WHEAT PRODUCTION FORECASTING FOR A PREDOMINANTLY UNIRRIGATED REGION IN MADHYA-PRADESH (INDIA)

The area under wheat was estimated and a forecast of production made i n a predominantly un-irrigated region (36 per cent irrigated wheat cro p, geographical area 5.61 Mha) of Madhya Pradesh (India) using digital data from LISS-I (Linear Imaging Self Scanner) onboard Indian Remote Sensing Satellite (IRS-1B), for the crop season 1991-92. A stratified sampling approach based on 5 km by 5 km sample segments, 10 per cent s ampling fraction in conjunction with supervised maximum likelihood (MX L) classification was used for wheat acreage estimation. Yield forecas ts were based on an optimal combination of forecasts from two differen t methodologies, viz., wheat yield-spectral relationship and time seri es analysis using ARIMA (Autoregressive Integrated Moving Average) app roach. In the former, a two-year (1989-90, 1990-91) pooled regression relating LISS-I derived Near Infrared/Red (NIR/R) radiance ratio to di strict wheat yields was developed and used to forecast wheat yields fo r the year 1991-92 based on classified wheat pixels. In the latter cas e, historical district-wise wheat yield data of 35 years was used to d evelop appropriate ARIMA models and used to forecast 1991-92 yields. T he relative deviation of remotely-sensed-based forecasted production, acreage and yield from the post-harvest estimates released later by th e State Department of Agriculture were -15.8, -10.02 and -6.01 per cen t, respectively. The acreage and yield meet the accuracy of 85 per cen t at 90 and 95 per cent confidence levels, respectively.