CLIMATE CHANGE STUDY ON HYDROMETEOROLOGICAL VARIABLES ON DIFFERENT SPATIO-TEMPORAL SCALES

Abstract

Climate change and agriculture are interrelated processes, both of which take place on a global scale. IPCC (2007) reported that there is a probability of 10-40% loss in crop production in India by 2080-2100 due to global warming despite beneficial aspects of increased CO2. The warming of indian sub-continent is reported to be between I °C and 2 °C by 2030 (iPCC, 2007). This will lead to a more vigorous hydrological cycle; this translates into prospects for more severe droughts and/or floods in some places and less severe droughts and/or floods in other places, with changes in precipitation and evapotranspiration rates. Therefore, there is a pressing need to study the trends and future forecast of hydrometeorological variables. Comprehensive knowledge of the characteristics of hydrometeorological variables including their variation in past and future is essential for proper planning and, in turn, overall development of the river basin. This study is planned to quantify the effect of climate change on hydrometeorological variables for whole MP and for Tons River Basin with specific objectives to (a) study the temporal and spatial variability in precipitation and temperature for MP; (b) analyze trends in various hydrometeorological variables i.e. reference evapotranspiration, temperature, wind speed, relative humidity, precipitation and strearnfiow in the Tons River Basin; (c) study the association of El Niño Southern Oscillation (ENSO) with precipitation and streamfiow in the Tons River basin; (d) study the spatio-temporal variability and forecasting of drought in the Tons River basin; and (e) derive local-scale hydrometeorological variables (precipitation, temperature and streamfiow) of the Tons River Basin by employing statistically downscaling technique using Global Climate Model (GCM) output. LONG TERM TREND ANALYSIS At District Level The spatial and temporal variability of precipitation and temperature is investigated for entire MP over the period of 102 years (1901-2002) on annual and seasonal basis. MK test is directly applied to the original time series, which are found serially independent. However, the series, which are found serially correlated, pre-whitening is used before applying the MK test to detect the trends in time series. Sen Slope estimator test is used to detect the magnitude of trends. A significant decreasing trend in annual precipitation was observed only at three stations (Balaghat, Dindori, and Shandol), when subjected to MK test in 102 years, and decrease in magnitude ranged from 0.989 (at Shandol) to 1.610 mm per year (at Balaghat). Seasonal analysis however, showed a significant decreasing trend at two stations in monsoon season and increasing trend at eight stations in summer season. The decrease in annual precipitation was -2.59% over the entire MP in last 102 years. The most probable year of change was 1978 in annual precipitation when subjected to cumulative deviations and Pettit-Mann-Whitney tests. The period of 1901-2002 exhibited large spatial variability in precipitation trends, rather a mix pattern was detected. After the change point (1979-2002), summer precipitation showed an increase, and winter and monsoon seasons a widespread decrease. This may adversely affect the cultivation of rice crop in eastern MP. The annual precipitation slopes were seen to be negative in the study area. A significant increase at the rate of 0.62 °C/102 years was observed over MP in the annual minimum temperature, at the rate of 0.60 °C/102 years in the annual maximum temperature, and at the rate of 0.60 °C/100 years in the annual mean temperature. Seasonal analysis showed that the magnitudes of the positive trends in winter, summer, and monsoon mean temperatures were 1.20, 0.59, and 0.07 respectively. The similar pattern was observed in seasonal maximum and minimum temperatures. However, the night temperature increase was more than day temperature during the study period. The increase in temperature during reproductive, grain formation and ripening phase of crops (February to March) is understood to be detrimental to productivity of wheat and other Rabi season crops due to terminal stress. The decadal variation of annual temperature variables shows that the tempeature anomaly was positive after the 1951 to 1960. The winter season mean temperature for sowing of Rabi