy and forest stratification
Field data collection can easily overwhelm any forest management organization wishing to determine forest degradation with reasonable accuracy levels (i.e. this value is a country decision, however an example of good accuracy would be a 10% maximum error at 90% confidence interval).
Most forest inventories are multipurpose. This means that several forest attributes are assessed and the inventory design frequently tries to optimize more than one forest attribute. When designing the sampling scheme of a Forest Inventory, there are at least two ways to reduce the sampling effort and to reduce uncertainties as far as practicable:
The first is a stratification of the territory and the forest land into more homogeneous units, under the assumption that the estimates of the variable of interest will be more similar within a stratum than among strata. Homogeneity here refers to a variable of interest: in this case carbon stocks.
Stratification can consider as many levels as necessary. For carbon stock measurement it should include, at least:
1.
Forest ecology, forest type : This determines the maximum biomass content and general properties of growth dynamics. Professional forest inventory at this level should determine general quantities of the forest – tree associations and density, basic wood density, average height . – and allometric equations of the forest type.
2.
Human practices that alter forests and supposedly result in similar carbon stocks: i) Degradation status : Understood as human induced activities that reduce the carbon stock, forest dynamic and forest composition for a selected time threshold (i.e. excessive fuelwood removal, wood charcoal production, fires, insect outbreaks, (see
Section 2.2). ii) Forest management for timber extraction ( i.e. low impact logging activities, illegal logging, plantations). iii) Conservation activities: Active management to either avoid human encroachment and/or to restore degraded areas.
A second consideration would be the cost-effectiveness principle, which not only takes into account the statistical requirements to obtain accurate estimates of forest parameters but also the costs associated with the implementation and logistics of the inventory. The designing and implementation of a Pilot Phase before the complete development of the inventory will offer clues about both statistical and cost requirements.
13

2.1.2.
Survey methods
Professional forestry organizations have typically used permanent sampling plots (PSP) to inventory forest resources and temporal dynamics. When historical records exist, it is worthwhile to continue the sampling on this basis. For assessment of forest degradation, however, it is advisable to work with sampling plots at random locations (in the forest stratum) in order to avoid bias in the estimates – the local population may avoid PSPs knowing that they are under special scrutiny of the forestry organizations.
Circular or square plots at random locations in the forest or transect surveys are the primary forms of survey organization. Circular plots have the advantage of being non-directional, while transect surveys are especially useful to detect gradients in some forest property. The distribution of the plots or transects has to be representative of the environmental conditions within the forest stratum, particularly exposure to the sun, elevation and soil hydrology.
Consistent and repeated measurements over time that cover all forest types in the entire country are key to offer reliable country-specific estimates of forest degradation.
A major issue affecting the reporting of forest degradation emissions is the estimation of its uncertainty. Among the REDD+ activities, both forest degradation and deforestation will require high levels of accuracy and certainty since they are major contributors of countries´ GHG forest budgets. Current degradation estimates can be used to reduce uncertainties but historic degradation will necessarily come with large uncertainties due to lack of available data to determine its accuracy.
2.1.3.
Data collection strategy
The data that needs to be collected can be broken down into two distinct types:
1.
Ecosystem or forest type properties.
Typical data in this category are basic wood density, average tree height, free branch height, taper, biomass expansion factor, description of tree associations, species identification to report on the incidence of minor species such as bamboo, rattan, lianas and vines, litter and dead wood, root biomass and soil properties.
These kind of properties require specialized skills for accurate assessment, but they tend to be stable over time and space and thus need only periodic surveying on a sampling basis.
The sampling is best performed by professional foresters.
2.
Forest dynamics data.
An essential property that has to be collected on a regular basis and distributed over the entire forest (stratum) is diameter at breast height (DBH). A follow up on other ecosystem gains and losses such as litter-fall, root growth, tree mortality, coarse woody debris, decomposition and changes in species composition would also be required for balancing the forest carbon cycle. With the properties from the first category, biomass can accurately be determined from DBH. DBH is also very easily measured and with little training people without formal knowledge of forest survey can be employed to collect DBH, making large-scale inventories feasible. When DBH is collected on a plot basis, the statistical distribution of individual tree DBH can be used as a proxy indicator for forest degradation, if the DBH distribution of undisturbed forest is used as a reference.
The data indicated here is routinely collected by forestry organizations in many countries and is not specific to the assessment of forest degradation. However, older inventories that have emphasized merchantable volumes of commercially interesting species – as was the case in most colonial forestry systems – can be correlated to similar inventories in the present era, supplemented with
14

forest properties that allow for the assessment of biomass, thus enabling an estimate of historical biomass content.
2.2
Remote sensing methods to measure Activity Data: changes in forest land uses that remain forest land uses
While most Annex I countries have been reporting their changes in forest area affected by degradation based on their National Forest Inventories, the measurement and monitoring of AD through remote sensing offers a series of advantages: i) it represents an operational, consistent, coherent, transparent and fairly accurate way of reporting on AD, which allows for near-real reporting on land use changes, ii) it is cost and time effective, iii) it offers data over remote and logistically complicated regions, iv) it offers a high frequency of data that help minimize seasonality problems, v) it is the only approach that objectively offers information on historical trends, and iii) it favours the control of leakage and permanence issues.
However, it also has several disadvantages: i) it is hampered by clouds, ii) it is limited by the technical capacity to sense and record the change in canopy cover with small changes likely not to be apparent unless they produce a systematic pattern in the imagery, iii) optical remote sensing is not useful to identify sub-canopy changes and therefore it is insensitive to under-canopy forest degradation (i.e. certain fire types, certain overgrazing, certain logging activities, and iv) not all degradation processes can be monitored with high certainty using remote sensing data. Table 3 offers a list of degradation processes that are best detected through remote sensing. Of course, a mixed approach would be desirable.
Highly Detectable Detection limited & increasing data/effort
Detection very limited
•
Deforestation
•
Forest fragmentation
•
Recent slash-and-burn agriculture
•
Major canopy fires
•
Major roads
•
Conversion to tree monocultures
•
Hydroelectric dams and other forms of flood disturbances
•
Large-scale mining
•
Selective logging
•
Forest surface fires
•
A range of edge-effects
•
Old-slash-and-burn
•
Small scale mining
•
Unpaved secondary
• agriculture roads (6-20m wide)
Selective thinning of canopy trees
•
Harvesting of most nontimber plants products
•
Old-mechanized selective logging
•
Narrow sub-canopy roads (<6m wide)
•
Understorey thinning and clear cutting
•
Invasion of exotic species
Table 3: Forest degradation activities and their degree of detection using Landsat-type data, Source: Peres et al.
(2006).
Independently of the approach chosen, the development of a monitoring system for degradation first requires that the causes of degradation be identified and the likely impact on the carbon stocks be assessed. FAO, together with the Members of the Collaborative Partnership on Forests (CPF) undertook a special study on forest degradation to identify the parameters of forest degradation and the best practices for assessing them. Through this initiative a series of country case studies on measuring and assessing forest degradation from across the globe were collected. They can be
15

