See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/290427306 Reliability of downscaling rainfall data in the estimation of rainfall trends : A case study Article · January 2012 CITATIONS READS 0 870 3 authors, including: Bapuji Rao Bodapati Pramod Parameswaran Central Research Institute for Dryland Agriculture, India Willis Towers Watson 45 PUBLICATIONS 425 CITATIONS 29 PUBLICATIONS 156 CITATIONS SEE PROFILE Some of the authors of this publication are also working on these related projects: Monsoon Mission II View project NICRA Agrometerology View project All content following this page was uploaded by Pramod Parameswaran on 14 January 2016. The user has requested enhancement of the downloaded file. SEE PROFILE Journal of Agrometeorology [ISSN: 0972-1665]4, Volume-14, Special Issue 2012 B. Bapuji Rao, V. P. Pramod and Vum Rao [Page 162-168] Reliability of downscaling rainfall data in the estimation of rainfall trends : A case study B. BAPUJI RAO, V. P. PRAMOD AND VUM RAO All India Coordinated Research Project on Agrometeorology, Central Research Institute for Dryland Agriculture Santoshnagar, HYDERABAD – 500 059 ABSTRACT Precipitation extremes have become more intense during the twentieth century, and that this trend is likely to persist in response to the continuing global warming and climate change. There is substantial climatological evidence to support this statement. However an understanding on these issues is limited by spatial scaling. Most evidence of past trends comes from rain gauge data, whereas the future trends are produced by climate models, which rely on gridded data. In this background, an attempt has been made to study the trends in rainfall on regional scale, over Andhra Pradesh, for the last 35 years (1971-2005) with gridded data (of different resolutions) and observed data, and thereby comparing the results. Results showed that with 0.5o x 0.5o IMD gridded data, 0.25o x 0.25o APHRODITE gridded data and observed mandal level data an increasing trend in rainfall was examined in some districts. No decreasing trend in rainfall was observed with the IMD data whereas a decreasing trend in rainfall was observed in Mahaboobnagar and Adilabad districts with APHRODITE. Decreasing trend in rainfall was identified with the observed data at several locations in majority of the districts. The difference in trend results carried out with mandal level and gridded data reflects the drawbacks in using gridded data in projecting the future trend in rainfall and shows how it can lead to errors. Keywords: Rainfall, Mann-Kandall test, Rainfall trend, Climate model. km falls between 12o 41’ North to 19o 50’ North latitude and 76o 45’ East to 84o 40’ East longitude situated on the south-eastern coast of India (Fig. 1). Considering the geographical position, Andhra Pradesh is divided into three distinct regions namely Coastal, Telengana and Rayalaseema. The rainfall of Andhra Pradesh is inclined by both the South-West and North-East monsoons. The state receives a mean annual rainfall of 910 mm. Major portion (67%) of the rainfall is contributed by SouthWest monsoon (June-Sept) followed by North-East monsoon (Oct-Dec- 24%). The rest (9%) of the rainfall is received during the winter (Jan-Feb) and summer (Mar-May) months. INTRODUCTION Trends and variations in rainfall have significant social and political impacts as Indian agriculture is largely controlled by the Indian summer monsoon rainfall. The extremes in precipitation have become more intense during the twentieth century, and that this trend is likely to continue with continued global warming and climate change (Karl and Knight, 1998; Zwiers and Kharin, 1998; Groisman et al., 1999; Kharin and Zwiers, 2000; Meehl et al., 2000; Frich et al., 2002; Kiktev et al., 2003; Hegerl et al., 2004 and Goswami et al., 2006). Future projections produced by global and regional climate models offer a way to characterize any trends in extreme behaviour. However, there is an issue of spatial scaling and ways to compare the output of regional and climate models with historical data. Climate model data represent an aggregate over a grid box whereas historical data are collected from raingauge at specific point locations. In this context, an attempt is made here to investigate the rainfall trends for a period of 35 years (1971-2005) with gridded data and observed