crops in October to November were increased which can be detrimental for seed germination. Similarly for Rabi crops in reproductive phase (March to April) may cause hastening in the maturity, reduce grain setting and grain number, and thereby reducing yield. The minimum temperature increased at a higher rate (0.42 °C) followed by the mean (0.36 C), and the maximum temperature (0.32 °C) during the More Urbanized Period (MUP) with population increase (by around 2.6 times) than the Less Urbanized Period (LUP) (1901-1951) to MtJP(1961-2001). At Basin Level The trends in hydro-meteorological variables namely maximum, minimum, mean, and dew point temperature; reference evapotranspiration; wind speed; and relative humidity are investigated for three stations (Rewa, Satria, and Allahabad) located in Tons River Basin using monthly, annual, and seasonal scales. The data of 1969-2008 are used for Satna and Allahabad, and from 1969-2003 are used for Rewa station. Data of 15 weather stations from 195 1-2004 are used to determine trends in precipitation (i.e. rainfall) data. From the data of 1979-2008, the trends in streamfiow are detected at Meja road and Satna gauging sites. The following text analyses the trends for each variable. Temperature: The analysis of annual time series revealed significant warming ranging from 1.4 to 1.7 °C per 100 years in maximum temperature. The minimum temperature showed an increase of about 1.9 to 4.0 °C per 100 years between the stations. The annual mean temperature increases in the range of 1.8 to 3.1 DC per 100 years in the study area. Significant increasing trends in maximum and mean temperatures were observed during winter and post-monsoon seasons. However, minimum and dew point temperatures showed increasing trends in all seasons. The increase in dew-point temperature varied from 0.27 °C per decade in monsoon and 1.32 DC per decade in winter at Allahabad and Rewa stations, respectively. By comparing the result over MP, it can be revealed that warming is more prominent at basin level than district level. Precipitation: Out of 15 stations in annual precipitation, significant negative trends were observed at Govindgarh (5.5 mm/year) and Meja (6.6 mm/year) stations only. The analysis indicated significant decreasing trends at 14 stations in pre-monsoon, 10 stations in winter, 2 stations in postmonsoon and only I station in monsoon season. Winter precipitation decreased by 0.33 mm/year at Maihar station, and 0.61 mm/year at Jaso station. In pre-monsoon, the decrease ranged from 0.05 mm/year at Sirmaur station, and 0.49 mm/year at Rewa station. The decrease was 4.1 mm/year at Govindgarh station in monsoon. During post-monsoon, the decrease was 0.7 mm/year and 0.3 mm/year at Birsinghpur and Jaso stations, respectively. Wind speed: Decreasing trends were observed at monthly, annual, and seasonal scales at all stations. The annual decrease varied from 0.10 to 0.23 meter per second per decade. The decrease ranged from (-) 0.06 meter per second per decade in post monsoon to (-) 0.36 meter per second per decade in monsoon season. Relative humidity: A significant increasing trend in annual relative humidity was observed at Allahabad (0.16% per decade) and Satna (0.12% per decade) stations. The pre-monsoon season showed a significant increasing trends at Allahabad (0.27% per decade) and Satna (0.2 1% per decade) stations. Only winter season shows a significant increase (0.20% per decade) at Allahabad station. Reference evapotranspiration (ETo): The decreases in annual ETo were 4% (i.e. 1.75 mm per year) and 23% (i.e. 8.987 mm per year) for Rewa and Allahabad stations, respectively. Seasonally, the decrement of winter ETo ranged between 8-9% between different stations. Pre-monsoon ETo showed decrease of 9, 25, and 29% at Rewa, Satna, and Allahabad stations, respectively. Similarly, monsoon ETo showed a decrease in the range (8, 26) % in the study area. Decrement of 6-7% was observed in post-monsoon ETo. ETo was found to be the most sensitive to maximum temperature, followed by net solar radiation, minimum temperature, vapor pressure deficit, and wind speed, respectively. Streamfiow: A significant decrease in streamfiow at Satna gauging site was found in all months, except August, September, and October. The decrease varied between -0.16 cumec/year in the month of April and 0.59 cumec/year in December. At Meja road station, January-May, November V and December showed significant