accessed at: “http://www.fao.org/forestry/cpf/forestdegradation/en/” Table 4 offers some examples of country approaches to measure forest degradation, based on these FAO-FRA reports.
Mapping forest degradation with remote sensing data is more challenging than mapping deforestation because the degraded forest is a complex mix of different land cover types
(vegetation, dead trees, soil, shade) and the spectral signature of the degradation changes quickly
(i.e., < 2 years) (Souza et al.
2009). High spatial resolution sensors such as Landsat, ASTER and
SPOT have been mostly used so far to address forest degradation. However, very high resolution satellite imagery, such as Ikonos or Quickbird, and aerial digital imagery acquired with videography have been used as well. Methods for mapping forest degradation range from simple image interpretation to highly sophisticated automated algorithms (GOFC-GOLD 2009).
There are several methods to evaluate forest degradation with remote sensing:

Direct detection of degradation processes (forest canopy damage)

Indirect approaches (observe human infrastructure)

Fire monitoring
Country Remote
Sensing
Mexico
Field data collection
X
Combination of both
Methodological details Forest indicators
Source
Mexico
Brazil
X
X
Ca. 25,000 1-ha plots established, of which 23,000 measured; 20% re-measured every year.
Forest disturbance: intact forest, secondary tree dominated; secondary shrub dominated
Relationship between
MODIS-derived NDVI values and aerial biomass volume derived from the NFI.
Relationship between spectral mixing analysis and
Aboveground Biomass
(AGB) measured through forest transects.
Comparison of methodologies used in Nepal to measure degradation
De Jong et al.
2010
Tovar 2009
Souza et al.
(2009)
Nepal X
DRC X
X X
Field measuring of forest degradation permanent plots. using
Acharya and Dangi
(2009)
Musampa
(2009)
Table 4 : Country examples to measure forest degradation
2.2.1
Direct methods
In the direct method, and under a degradation definition based on changes in carbon stocks, forest canopy gaps, small clearings, and the structural forest changes resulting from disturbance are the features of interest to be enhanced and extracted from the satellite imagery. Among the most classically used techniques are: i) Visual interpretation, which can easily detect canopy damage
16

areas in very high spatial resolution imagery; ii) automatic segmentation; iii) spectral mixing analysis for logging disturbances (Asner et al . 2005, Oliveira et al . 2007) and fire (Souza et al .
2005); iv) lacunarity indices for canopy structural characterization (Malhi and Román-Cuesta
2008); vi) Hyperspectral automated canopy identification (Palace et al.
2008). Figure 1 shows a couple of examples of direct methods to estimate changes in forest structure through remote sensing. Box 1 offers a brief summary of these techniques.
Visual interpretation : The interpreter does not apply any classification algorithm to the image, s/he simply uses the satellite image as a photograph of the forests and delineates forest canopies / changes and forest types (i.e. similar method to the ones employed for aerial photography interpretation). This technique is particularly useful for high resolution images such as IKONOS or Quickbird.
Automatic segmentation: The interpreter uses an automatic algorithm to divide the satellite image into homogeneous regions which will likely correspond to areas of similar structural properties. Similar forest types are likely to be captured by this technique. Large scale degradation of the size of a forest stand could be identified in this technique.
Spectral mixing analysis: The interpreter uses an automatic algorithm to classify the image into percentages of certain pure elements. This technique considers that each pixel is a combination of certain pure elements (i.e. Soil, Vegetation, Non-Photosynthetic vegetation
(NPV) (i.e branches), Shade, Burnt) that combine to produce a given response per pixel.
Each pixel will be characterized by a percentage of each pure element (i.e. 80% Vegetation,
10% Soil, 5% NPV and 5% Shade) and the interpreter can choose the thresholds of each element that will help classify the image (i.e. those pixels with ≥60% of Vegetation-pure element, will be classified as vegetation cover). Some of these fractions can be used to develop indices such as the Normalized Difference Vegetation Index. This technique has been successfully used to detect degradation by logging and fire, through increases in shade fractions and NPV, and reduction in vegetation fractions.
Lacunarity Indices: This technique assumes that tropical forests have fractal geometry and that they can be classified using fractal mathematical algorithms such as Lacunarity Indices.
This technique uses brightness patterns to define the fractal response of forests and to classify areas with high and low vegetation, and to automatically distinguish between tree crowns and forest gaps. With high resolution images such as IKONOS, this technique offered promising results for identifying tropical Amazonian forest structures, especially through the use of the Index of Translational Homogeneity (ITH), which derives from the fractal behaviour of forest reflectance and allows to estimate the width of the canopy crowns
(in meters). The identification of individual crowns allows to detect vegetation changes and degradation processes.
Automated canopy identification: Like the method above, this technique allows the identification of individual crowns, using a similar approach. It is based on a crown-edge detection algorithm that relies on the image brightness patterns, and allometric equations based on field data.
Box 1: Brief summary of several direct methods that can be used to detect degradation
There are limiting factors to be taken into consideration when mapping direct forest degradation
(Souza et al.
2009). First, it requires frequent mapping, at least annually, because the spatial
17

signatures of the degraded forests change after one year. Additionally, it is important to keep track of repeated degradation events that affect more drastically the forest structure and composition resulting in greater changes in carbon stocks. Second, the human-caused forest degradation signal can be confused with natural forest changes such as wind throws and seasonal changes. Confusion due to seasonality can be reduced by using more frequent satellite observations. Third, all the methods described above are based on optical sensors which are limited by frequent cloud conditions in tropical regions. Finally, higher levels of expertise are required to use the most robust automated techniques requiring specialized software and investments in capacity building.
2.2.2
Indirect methods
The indirect method is useful when degradation intensity is low and the area to assess is large, when satellite imagery is not easily accessible, or when the direct approach cannot be applied for whatever other reason. Figure 2 offers a few examples of indirect methods to evaluate forest degradation. An example of a useful indirect approach is the “intact forest” approach where the spatial distribution of human infrastructures (i.e. roads, population centres) are used as proxies, so that the absence of these are used to identify forest land without anthropogenic disturbance (intact forests) so as to assess the carbon content present in the disturbed and non-disturbed forest lands
(Mollicone et al . 2007; Potopov et al . 2008):