data on regional scale, over Andhra Pradesh, and thereby comparing the results of both. The variability in inter-district rainfall is high and six coastal districts and two Telengana districts receive more than 1,000 mm per annum, while the districts of Kurnool, Anantapur and Y.S.R.Kadapa register rainfall below 700 mm. The coastal district of East Godaveri receives maximum annual rainfall (1133 mm) while Anantapur (Rayalaseema region) receives the minimum (548 mm). Because of this wide spatial variation in annual rainfall an attempt has been made to study the trends in rainfall over a period of 35 years with data available from local sources (Revenue records), closely located weather stations and gridded data with different spatial resolutions and a comparison is made. STUDY AREA Andhra Pradesh is India’s fifth largest state (in terms of area) spreading over an area of 2, 76, 754 sq. 162 Journal of Agrometeorology [ISSN: 0972-1665]4, Volume-14, Special Issue 2012 B. Bapuji Rao, V. P. Pramod and Vum Rao [Page 162-168] Fig. 1 Images showing the study area. rainfall was carried out with Mann-Kendall trend test using Toolkit software V 1.02 (CRC for Catchment Hydrology, Australia, 2005). Geographical Information system (GIS) tool is used in preparing the plots in response to the trend result. DATA AND METHODOLOGY Datasets from three different sources are used in this study, the mandal level (1128 mandals) daily rainfall from Directorate of Economics and Statistics, Govt. of Andhra Pradesh for the period of 1971-2010 were the first source and the distribution of raingauge stations is presented in Fig. 2a. India Meteorological Department (IMD) daily rainfall data on 0.5o x 0.5o spatial resolution for the period 1951-2007 and Asian Precipitation – Highly Resolved Observational Data Integration Towards Evaluation of the Water Resources (APHRODITE’s water resources) 0.25o x 0.25o gridded daily rainfall data are the other two sources. Both IMD and APHRODITE used a well-tested interpolation method (Shepard, 1968) to interpolate the station data into regular grids. In order to maintain uniformity, the length of all datasets is restricted to 35 years from 1971 to 2005 to study the trends in daily rainfall. Mandal and grid-wise annual rainfall data was arrived by summing the daily rainfall data. A trend analysis of the annual RESULTS AND DISCUSSION The data points of IMD and APHRODITE representing Andhra Pradesh with 0.5o x 0.5o and 0.25o x 0.25o are 102 and 382 grid points, respectively (Fig 2b and 2c). In both gridded dataset there is no grid point to represent the Hyderabad district. Mandals with rainfall data less than 10 years are not considered in carrying out the trend analysis. Mann-Kendall test with mandal level data shows that out of 1039 mandals, 752 mandals showed non significant trend, 55 mandals increasing trend and 232 mandals decreasing trend (Table 1). In Mahaboobnagar, Medak, Nalgonda, Prakasham and West Godavari districts none of the mandals showed any increasing trend in rainfall. About 94% of total geographical area of Chittor district does not show any 163 Journal of Agrometeorology [ISSN: 0972-1665]4, Volume-14, Special Issue 2012 B. Bapuji Rao, V. P. Pramod and Vum Rao [Page 162-168] showed increasing APHRODITE data. significant trend. Medak is the only district in which 45% of the total geographical area showed non significant trend. In this district 35% of the total geographical area showed decreasing trend at 99% significant level. The remaining 14% and 6% of the area showed declining trend at 95% and 90% significant level respectively. This shows that decreasing trend in rainfall is pronounced more in Medak district for the last 35 years when compared to other distrcits. Karimnagar falls second in this category after Medak. About 46% of the total geographical area of Karimnagar district shows no significant trend, whereas 21% shows a decreasing trend at 99% significant level, 8% at 95% significant level and the remaining 7% shows a negative trend at 90% significant level. Among the 102 grid points of 0.5o x 0.5o gridded data, 15 points showed increasing trend over the state which includes Srikakulam, Khammam, Guntur, Prakasam, Nellore, Cuddapah, Kurnool and Mahbubnagar districts (Fig. 3b). There is no decreasing trend in rainfall captured with the IMD dataset. The results are surprising