期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2006
卷号：XXXVI-8/W48
页码：85-90
出版社：Copernicus Publications
摘要：The present paper highlights the experience of development of satellite-based methodology, results and achievements of wheat production forecasting in Haryana state of India under the project since 1989-90. Haryana with a geographical area of 44,212 km2 is the major agricultural state of the country. At the beginning of the project the area estimates were provided at the state level and then at district-level. Subsequently, the work on production estimation was also started in the year 1990 at the district level. Single date in-season digital satellite data coinciding with peak vegetative stage of wheat crop was analyzed for wheat acreage estimation using Maximum Likelihood (MXL) supervised classification. Ground truth is collected near synchronous to satellite pass from large contiguous and homogeneous sites of wheat grown under different physiographic conditions. To save the computer timings and to complete the estimates in the large areas of the state well in time, stratified random sample segment approach has been adopted for analysis. The segment size was initially kept at 10*10 km which was reduced to 7.5*7.5 km and finally to 5*5 km. The sampling fraction was also increased from 10 to 20 percent. However, complete enumeration approach was followed subsequently, that involved analysis of the satellite data for the complete district, as the computer speed and capacity increased and the size of the districts decreased with creation of new districts in the state. Yield of wheat was estimated using multiple regression models developed using satellite based spectral vegetation indices along with meteorological and historical yield data. Zonal Spectro-trend, Trend-agromet and Spectro-trend-agromet models for wheat production forecasting were developed for different regions of the state of which the last one gave the best results. The performance of wheat acreage and yield estimates was evaluated by computing Relative Deviations (RD%) with Department of Agriculture (DOA) estimates. It was observed that the area and production estimates improved with decrease in the segment size and increase in the sample segment as the size of various district decreased gradually. During the last 15 years of estimation the spatial resolution of the satellite data has gradually improved from 80 meter of MSS to IRS-IA/ IB/ lC/ ID (LISS-I/ II) and finally 23.5m LISS-III of Resourcesat which is being currently used for this purpose. Besides the spatial resolution, the spectral and temporal resolution of the satellite data has also improved which drastically improved the crop discrimination. Both accuracy as well as precision of the estimates improved over the years, as reflected by the relative deviation and CV values respectively, for various districts. However, the northern districts having large contiguous areas of the crop showed better accuracy and precision as compared to southern districts having a mix of various crops. Presently remote sensing technology is able to provide district level acreage and production estimates with 95 % accuracy
