Forecasting Crop Production
under Abnormal Weather Conditions: A Case Study for Groundnut Crop in
Gujarat
S. S. Pokharna and S. S. Ray ARD/RSAG/RSA, Space Applications Center, Ahemdabad 380053, India A. M. Shekh, H. Venkatesh, K. I. Patel and R. B. Gajjar Department of Agrometeorology, Gujarat Agricultural University, Anand 38001, India S. P. Nanawati Gujarat State Oilseeds Growers' Federation Ltd. Ahmedabad 380009, India Abstract This study shows the use of remote sensing to find the change in groundnut crop inventory in Gujarat for two consecutive years(1992-93 and 1993-94). The 1993-94 being a rainfall deficient year, the groundnut production had drastically reduced from the production had drastically reduced from the production in the nrmal year 1992-93. Crop acreage was estimated using satellite based remote sensing data. A cumulative soil moisture balance model, which computer water requirement satisfaction Index (WRSI) was used for yield estimation. The district - level acreage, yield and production estimates were given for five district of Saurashtra region. The groundnut production of the group of five districts decreased fro 1497 thousands tones in 1992-93 to 547 thousand tones in 1993-94. These estimates were in fairly good agreement with the estimates made by the deptt. Of agriculture. 1. Introduction Remote sensing (RS) technology has been extensively used for forecasting crop production for major crops under normal weather conditions in India (Anon. 1990 and Navalgund et al. 1991). However, large areas in the country are rainfed and hence crop production varies widely from year to year depending on weather conditions. RS technology coupled with metereological models can prove very useful for crop inventory study in the abnormal weather conditions. India strands first both in acreage and production of groundnuts (Arachis hypogaea L.) in the world. Among the states in India, Gujarat occupies first position in groundnut acreage and production. Hence, it is essential to estimates groundnut production for Gurajat in both normal and abnormal weather situations because they play and important role in determining the oil economy in the country. Pokharna et al 91991) and Medhavy et. Al. (1989) have studied the possibility of oilseeds acreage estimation at district level using a single date RS data. However, for yield estimation in rainfed situation, agrometereological models prove to be very useful (Ray et. Al 1994). Frere and Popov (1979) computed a Water requirement satisfaction in Index (WRSI) from their simplified cumulative water balance model for which the major input is rainfall. This model has an advantage over the RS based yield model as groundnut is a Kharif crop and it is difficult to get cloud free RS data during this season (Ray et.al 1993). Also since this model is strongly rainfall dependent, it could be very useful under abnormal weather conditions. In this context the present study deals with i) district level acreage estimation using RS technique for groundnut and ii) district level yield estimation using WRSI based models under both normal as well as abnormal weather conditions. 2. Materials and Methods 2.1 Study Area Groundnut in Gurajat is mainly grown in five districts of Saurashtra lying between Gulf of Kachch and Gulf of Khambat. These district are Amreli, Bhavnagar, Jamnagar, Junagadh and Rajkot. These five districts together, covering a geographical area of 53,850 sq km, contribute around ninety per cent to the groundnut production of the state. Groundnut is grown as rainfed Kharif crop in this area. Together crops grown in Saurashtra region during Kahrif season are seasamum, pearl millet, sorghum and cotton etc. 2.2 Data Used 2.2.1 Satellite Date : The five districts of groundnut are covered by five IRS LISS I scenes. The path and row of each scene is given in Table 1. The dates of acquisition were so chosen that these corresponded to the maximum vegetative stage of the concerned crop. In 1993-94 cloud free Rs data were not available during the peak vegetative stage i.e., second week of September. Hence, initially NDAA AVHRR data was used to give combined acreage estimates for the group of five districts. Later on, Landsat TM as well as IRS-IB data of October first week (Table 1) were used to give district level estimates.