Intact forests: fully-stocked (any forest with tree cover between 10% and
100% but must be undisturbed, i.e. there has been no timber extraction)

Non-intact forests: not fully-stocked (tree cover must still be higher than
10% to qualify as a forest under the existing UNFCCC rules, but in our definition we assume that in the forest has undergone some level of timber exploitation or canopy degradation).
Scenario modelling for forest degradation would be another indirect method which could be applied to estimate both future and historical forest degradation dynamics. Soares-Filho et al.
(2006) published an example of a deforestation modelling approach for the Amazon Basin that produced annual maps of simulated future deforestation under user defined scenarios of highway paving,
Protected Area (PA) networks, PA effectiveness, deforestation rates and deforested land ceilings.
With the right support from field data, a similar modeling approach could be used for
(re)constructing historical and future scenarios of forest degradation.
18

Figure 1 : Spectral mixing analysis (SMA) and estimations of Aboveground Biomass (AGB) as a way to follow the degradation dynamics of Amazonian lowland forests (Souza et al.
2005), and the use of lacunarity analysis and the index of translational homogeneity to estimate crown widths in Amazonian forest landscapes (Malhi and Román-Cuesta 2008).GV: Green Vegetation fraction, NPV: Non-photosynthetic vegetation fraction, NDFI: Normalized
Difference Fraction Index ; ITH: Index of Translational Homogeneity; S: South, W: west)
19

Figure 2 : Estimation of intact and non-intact forests based on areas of influence (buffers) from human infrastructures. The example depicts the evolution of a forest landscape where new roads are built, reducing the total area of intact forests (green grid) (Mollicone et al.
2007). Future deforestation models for the
Amazon basin based on two possible scenarios: Bussines-As-Usual (BAU) (a), and effective governance (b) (Soares-Filho et al.
2006).
20

The estimation of emissions due to historical forest degradation is jeopardized by the aspects already mentioned in previous sections in this paper. In spite of those limitations, the procedure to estimate the emissions from a historical period remains the same:
EF x AD = Carbon emissions due to forest degradation
The estimation of the changes in forest carbon stocks due to historical degradation processes are unlikely to rely on real past data, especially true in non-Annex I countries where no or limited past field data are available and consistent measurements of carbon stock change are not established.
Historic EF will therefore have to be derived from present day data on carbon stock losses due to similar degradation processes. For certain degradation activities, data might be collected from the records of the companies that performed the activities (i.e. company records for wood volume extracted in selective logging activities in the past).
The direct estimation of AD through the identification of forest land uses that remain forest land uses for the historical period (AD) is challenged by the problem of defining a temporal threshold to measure the loss of forest carbon stocks. The estimation of AD to evaluate historical degradation will rely on remote sensing techniques (direct and indirect). Due to the large pixel size of past sensors (i.e. Landsat MSS with 60m of spatial resolution), non-satellite data such as aerial photography, will also be used. When using satellite date, sub-pixel approaches that rely on spectral signatures of different pure elements is preferable than the “whole-pixel” approach, since studies have demonstrated that they are more sensitive than “whole-pixel” classifiers, in detecting degraded forest environments (Asner et al. 2005; Oliveira et al.
2007; Souza et al.
2009).
Moreover, and in spite of its complications, identifying historical degradation drivers is fundamental in order to distinguish between human and non-human disturbances and therefore offer more consistent estimates of forest degradation. In case of doubt, the conservativeness principle, which minimizes the overestimation of emission reductions, must be applied.
3.1. Available methods for assessing historical rates of degradation – selective logging
Degradation of tropical forests can occur by many means ranging from timber harvest and chronic anthropogenic disturbance such as fire to decreased tree seed dispersal via over-hunting of key dispersal species. Pre-anthropogenic rates of degradation are hard to define, while contemporary rates also pose major challenges for most scientific methods. Fire is thought to have been relatively rare in humid tropical forests, and large-scale defaunation likely did not occur in regions like the
Congo until recent times (Wilkie et al . 2000). Of all forms of forest degradation, none has been as widespread geographically as selective logging. ITTO (2005) estimates that 350 million hectares of humid tropical forest are under timber production today, although these figures are likely to be very rough at the global scale.
Whereas deforestation mapping and monitoring have become commonplace, most efforts have not resolved the global geographic extent or intensity of the forest disturbance associated with selective logging. Although timber production rates have been estimated from sawmill, sales, and export statistics (Nepstad et al.
1999), these estimates cannot directly quantify the area of degradation caused by logging. Moreover, single-entry and/or low-intensity (< 5 m
3
ha
-1
) logging may not result in degradation, or it may not be considered so unless it can be assessed in terms of persistent
21