increasing trends, which varied from 0.99 cumec/year and 7.6 cumec/year for May and November, respectively. Significant increasing trends at the Meja road site were observed at annual, winter, pre-monsoon, and post-monsoon scales. However, at Satna station, winter and pre-monsoon seasons showed significant decreasing trends. Increases in annual discharge were 571 cumec per decade and 29.4 cumec per decade at Meja road and Satna stations, respectively. At Meja road, the increase was 15.92, 7.33, and 23.13 mm/year during winter, premonsoon, and post-monsoon, respectively. The decreases at Satna station were -1.4921 mm/year and -0.6429 mm/year in winter and pre-monsoon seasons. EFFECT OF ENSO ON PRECIPITATION AND STREAMFLOW If To evaluate the effect of ENSO on precipitation and streamfiow, cumulative frequency distributions (CDF), correlation analysis, and composite analysis techniques were employed. The monthly precipitation series of Tons River Basin was used for 54 years (195 1-2004) and monthly streamfiow of Meja road and Satna stations for the year 1978-2008. The monthly Sea Surface Temperature (SST) anomalies ofNiño 3.4 region (5° N to 5° S and 1700 W to 1200 W) were used. Pearson correlation and Kendall's i-correlation between ENSO and precipitation indicated significant negative correlation in monsoon season for all stations. Therefore, further analysis was carried out for monsoon season precipitation only. The CDF of standardized monsoon precipitation during El Niño, La Nina, and full time series shows that the exceedance probability for long-term average monsoon precipitation is more than that of El Niño year and less than that of La Nina year. It is concluded that the zone of highest precipitation had the high exceedance probability during El Niño and La Nina years, and vice versa. Composite analysis results during El Niflo phase showed iv that, in high precipitation zone, there are equal chances of occurrence of below or near normal precipitation. However, in low precipitation zone, there are more chances of below normal precipitation. During La Nifla phase, there are almost equal chances of occurrence of above and near normal precipitation for each climatic division. A comparison of the results of data from 195 1- 1974 and 1975-2004 showed that the relationship between precipitation and ENSO (El Niño and La Nina events) has been quite weak in recent years. Similar to the above, correlations between ENSO and streamfiow indicated a significant negative correlation during pre-monsoon season, and a positive correlation in monsoon season at both Meja road and Satna stations. The CDF of standardized annual streamfiow during ENSO phases showed that the exceedance probability of La Nina year was higher than El Niño year and full time series. Similar results were obtained for monsoon season at both the stations. During El Niflo, comparison of the results from perennial (Meja road), and intermittent Satna station) monsoon streamfiow showed that the maximum chance of near normal or above normal streamfiow at perennial streamfiow and maximum chance of near normal streamfiow at intermittent streamfiow. During La Nina phase in monsoon season, there is more probability of occurrence of above normal streamfiow at perennial streamfiow and equal probability of below normal or near normal streamflow at intermittent streamfiow. SPATIO-TEMPORAL VARIABILITY AND FORECASTING OF DROUGHT The standardized precipitation index (SPI) was used to determine the drought at multiple time scales (viz., 3, 6, 12, 24, 48-month time steps) using monthly historical precipitation of 18 stations during 1951 to 2004. Wet and dry years were identified based on minimum and maximum precipitation of all stations. Spatial distributions of drought (on annual basis) for wet and dry years were explored using ArcGlS 9.3. The maximum number of droughts occurred in north side of study area (at Karchana station), and the minimum number in West side (at Baraundha station), at 3 and 6-month time scales. However, at 12, 24, and 48-month time scales, maximum and minimum number of droughts occurred in the West (at Birsinghpur station) and North (at Allahabad Chailk) side of the study area. The number of occurrence of drought events was high in case of mild drought followed by moderate, severe, and extreme droughts; and it was found to increase in recent decade. The years 1961 and 1981 were the dry and wet years, respectively, at