2.2.2 Agrometeroloqical Data : The district-wise historical yield data for groundnut for 32 years i.e. 1961-1991 were collected from DES (Dept. of Economics and Statistics) publications: District-wise fortnightly rainfall and number of rain days data (needed for yield forecast for the same period were collected from Dept. of Agriculture, Gujarat state (DDA). The soil physical data were collected from the soil science department of the Gujarat Agricultural University, Junagadh. Crop coefficients for were taken from Dorrendbos and Kassam (19979). 2.3 Acreage Estimation For analysis of digital data, a stratified random sampling approach was and a sampling fraction of ten per cent. For stratification, historical taluka-wise acreage estimates as well as vegetation density found in he FCC prints of satellite data of previous year was used The ground truth information about different land cover classes was collected in fourth week of September and the first week of October 1992. for the year 1993-94, groundtruth information was collected in fourth week of September and also during November 9-8 for yield information. Digital analysis of satellite data for the year 1992-93 was carried out on VIPS-32 image processing system (VAX 11/780) at RRSSC, Jodhpur using supervised classification method with a maximum likelihood classifier. Aggregatin was done on both Dipix image processing system at SAC, Ahmedabad and VIPS-32 system at Jodhpur. 2.4 Yield Estimation A cumulative water balance model developed by Frere and Popov (1979) based on periodic values of precipitation and PET was used. The main relationship is given by: Where S = net water added to soil in ith period Pi = precipitation in ith Period WRi = water requirement ith Period WR = KC * PET and KC = crop coefficient The WRSI indicates (in percent) the extent to which the water requirements cumulatively, duing the growth cycle. It is taken as equal to 100 at the time of sowing and in case of a deficit the WRSI is reduced by the percentage ratio of the water deficit and the total crop water requirements over the whole crop season. If the water surplus is greater than 100 mm in a fortnight and the number of rainy days is less than 6, WRSI is decreased by 2.1 units for each 100 mm of excess water. The final value of WRSI at the end of growing season is related exponentially to the yield. The model was run on daily basis starting from June 1 of every year. Normally onset of the monsoon in the study area takes place after June 10. The value of KC for bare soil evaporation, was taken to be 0.2. For three major stages of the crop viz. vegetative, flowing and maturity, the values of crop coefficient chosen were 0.6, 1.1 and 0.8 respectively. The field capacity, permanent wilting point and solid depth were 150mm, 25,, and 60 cm, respectively. It was assumed that the sowing started after the cumulative rainfall exceeded 30 mm. District level production estimation was made using the acreage values from RS method and yield values estimated by yield -WRSI models. 3. Results and Discussion: 3.1 Normal year (1992-93) 3.1.1 Acreage Estimates : The acreage estimated by Rs method (Table 2), when compared with DDA gave the relative deviation n(RD) wich varied from - 13.0 to 19.0 per cent. In Jamnagar as the crop condition in some talukas (Okha, Jodiya) was very poor they could not be classified under groundnut, which resulted in underestimation. In Amreli other field crops had been misclassified under groundnut. So there is a need for representative ground truth site in case of every type of crop cover.
3.1.2 Yield Estimates : The WRSI, estimated by the cumulative water baltially relatied with the yield with R2 varying from 0.46 to 0.67 (Table 3). Groundnut yield forecast for the current year were made using the yield-WRSI and Y represents yield. The relative deviations from Dept. of Agriculture's estimates were ranging from 43.7 to 9.2 per cent (Table 4).
3.1.3 Production Estimates : The district level estimates of production for groundnut were 401.7, 244.8 187.8, 365.8, and 297.2 thousand tones, for Amreli, Bhavnagar, Jamnagar, Junagadh and Rajkot districts, respectively (Table 5). The agreement for the all the districts except Jamnagar are fairly good for this preliminary study. For Jamnagar, both acreage and yield are not matching. This problem is under further investigation.
3.2 Abnormal Year (1993-94) 3.2.1 Acreage Estimates : In the NOAA AVHRR image a small part was covered by cloud and shadow. As the land cover in this is mostly forest, it did not have much effect on the result. The acreage for the five districts was estimated to be 675.3 thousand hectares in comparison to the acreage of 1532.0 thousand ha in 1992-93. This estimate, which was available by 11th October, 1993, was corroborated by the state government frecasts (Ppublished in the national news papers) which reported a decine of 10 lakh ha. In groundnut area compared to the previous year. The district level groundnut area was estimated using Landsat TM data for Bhavnagar, Janmagar and Junagadh, IRS-IB data for Rajkot and NOAA data for Amreli. There were large differences between Rs and DOA estimates (Table 2) This is because DOA figures wre based on estimates made at the time of sowing. Hence they did not include the portion of the crop which ws uprooted due to very low rainfall in the later season. Hence RS estimates give the area of the crop remaining in the field after September. Because of this discrepancy, relative deviations were not computed. 3.2.2 Yield Estimates : The district level yield estimated form the Yield-WRSI models showed a marked decline from previous year's yield estimates (table 4). The decline was low in Amreli and Bhavnagar and was sharp in Jamnagar and Rajkot. Again as the yield estimates of DDA was derived from production and acreage (estimated at the time of sowing) WRSI based yield estimates were different than DOA estimates but were obviously more realistic. 3.2.3 Production Estimates : The district level estimates of production for groundnut were 164.3, 143.2, 19.4 133.7 and 86.6 thousand tones for Amreli, Bhavnagar, Jamnagar, Junagadh and Rajkot, respectively (Table 5). The total groundnut production for the group of these five districts was estimated to be 547.21 thousand tones of the normal monsoon year. The corresponding DOA estimates were 386.82 thousand tones. District-wise estimates for Amreli, Bhavnagar and Junagadh seems to be fairly reasonable looking at the complexity of the problem whereas for Jamnagar and Rajkot, they are widely different. This problem is under invest discrepancy is that a significant fraction of uprooted groundnut plants wre not removed form the field during the time of satellite pass, which will increase the apparent area. Acknowledgement The authors are grateful to Dr. Baldev Sahai, Mission Director, RSAM, Dr. R.R. Navalgund, Group Director, RSAG, Shri J.S. Parihar, Head, ARD and Dr. Ajai, Head, EISD for their keen interest and encouragement. Thanks are due to the Staff of RRSSC, Jodhpur for their help during digital analysis. This project was funded by the National Dairly Development Board (NDDB), Anand. We are thankful to the staff members of Oil Prcuring Group of NDDB for discussion and their keen interest in this work. Authors are also thankful to Shri Sandip Oza, S.A. Sharma and Dr. M. Oza for allowing us to use their softwares. References
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||