canopy damage and ecosystem fragmentation. Satellite mapping of selective logging began an expansive research phase earlier in this century, with results steadily improving over time (Asner et al.
2002, 2004, Souza et al . 2003, 2005). The first large-scale, high-resolution satellite mapping of selective logging and degradation was published just five years ago for a large portion of the
Brazilian Amazon (Asner et al . 2005). In the past five years, major advances have been made in monitoring logging and degradation across the Legal Amazon, and both the Brazilian space agency and Imazon provide highly transparent reporting of degradation (Monteiro et al . 2010). The results often indicate that selective logging can extend over as large a forested territory as deforestation each year. Other satellite-based work suggests that logging and/or logging concessions are now widespread throughout Africa (Laporte et al . 2007), parts of Oceania (Shearman et al . 2008), and other Amazonian countries (Oliveira et al . 2007). And recently, a first global-scale direct mapping of selective logging in humid tropical forests showed that logging activities strike deep into forest interiors, often far from deforestation fronts (Asner et al.
2009). Globally, about 20%, or more than
3.9 million km 2 , of humid tropical forests are selectively harvested, a value that is in rough agreement with the tropical timber production estimate of 3.5 million km
2
(Putz et al . 2008).
Analysis of historical, let alone current, rates of illegal logging are difficult to come by because most illegal logging operations are carried out in remote areas, at low logging intensity (poaching) and/or in a clandestine manner. Perhaps the best estimates of illegal logging rates come from regional-scale analyses involving comparison of roundwood reaching sawmills and reported fieldbased extraction rates. However, this only works in areas where field-based extraction rates are required. The type of illegal logging detected under these conditions is often associated with leakage around concessions.
3.2 Aerial photography – assessment of structural changes in the canopy from successive campaigns
Aerial photography has played an important role in forest survey (Caylor 2000; Hall 2003) and was a unique means to monitoring canopy condition in detail until the first launch of a high resolution satellite in 1999, namely IKONOS. Aerial photographs can provide structural information of the forest canopy for assessing historical rates of forest degradation. Forest degradation is recognized as structural changes of the canopy from successive aerial photographs.
Selective logging (both legal and illegal) is a cause of forest degradation and it is recognized as new gaps in the canopy. Several methods to identify canopy gaps and their dynamics using aerial photographs were proposed in previous studies. Fox et al.
(2000) mapped forest canopy gaps by interpreting aerial photographs. The methods to detect gaps from multi-temporal digital surface model (DSM), which presents the elevation of canopy surface, derived from stereo-pair aerial photographs (Nakashizuka et al ., 1995; Tanaka and Nakashizuka, 1997; Itaya et al ., 2004; Ticehurst et al ., 2007) were used for long-term canopy dynamics studies. DSM derived from aerial photographs is also available for estimating forest growth from comparison between multi-temporal elevations of canopy surfaces.
Either the stock-difference method or the gain-loss method was recommended for estimating carbon stock change in “Forests Remaining Forests” in the “2006 IPCC Guidelines for Agriculture,
Forestry and Other Land Use (AFOLU)”. The estimation of carbon stock changes would benefit from an evaluation of carbon stock changes of individual trees in order to assess historical rates of forest degradation. Crown areas of individual trees can be estimated from aerial photographs and the assumption of allometry between crown area and carbon stock (or alternatively stem diameter and tree height) of individual trees is needed, although not always existent in tropical forests. For
22

extracting individual crown areas from aerial photographs, the valley-following method (Leckie et al ., 2003, 2004; Gougeon and Leckie, 2006) and the watershed method (Wang et al ., 2004; Hirata et al ., 2009) are applicable after converting aerial photographs into digital format. Box 2 explains these techniques in a bit more detail. Both techniques aid at delineating individual trees in the canopy, and therefore would allow the identification of degradation at the canopy level.
In these methods, however, there is a likelihood that the difference in shade between aerial photographs in different seasons leads to a wrong detection of gaps particularly on steep slopes. In addition, the different altitudes of aircraft when successive aerial photographs were taken often leads to unmatched individual crowns.
Valley-following methodology : In a valley following approach the image structure is considered analogous to image intensity topography. Valleys of shade or lower intensity areas between tree crowns are identified and remaining tree material is outlined into crownlike shapes by a rule-based system. Local-maxima approaches identify the peaks in image intensity representing the location of a tree crown but not its outline. Several methods use local maxima as a starting point and try to find the crown boundary by the intensity structure in the image. Template matching creates a model of what trees should look like under different illumination and viewing conditions and matches these to features on the imagery. This provides a tree location, estimation of crown size, and a model output shape that best fits the real crown.
Watershed methodology : A grey-level image may be seen as a topographic relief, where the grey level of a pixel is interpreted as its altitude in the relief. A drop of water falling on a topographic relief flows along a path to finally reach a local minimum. Intuitively, the watershed of a relief correspond to the limits of the adjacent catchment basins of the drops of water. In image processing, different watershed lines may be computed. In graphs, some may be defined on the nodes, on the edges, or hybrid lines on both nodes and edges. Watersheds may also be defined in the continuous domain. There are many different algorithms to compute watersheds. For a segmentation purpose, the length of the gradient is interpreted as elevation information.
Other methods based on grouping textures (Warner et al. 1999), morphological operators
(Barbezat and Jacot, 1999), and joining convexshaped edges (Brandtberg and Walter, 1998) have also been explored.
Box 2: Brief technical description of the methods used to identify tree canopies.
3.3
The estimation of biomass burning emissions
There is plenty of evidence that wildfire presence, severity and contribution to global emissions have increased since the mid 1980s in most terrestrial ecosystems (Grissino-Mayer and Swetnam,
2000; Pausas, 2004; Flannigan, 2005; Skinner et al.
2006; Westerling et al.
, 2006; Balshi et al.
,
2007). This has been the result of more frequent and intense global climate pressures (i.e. strong El
Niño events, delayed rainy seasons, increased lighting sources) combined with human activities and modified landscapes.
The evaluation of historical forest degradation and deforestation linked to past fire events is, therefore, an urgent task.
23