almost all stations. In wet year, the entire study area was wet at 3, 6, 12, 24 and 48-month time scales and in dry year, only a part of the area was dry. The results indicated that droughts could be successfully forecasted with reasonable accuracy by the best identified ARIMA models. V STATISTICAL DOWNSCALING USING LEAST SQUARE SUPPORT VECTOR 4- MACHINE Least Square Support Vector Machine (LS-SVM) was used for projecting monthly precipitation, streamfiow, maximum and minimum temperatures for the period 2001-2100. To develop models about 70% of randomly selected data was used for training, and remaining 30% for testing of the models. The cross-correlation was used to select the potential predictors between predictand and predictors. Principle Component Analysis was applied to reduce the dimensionality of predictors. The LS-SVM model performance was evaluated employing four indicators, correlation coefficient, root mean square error, normalized mean square error and Nash-Sutcliffe coefficient. The required rcgularization parameter (gam and squared kernel) for training of LS-SVM was selected using possible combination of three kernel types (linear, polynomial and Radial Basis Function (RBF)) and three optimization algorithms (simplex, grid-search and line-search). The future predictands were dowuscaled using CGCM3 data of A2 emission scenario. The developed LS-SVM models were found to be efficacious to simulate future (2001-2100) monthly maximum and minimum temperature at three stations namely Allahabad, Satna and Rewa in Tons River basin. The maximum and minimum temperatures are seen to increase in future at all stations. A significant increasing trend was seen in annual maximum temperature when MK test was employed; the increase (derived from Sen Slope estimator) was 0.47, 0.05, and 0.1 °C per decade for Allahabad, Rewa, and Satna stations, respectively. The increase in minimum temperature was 0.42, 0.76, 0.68 °C per decade at Allahabad, Rewa and Satna stations, respectively. Significant increasing trends were observed in all seasons in both minimum and maximum temperatures except in I1OnSOOrI maximum temperature at Rewa and Satna stations, which show a significant decrease. The increasing trends varied between 0.09 °C per decade in pre-monsoon season at Satna station and 0.7 per decade in winter season at Allahabad station. For minimum temperature, increases in iiagnitude are 0.1 °C per decade in winter season and 1.42 °C per decade in monsoon at Allahabad and Satna stations, respectively. It can be inferred that the climate of the study area would be warmer in future and this warming is more pronounced during night than day. The i ntercompari sons of temperature at different stations show that the station at higher altitude would experience more heat than station at lower altitude. LS-SVM downscaling models were developed for projection of monsoon (June- September) and non-monsoon (October-May) seasons precipitation for Tons River basin using data of 195 1-2004. The results of calibration and validation indicated the LS-SVM model is able to downscale both wet vi and dry seasons precipitation. The results of projections showed that wet season precipitation would increase in future. However, in dry season, there would be no or little change. A significant increase was 14.36 mm/year and 0.20 mm/year during the next 100 years (2001-2100) for wet and dry precipitation, respectively. LS-SVM downscaling models for projecting monthly streamfiow for monsoon and non-monsoon seasons were developed for Meja road station using data of 1979-2008. The selected potential predictors were air temperature, geopotential height, zonal wind, meridional wind, and specific humidity at different grid points for different pressure levels; and precipitable water content, surface pressure and soil moisture content at surface. This study incorporated the meridional wind, zonal wind (which is a main mechanism to transport moisture into the region) and soil moisture (which is important for streamfiow generation), which have not been used as a probable predictor for downscaling streamflow previously in India. The results of future projections showed that during the monsoon season, average streamflow would be less than past streamflow, and in non-monsoon season, it would be more. The magnitude of decrease and increase were -6.2 cumec/year and 2.06 cumec/year during monsoon and non-monsoon seasons respectively.