when trend analysis was carried out with 0.25o x 0.25o APHRODITE data. If increasing trend was able to find out with 0.5o x 0.5o IMD data, one would naturally expect an increasing trend with more fine resolution datasets in the same districts. But not even a single point of the 0.25o x 0.25o gridded data depicted any significant trend in these districts, except Guntur (Fig. 3c). In fact with mandal level data, seven mandals in Kurnool, one in Guntur and six in Khammam showed increasing trend. Mahaboobnagar, Guntur, Rangareddy, Adilabad, Visakhapatnam and Vizianagaram are the districts in which increasing/decreasing trend was observed with the 0.25o x 0.25o gridded data. Mahaboobnagar and Adilabad are the only two districts in which decreasing trends were noticed with APHRODITE data. In comparison with IMD gridded data, APHRODITE could able to capture decreasing trends in rainfall. Three grid points in Mahaboobnagar and a grid point in Adilabad district showed decreasing trend. The decreasing trend in Mahaboobnagar that was captured with fine resolution gridded dataset is also reflected with the mandal level data as well. In the remaining districts, along with the IMD gridded data, 0.25o x 0.25o gridded data also could not express any significant trends in the rainfall. The accuracy in downscaling data to different scales on a state wide basis is show in Table 2. With mandal level data 5% of the total geographical area showed increasing trend, 17% showed decreasing trend and the remaining 71% showed no significant trend in rainfall. About 15% of the area showed increasing trend in rainfall with 0.5o x 0.5o IMD gridded data whereas only 2% of the total geographical area trend with 0.25o x 0.25o The discrepancy in trend results with 0.5o x 0.5o IMD and 0.25o x 0.25o APHRODITE data may be because of the difference in number of raingauge stations, even though it uses the same interpolation technique (Shepard, 1968) in converting the station data into regular grids. The difference in trend results carried out with mandal level data and gridded dataset reflects the limitation in using grid data for trend analysis and shows how it can give negative results. Also depending upon the trends a rainfall examined with historic data, a suspicion may arise on the the future projections in rainfall trends using the gridded data. CONCLUSIONS Using the Mann-Kandall trend test, an increasing trend in rainfall was observed over a period of 35 years (1971-2005) with 0.5o x 0.5o IMD gridded data in eight districts districts (Srikakulam, Khammam, Guntur, Prakasam, Nellore, Cuddapah, Kurnool and Mahbubnagar) of Andhra Pradesh. When the same trend test was carried out with much more fine resolution 0.25o x 0.25o APHRODITE gridded data no significant trend in rainfall was observed in these districts, except Guntur. In fact it shows an increasing trend in rainfall over Rangareddy, Visakhapatnam and Vizianagaram. This difference in trend results with the two gridded dataset (IMD and APHRODITE) may be because of the difference in number of rain gauge stations, even though it uses the same interpolation technique in converting the station data into regular grids. With the mandal level data, in addition to the increasing trend, a declining trend in rainfall is also captured. No increasing trend in rainfall was observed in Mahaboobanagr, Medak, Nalgonda, Prakasham and West Godavari districts. IMD 0.5o x 0.5o grid data could not sense any decreasing trend. The difference in trend results carried out with mandal level data and gridded dataset reflects the limitation in using grid data for trend analysis and shows how it can lead to errors in climate change projections. Since rainfall is a crucial agroclimatological factor and its analysis is an important factor for agricultural planning in India, it is worthwhile to use the data collected from local sources, wherever available. ACKNOWLEDGEMENTS This research is supported by National Initiative for Climate Resilient Agriculture (NICRA), Government of India and authors are thankful to Mrs. Latha and Ms. Pallavi for their help in analysis and preparing the GIS maps. 164 Journal of Agrometeorology [ISSN: 0972-1665]4, Volume-14, Special Issue 2012 B. Bapuji Rao, V. P. Pramod and Vum Rao [Page 162-168] Fig. 2 Location maps of 1128 raingauge stations in Andhra Pradesh (a), IMD grid points on 0.5o x 0.5o (b) and APHRODITE’s 0.25o x 0.25o scale in (c) in Andhra Pradesh. 