Satellite systems have proved useful to detect and monitor fire episodes (for further information review GOCF-GOLD 2009). Fire monitoring from satellites falls into three primary categories: detection of active fires, mapping of post fire burned areas (fire scars) and fire characterization (e.g. fire severity, energy released). For the purposes of emission estimation we are primarily interested in the latter two categories.
The effects of fire in forest environments are, however, widely variable and the contribution of biomass burning emissions is still uncertain (Van der Werf et al.
2003; Bowman et al.
2009).
Reasons behind these uncertainties relate to the variables needed to estimate biomass burning emissions. There are two main approaches (GOFC-GOLD 2009):
1) A ‘bottom up’ –indirect- method (Seiler and Crutzen, 1980) as:
L = A × Mb × Cf × Gef Eq. 1 where the quantity of emitted gas or particulate L [g] is the product of the area affected by fire A
[m
2
], the fuel loading per unit area Mb [g m
-2
], the combustion factor Cf, i.e. the proportion of biomass consumed as a result of fire [g g
-1
], and the emission factor or emission ratio Gef, i.e. the amount of gas released for each gaseous specie per unit of biomass load consumed by the fire [g g
-
1
]. The main issue with the indirect method is the large uncertainty associated with the area burned
(A). This is especially true for historical assessment of biomass burning events, where few datasets exist and those available are better suited for estimating trends than absolute burnt area values (i.e.
Global Burned Surface (1982-1999), (Carmona-Moreno et al . 2005). Moreover, since currently operative, automated, high frequency, fire products such as MODIS fire datasets only start in the
2000´s (Roy et al . 2008), historical burnt area estimates must rely on individual characterization of other available satellites (i.e. Landsat, SPOT, IRS). While these systems do provide the required spatial resolution to assess burnt areas, there are currently no systematic products using those data, and issues related to data availability (satellite overpass, cloud, receiving stations) make it challenging, at the current stage, to envisage the development of present or past automatic processes to develop current or historical maps of burnt area at country level.
At least three main problems hinder the identification of burnt areas leading to degradation, when using optical satellites: i) Forest fire degradation does not result in complete forest dieback and its spectral signature is more difficult to identify than deforestation fires. Fires leading to forest degradation are mainly surface fires and ground fires, but under-storey fire episodes are difficult to track. This is particularly true for surface fires. Ground fires are likely to result in tree mortality since they affect tree roots, and might be easier to identify through spectral changes in the canopy cover. ii) There is evidence of long term, delayed, effects of forest fires on tree mortality.
Increases in large tree mortality three years after fire can potentially double current estimates of biomass loss and committed carbon emissions from low-intensity fires in tropical forests (Barlow et al . 2009). iii) Satellite systems for Earth Observation are currently providing data with a wide range of spatial resolutions. In principle, only hyperspatial and high resolution data can provide the sub-hectare mapping required for good area identification of fire-degraded forests
(unavailable for historical assessments). However, higher resolution images have a low temporal resolution (15 - 20 days in the case of Landsat) and non-systematic acquisition
(especially the hyperspatial sensors). Moreover, for technological and commercial reasons hyperspatial sensors acquire data almost exclusively in the visible and near
24

infrared wavelengths, and do not have the spectral bands required for adequately mapping fires and burned areas (e.g. thermal and shortwave infrared) and for their characterization (i.e. middle - infrared).
Another factor that remains difficult to estimate is the burning efficiency (Cf) of fire. Cf is a function of fuel loads and moisture content (availability), fuel types (flammability), fuel arrangement in the landscape (continuous versus fragmented fuels), and forest resilience. Since both fuel availability and flammability strongly depend on climatic conditions and forest resilience is species and site specific (strongly related to soil conditions), estimating Cf from fires resulting in forest degradation is still an open challenge. The total burned area and its spectral response can offer clues of Cf values but the limitations listed above still apply.
Fuel load (Mb) remains an uncertain parameter and has been variously estimated from sample field data, satellite data and models (including those partially driven by satellite data) calculating Net
Primary Production to provide biomass increments and partitioning between fuel classes (Van der
Werf et al ., 2003).
The indirect method is currently the only method that allows the reporting of biomass burning emissions under the Climate Change Convention and under REDD+. However, its main disadvantage is that, L being the result of the multiplication of four independent terms, their uncertainties will propagate into the uncertainty of the estimate L. As a consequence, a precise estimate of L requires a precise estimate of all the terms of equation (1). Moreover, for the indirect method, the computation of the total emissions requires burnt area maps at a spatial resolution which is not currently provided by any of the on-going automatic systems, let alone historical maps.
Furthermore, the areas burnt must be characterised in terms of fire behaviour (surface fires, ground fires, crown fires) and in terms of land use change (fires in forest remaining forest, fires related to deforestation). This information is not routinely available as ancillary information of the systematic global and continental products. Hyperspatial and high resolution satellites are the best fit to detect sub-hectare forest degradation. However, they are acquired at low temporal frequencies and do not count on the spectral bands required for mapping active fires and burnt areas (e.g. thermal and shortwave infrared), as well as fire characterization (i.e. middle - infrared). Conversely, coarse resolution systems do not have the spatial resolution required for sub-hectare mapping (as an example, a single nadir pixel from MODIS covers 6.25 to 100 ha depending on the band), but their daily temporal resolution and multispectral capabilities have allowed the development of several fire-related global, multiannual products in recent years .
Many of the factors required for the indirect method can be obtained from the IPCC default values,
AFOLU Tables 2.4 through 2.6 (IPCC 2006). These values refer to each forest type according to the fire characteristics. The use of default estimates will lead to Tier 1 reporting of biomass burning estimates. Tier 2 methods employ the same general approach as Tier 1 but make use of more refined country-derived emission factors and more refined estimates of fuel densities and combustion factors than those provided in the default tables. Tier 3 methods are more comprehensive and include considerations of the dynamics of fuels (biomass and dead organic matter).
2) A recently proposed alternative is to directly measure the power emitted by actively burning fires and from this estimate the total biomass consumed. The radiative component of the energy released by burning vegetation can be remotely sensed at mid infrared and thermal infrared wavelengths (Ichoku and Kaufman, 2005, Wooster et al . 2005, Smith and Wooster 2005). This instantaneous measure, the Fire Radiative Power (FRP) expressed in Watts [W], has been shown to be related to the rate of consumption of biomass [g/s].
25

This method directly measures the power emitted by actively burning fires and from this estimate the total biomass consumed. Among its advantages, this method provides accurate (i.e. ± 15%) estimates of the rate of fuel consumed and the derived satellite variable to estimate the energy released (FRP) has been shown to be linearly related to the total biomass consumed by fire.
However, the accuracy of the integration of FRP over time to derive Fire Radiative Energy (FRE) depends on the spatial and temporal sampling of the emitted power. Ideally, the integration requires high spatial resolution and continuous observation over time of the burning fire, while the currently available systems provide low spatial resolution and high temporal resolution (geostationary satellites) or moderate spatial resolution and low temporal resolution (polar orbiting systems). For these reasons, direct methods have yet to transition from the research domain to operational application and are not yet useful for biomass burning estimates under REDD+.
Measuring forest degradation and related forest carbon stock changes is more complicated and less efficient than measuring deforestation since the former is based on changes in the structure of the forest that do not imply a change in land use and therefore it is not easily detectable through remote sensing. There is not one method to monitor forest degradation. The choice of different approaches depends on a number of factors including the type of degradation, available (historical) data, capacities and resources and the potentials and limitations of various measurement and monitoring approaches.
Measuring all carbon stock changes caused by forest degradation within a country at the same level of detail and accuracy will likely not be efficient. In particular the considerations of IPCC source category analysis, and the fact that many degradation activities are focused on specific areas within the country help to make the monitoring more targeted and efficient to capture the most important components with priority.
To estimate forest degradation, countries need to assess carbon stock changes (emission factors) and the total area undergoing degradation (activity data), ideally for different types of degradation (i.e. fire, logging, fuel wood harvesting). The assessment of changes in carbon stocks requires consistent ground data while the evaluation of the total area undergoing degradation is more reliably measured through remote sensing for the major degradation processes, in particular for developing countries.
Both current and historic assessments of forest degradation will need to consistently collect data on emission factors and on activity data to report their forest degradation emission estimates. The particular problem of measuring historical degradation is the lack of field based forest data for most non-Annex I countries. As a consequence, many countries will need to derive current values of emission factors to estimate past degradation emissions and to rely more heavily on remote sensing techniques, which did not always offer the right spatial resolution in the earlier decades (i.e. Landsat
MSS were 60m pixel). In general terms, forest degradation areas can be mapped through direct and indirect methods, the former approach is based on direct observations of forest structural changes
(i.e. canopy gaps), while the second considers modelling approaches based on known drivers of forest degradation (i.e. distance to roads and human influenced areas, fires, etc). For historical assessments of degradation, indirect methods might be needed if satellite data cannot detect degraded areas.
The major issue affecting the assessment and reporting of forest degradation emissions is the estimation of its uncertainty. Among the REDD+ activities, both forest degradation and
26

deforestation will require high levels of accuracy and certainty since they are major contributors of countries´ GHG budgets for forests. While consistent measurements of forest carbon stock changes have not been of a high priority in many countries and monitoring programs in the past, this situation is changing now and new investments in systematic forest degradation estimates can help reduce uncertainties even for historical estimates. However, historical degradation estimates will necessarily come with large uncertainties due to the lack of available data to determine their accuracy.
27

Acharya, K., and Dangi, R. 2009. Forest Degradation in Nepal: Review of Data and Methods. FAO
Forest Resources Assessment Programme, Rome, Italy
Asner, G. P., M. Keller, R. Pereira, and J. C. Zweede. 2002. Remote sensing of selective logging in
Amazonia: Assessing limitations based on detailed field observations, Landsat ETM+, and textural analysis. Remote Sensing of Environment 80:483-496.
Asner, G. P., M. Keller, R. Pereira, J. C. Zweede, and J. N. M. Silva. 2004. Canopy damage and recovery after selective logging in Amazonia: Field and satellite studies. Ecological
Applications 14:S280-S298.
Asner, G. P., D. E. Knapp, E. N. Broadbent, P. J. C. Oliveira, M. Keller, and J. N. M. Silva. 2005.
Selective logging in the Brazilian Amazon. Science 310:480-482.
Asner, G., Rudel, T., Aide, M., Defries, R., Emerson, R. 2009. A contemporary assessment of change in humid tropical forests. Conservation Biology 23:1386-1395.
Barbezat, V., Lamas, J., and Jacot, J. 1999. Automated recognition of forest patterns with aerial photographs. Journée scientifique "Télédétection et analyse d'image". LPM-CONF-1999-002
Bowman, D., Balch, J., Artaxo, P., Bond, W., Carlson, J., Cochrane, M., D’Antonio, C., DeFries,
R., Doyle, J., Harrison, S., Johnston, H., Keeley, J., Krawchuk, M., Kull, C., Marston, B.,
Moritz, B., Prentice, C., Roos, C., Scott, A., Swetnam, T., van der Werf, G., Pyne, S., 2009.
Fire in the Earth System. Science 324, 481 – 484.
Brandtberg, T., and Walter, F. 1998. Automated delineation of individual tree crowns in high spatial resolution aerial images by multiple-scale analysis. Machine Vision and Applications, 11,
64–73.
Caylor, J. 2000. Aerial photography in the next decade. Journal of Forestry, 98, 17-19.
De Jong, B., Anaya,C, Masera, O., Olguín, M., Paz, F., Etchevers, J., Martínez, R., Guerrero, G.,
Balbontín, C. 2010. Greenhouse gas emissions between 1993 and 2002 from land-use change and forestry in Mexico. Forest Ecology and Management. In press.
Dwyer, E., J. M. C. Pereira, J.-M. Gre´goire, and C. C. DaCamara (1999), Characterization of the spatio-temporal patterns of global fire activity using satellite imagery for the period April
1992 to March 1993. Journal of. Biogeography, 27, 57–69.
Eva, H., Carboni, S., Achard, F., Stach, N., Durieux, L., Faure, J.-F., Mollicone, D. 2009.
Monitoring forest areas from continental to territorial levels using a sample of medium spatial resolution satellite imagery. Journal of Photogrammetry and Remote Sensing
Volume 65, Issue 2, March 2010, Pages 191-197.
FAO. 2002. Food and Agriculture Organization. Proceedings: Second Expert Meeting on
Harmonizing Forest-related Definitions for Use by Various Stakeholders. Rome, 11-13
September 2002. Rome. http://www.fao.org/docrep/005/y4171e/y4171e00.htm
FAO. 2006a. Food and Agriculture Organization .Global Forest Resources Assessment
2005.Progress towards sustainable forest
FAO.2006b.
Summaries of FAO’s work in forestry. Rome, Italy.
<http://www.fao.org/forestry/foris/webview/forestry2/
FAO. 2007. Food and Agriculture Organization. State of the World's Forests. United Nations,
Rome. Available: http://www.fao.org/docrep/009/a0773e/a0773e00.HTM.
Flannigan, M.D., Logan, K.A., Amiro, B.D., Skinner, W.R., and Stocks, B.J. 2005. Future area burned in Canada. Climatic Change, 72,1-16.
Freeborn, P.H., Wooster, M.J., Hao, W.M., Ryan, C.A.,Nordgren, B.L. Baker, S.P. and Ichoku, C.
2008. Relationships between energy release, fuel mass loss, and trace gas and aerosol emissions during laboratory biomass fires, J. Geophys. Res., 113, D01102, doi:10.1029/2007JD008489
28