165 Journal of Agrometeorology [ISSN: 0972-1665]4, Volume-14, Special Issue 2012 B. Bapuji Rao, V. P. Pramod and Vum Rao [Page 162-168] Fig. 3 Annual rainfall trends using mandal level data (a), IMD 0.5o x 0.5o gridded data (b) and APHRODITE’s 0.25o x 0.25o gridded data 166 Journal of Agrometeorology [ISSN: 0972-1665]4, Volume-14, Special Issue 2012 B. Bapuji Rao, V. P. Pramod and Vum Rao [Page 162-168] Table 1: Area under increasing/decreasing trend in annual rainfall in different districts of Andhra Pradesh based on block level data District Total Increasing Increasing Increasing Decreasing Decreasing Decreasing Geographi Trend Trend Trend Trend Trend Trend No Not Significant cal Area of (99% (95% (90% (99% (95% (90% Available District Significant Significant Significant Significant Significant Significant Trend (Acres) Level) Level) Level) Level) Level) Level) Adilabad 3447417 8% 3% 1% Nil 5% 4% 67% 11% Anantapur 4024417 Nil 2% 3% 2% 2% 4% 87% Nil Chittoor 3116714 Nil 4% Nil Nil 1% 1% 94% Nil Y.S.R.Kadapa 3171945 Nil 5% Nil 2% Nil 1% 87% 4% East Godavari 2280711 Nil 9% Nil 9% 10% 2% 67% 3% Guntur 2250910 6% 2% 7% 16% 1% 1% 56% 11% Hyderabad 36378 Nil Nil Nil Nil Nil Nil Nil Nil Karimnagar 2427396 2% Nil Nil 21% 8% 7% 46% 16% Khammam 3352217 3% 6% 2% Nil 4% 7% 69% 8% Krishna 1852116 Nil Nil 2% 7% 18% 17% 49% 6% Kurnool 3684690 Nil 7% 5% 3% 1% 1% 81% 2% Mahaboobnagar 3882951 Nil Nil Nil 5% 7% 7% 81% Nil Medak 1988944 Nil Nil Nil 35% 14% 6% 45% Nil Nalgonda 3034113 Nil Nil Nil 11% 3% 8% 68% 10% S.P.S.R.Nellore 2772851 Nil 1% 2% 2% 9% Nil 83% 2% Nizamabad 1657969 5% Nil 2% 5% 7% 13% 69% Nil Prakasham 3655788 Nil Nil Nil 9% 15% 1% 50% 24% Ranga Reddy 1584109 Nil 1% 2% Nil Nil 6% 88% 3% Srikakulam 1257898 Nil Nil 3% 5% 7% 8% 78% Nil Visakhapatnam 2396860 1% 1% 2% Nil 1% 5% 87% 4% Vizianagaram 1288026 2% Nil 6% 14% 4% Nil 69% 5% Warangal 2679450 Nil 2% 1% 12% 5% 10% 68% 3% West Godavari 1606901 Nil Nil Nil 9% 19% 3% 68% Nil 167 Journal of Agrometeorology [ISSN: 0972-1665]4, Volume-14, Special Issue 2012 B. Bapuji Rao, V. P. Pramod and Vum Rao [Page 162-168] Table 2: Accuracy in downscaling data to different scales on a state wide-basis Data Source % area in the total geographical area showing Increasing Trend Decreasing Trend No Trend Mandal level data 5 17 71 0.5o x 0.5o 15 - 85 0.25o x 0.25o 2 1 97 Karl TR and Knight RW (1998) Secular trends of precipitation amount, frequency and intensity in the United States. Bull Am Meteorol Soc., 79: 231-241. REFERENCES Frich P, Alexander LV, Della-Marta P, Gleason B, Haylock M, Klein Tank AMG and Peterson T (2002) Observed coherent changes in climatic extremes during the second half of the twentieth centuary. Clim Res., 19: 193-212. Kharin VV and Zwiers F W (2000) Changes in the extremes in an ensemble of transient climate stimulations with a coupled atmosphere-ocean GCM. J Climate, 13: 3760-3788. Goswami BN, Venogopal V, Sengupta D, Madhusoodanan MS and Prince K Xavier (2006) Incresing trend of extreme rain events over India in warming environment. Science, 314: 1442 1445. Kiktev D, Sexton D, Alexander L and Folland C (2003) Comparison of modelled and observed trends in indicies of daily climate extremes. J Climate., 16: 3560-3571. Groisman PY, Karl TR, Easterling DR, Knight RW, Jamason PF, Hnnessy KJ, Suppiah C, Wibig J, Fortuniak K, Razuvaev NV, Douglas A, Forland E and Xhai PM (1999) Changes in the probability of heavy precipitation: Important indicators of climate change. Climate Change, 42: 243-283. Meehl GA, Zwiers F, Evans J, Knutson T, Mearns L and Whetton P (2000) Trends in extreme weather and climate events: Issues related to modelling extremes in projections of future climate change. Bull Am Meteorol Soc 81: 427-436. Shepard D (1968) A two dimensional interpolation function for irregularly spaced data. Proc. 1968 ACM Nat. Conf 517-524. Hegerl GC, Zwiers FW, Stott PA, Kanamitsu M, Kistler R, Collins W, Deaven D, Gandin L, Iredell M, Saha S, White G, Woolen J, Zhu Y, Leetmaa A and Reynolds R (2004) Detectability of anthropogenic changes in annual temperature and precipitation extremes. J. Climate, 17: 36833700. Zwiers FW and Kharin VV (1998) Changes in the extremes of climate simulated by CCC GCM2 under CO2-doubling. J Climate, 11: 2200-2222. 168 View publication stats