Fox, T.J., Knutson, M.G., and Hines, R.K. 2000. Mapping forest canopy gaps using air-photo interpretation and ground surveys. Wildlife Society Bulletin, 28, 882-889.
Giglio, L., Csiszar, I., and Justice, C. 2006 Global distribution and seasonality of active fires as observed with the Terra and Aqua Moderate Resolution Imaging Spectroradiometer
(MODIS) sensors. Journal of Geophysical Research G02016, doi:10.1029/2005JG000142
GOFC-GOLD. 2009. A sourcebook of methods and procedures for monitoring and reporting anthropogenic greenhouse gas emissions and removals caused by deforestation, gains and losses of carbon stocks in forest remaining forests, and forestation.GOFC-GOLD Report version COP15-1.Available at www.gofc-gold.uni-jena.de/redd/.
Gougeon, F.A., and Leckie, D.G. 2006. The individual tree crown approach applied to Ikonos images of a coniferous plantation area. PE & RS 72, 1287-1297.
Grissino-Mayer, H. and Swetnam, T. 2000. Century-scale climate forcing of fire regimes in the
American Southwest. Holocene, 10, 207-214.
Hall, R.J. 2003. The roles of aerial photographs in forestry remote sensing image analysis. In
Remote Sensing of Forest Environments: Concepts and case studies. Wulder, M.A. and
Franklin, S.E. (eds.), Kluwer Academic Publishers, Boston, Dordrecht and London, 47-75.
Hirata, Y., Sakai, A., and Tsuboto, Y. 2009. Allometric models of DBH and crown area derived from QuickBird panchromatic data in Cryptomeria japonica and Chamaecyparis obtusa stands. International Journal of Remote Sensing, 30, 5071-5088.
Ichoku, C., and Kaufman, Y. 2005. A method to derive smoke emission rates from MODIS Fire
Radiative Energy Measurments, IEEE Transaction on Geosciences and Remote Sensing, 43,
2636-2649DOI 10.1109/TGRS.2005.857328.
IPCC. Intergovernmental Panel on Climate Change. 2003. Good Practice Guidance on Land Use,
Land-Use Change and Forestry. Eggleston, H.S.,Buendia, L., Miwa, K., Ngara, T. and
Tanabe, K. (eds.). National Greenhouse Gas Inventories Programme. Institute for Global
Environmental Strategies (IGES). Japan. http://www.ipccnggip.iges.or.jp/public/gpglulucf/gpglulucf_contents.html
IPCC. Intergovernmental Panel on Climate Change. 2007. Climate Change 2007: Mitigation.
Contribution of Working Group III to the Fourth Assessment Report of the
Intergovernmental Panel on Climate Change. Cambridge, United Kingdom and New York,
NY, USA. http://www.ipcc.ch/ipccreports/ar4-wg3.htm
Itaya, A., Miura, M., and Yamamoto, S. 2004. Canopy height changes of an old growth evergreen broad-leaved forest analyzed with digital elevation models. Forest Ecology and
Management, 194, 403-411.
ITTO. 2005. International Tropical Timber Organization. 2005. Status of tropical forest management 2005. Available: http://www.itto.or.jp/live/PageDisplayHandler?pageId=270.
Laporte, N. T., J. A. Stabach, R. Grosch, T. S. Lin, and S. J. Goetz. 2007. Expansion of Industrial
Logging in Central Africa. Science 316:1451.
Leckie, D.G., Gougeon, F.A., Walsworth, N., and Paradine, D. 2003. Stand delineation and composition estimation using semi-automated individual tree crown analysis.
Remote
Sensing of Environment 85, 355-369.
Leckie, D.G., Jay, C., Gougeon, F.A., Sturrock, R.N., Paradine, D. 2004. Detection and assessment of trees with Phellinus weirii (laminated root rot) using high resolution multi-spectral imagery. International Journal of Remote Sensing 25, 793-818.
Lund, H. 2009. What is a degraded forest. Forest Information Services. Gainesville, VA. USA. http://home.comcast.net/~gyde/2009forestdegrade.doc
MEA. Millenium Ecosystem Assessment. 2005. Ecosystems and human well-being: Scenarios.
Findings of the Scenarios Working Group. Island Press, Washington D.C.
29

Mollicone D., Achard F., Federici S., Eva H.D., Grassi G., Belward A., Raes F., Seufert G., Stibig
H.-J., Matteucci G., Schulze E.-D. 2007. An incentive mechanism for reducing emissions from conversion of intact and non-intact forests. Climatic Change 83: 477–493.
Monteiro, A., Cardoso, D., Conrado, D., Souza Jr., C. and Veríssiom, A. 2010. Transparency in
Forest Management: State of Pará. Imazon Bulletin, 16p.
Murdiyarso, D., Margaret Skutsch, Manuel Guariguata, Markku Kanninen, Cecilia Luttrell, Pita
Verweij and Osvaldo Stella. 2008. Measuring and monitoring forest degradation for REDD
Implications of country circumstances CIFOR InfoBrief. No.16, www.cifor.cgiar.org.
Musampa, C. 2009. La dégradation des forêts en République Démocratique du Congo, FRA
Document de travail 169. http://www.fao.org/forestry/cpf/forestdegradation/en/
Nabuurs, G.J., O. Masera, K. Andrasko, P. Benitez-Ponce, R. Boer, M. Dutschke, E. Elsiddig, J.
Ford-Robertson, P. Frumhoff, T.Karjalainen, O. Krankina, W.A. Kurz, M. Matsumoto, W.
Oyhantcabal, N.H. Ravindranath, M.J. Sanz Sanchez, X. Zhang, 2007: Forestry. In Climate
Change 2007: Mitigation. Contribution of Working Group III to the Fourth Assessment
Report of the Intergovernmental Panel on Climate Change [B. Metz, O.R. Davidson, P.R.
Bosch, R. Dave, L.A. Meyer (eds)], Cambridge University Press, Cambridge, United
Kingdom and New York, NY, USA.
Nakashizuka, T., Katsuki, T., and Tanaka, H. 1995. Forest canopy structure analyzed by using aerial photographs. Ecological Research, 10, 13-18.
Negrón-Juárez, R., Chambers, J., Guimaraes, G., Zeng, H., Raupp, C., Marra, D., Ribeiro, G.,
Saatchi, S., Nelson, B., and Higuchi, N. 2010. Widespread Amazon forest tree mortality from a single cross-basin squall line event. Geophysical Research Letters, l16701-116706
Nepstad, D. C., A. Verissimo, A. Alencar, C. Nobre, E. Lima, P. Lefebvre, P. Schlesinger, C.
Potter, P. Moutinho, E. Mendoza, M. Cochrane, and V. Brooks. 1999. Large-scale impoverishment of Amazonian forests by logging and fire. Nature 398:505-508.
Oliveira, P. J. C., G. P. Asner, D. E. Knapp, A. Almeyda, R. Galvan-Gildemeister, S. Keene, R.
Raybin, and R. C. Smith. 2007. Land-use allocation protects the Peruvian Amazon. Science
317:1233-1236.
Palace, M., Keller, M., Asner, G., Hagen, S., and B. Braswel. 2008. Amazon Forest Structure from
IKONOS Satellite Data and the Automated Characterization of Forest Canopy Properties.
Biotropica 40: 141–150.
Pausas, J. 2004. Changes in fire and climate in the eastern Iberian Peninsula (Mediterranean Basin).
Climatic Change, 63, 337-350.
Peres, C., Barlow, J., and Laurance, W. 2006. Detecting anthropogenic disturbance in tropical forests. TRENDS in Ecology and Evolution 21, 227-229.
Potapov P., Yaroshenko A., Turubanova S., Dubinin M., Laestadius L., Thies C., Aksenov D.,
Egorov A., Yesipova Y., Glushkov I., Karpachevskiy M., Kostikova A., Manisha A.,
Tsybikova E., Zhuravleva I. 2008. Mapping the World’s Intact Forest Landscapes by
Remote Sensing. Ecology and Society 13:51
Putz, F. E., P. A. Zuidema, M. A. Pinard, R. G. A. Boot, J. A. Sayer, D. Sheil, P. Sist, Elias, and J.
K. Vanclay. 2008. Improved tropical forest management for carbon retention. PLoS Biology
6:1368-1369.
Roy, D.P. and Landmann, T. 2005. Characterizing the surface heterogeneity of fire effects using multi-temporal reflective wavelength data, International Journal of Remote Sensing,
26:4197-4218.
Shearman, P., J. Bryan, J. Ash, P. Hunnam, B. Mackey, and B. Lokes. 2008. The State of the
Forests of Papua New Guinea. University of Papua New Guinea Press, Port Moresby, Papua
New Guinea.
30

Simula, M. 2009. Towards defining forest degradation: comparative analysis of existing definitions.
Forest Resources Assessment. Pp 57. Working Paper 154. FAO, Rome. ftp://ftp.fao.org/docrep/fao/012/k6217e/k6217e00.pdf
Skinner, W., Shabbar, A., Flannigan, M. & Logan, K. 2006. Large forest fires in Canada and the relationship to global sea surface temperatures. Journal of Geophysical Research, doi:
10.1029/2005JD006738.
Smith A.M.S., and Wooster, M.J. 2005. Remote classification of head and backfire types from
MODIS fire radiative power observations, International Journal of Wildland Fire, 14, 249-
254.
Soares-Filho, B.S., Nepstad, D.C., Curran, L.M., Cerqueira, G.C., Garcia, R.A., Ramos, C.A., Voll,
E., McDonald, A., Lefebvre, P. & Schlesinger, P. 2006. Modelling conservation in the
Amazon basin. Nature, 440, 520-523.
Souza, C., L. Firestone, L. M. Silva, and D. Roberts. 2003. Mapping forest degradation in the
Eastern Amazon from SPOT 4 through spectral mixture models. Remote Sensing of
Environment 87:494-506.
Souza, C., D. A. Roberts, and M. A. Cochrane. 2005. Combining spectral and spatial information to map canopy damages from selective logging and forest fires. Remote Sensing of
Environment 98:329-343.
Souza, C., Cochrane, M., Sales, M., Monteiro, A., Mollicone, D. 2009. Integrating forest transects and remote sensing data to quantify carbon loss due to forest degradation in the Brazilian
Amazon, FRA Working Paper 161. http://www.fao.org/forestry/cpf/forestdegradation/64442/es/
Stach, N., Salvado, A., Petit, M., Faure, J. F., Durieux, L., Corbane, C., Lasselin, D. and Deshayes.
2009. Land use monitoring by remote sensing in tropical forest areas in support of the Kyoto
Protocol: the case of French Guiana International Journal of Remote Sensing 30, 5133 –
5149.
Tanaka, H., and Nakashizuka, T. 1997. Fifteen years of canopy dynamics analyzed by aerial photographs in a temperate deciduous forest, Japan. Ecology, 78, 612-620.
Tichehurst, C., Phinn, S., and Held, A. 2007. Using multitemporal digital elevation model data for detecting canopy gaps in tropical forests due to cyclone damage: An initial assessment. Austral
Ecology, 32, 59-69.
Tovar, C. 2009. Análisis del índice normalizado de la vegetación (NDVI) para detección de degradación de la cubierta forestal en México 2008-2009, FRA Documento de trabajo 173 . http://www.fao.org/forestry/cpf/forestdegradation/64442/en/
Wang, L., Gong, P., and Biging, G.S. 2004a. Individual tree-crown delineation and treetop detection in high-spatial-resolution aerial imagery. PE&RS, 70, 351-357
Warner, T.A., Lee, J.Y., and McGraw, J.B., 1999. Delineation and identification of individual trees in eastern deciduous forest. In: Hill, D.A. and Leckie, D.G., Editors, 1999. Proc. international forum: automated interpretation of high spatial resolution digital imagery for forestry,
Canadian Forest Service, Pacific Forestry Centre, Victoria, British Columbia, February 10–12,
1998, Canadian Forest Service, Pacific Forestry Centre, Victoria, BC, pp. 81–91.
Westerling, A., Hidalgo, H., Cayan, D., Swetnam, T. 2006. Warming and early spring increase western US forest wildfire activity. Science, 313, 940-943.
Wilkie D., Shaw E., Rotberg F., Morelli G., Auzel, P. 2000. Roads, Development, and Conservation in the Congo Basin. Conservation Biology, 14, 1614-1622.
31