The Study On Cotton Yield Estimate By Remote Sensing In Xinjiang

A cotton yield estimation operational system using remote sensing is essential for cotton production and trade management to collect objective, economic, efficient and timely information of dynamic monitoring cotton growth, growing area and yearly yield fluctuation, so as to provide timely and precisely significant references for knowing well situation of the same year cotton production, confirming reasonably reserve scale, publishing procurement prices, guiding planting plan, and policy execution evaluation. The objective of this research is to establish a remote sensing model for cotton yield estimation in Xinjiang using RS, GIS, GPS and combining canopy spectrum and main agriculture parameter. The main works and achievements are as follows:1. Owing to the difference of the vegetation indexes between the field measurement and distilled from remote sensing images directly, in the cotton growing periods, the correlation analysis of the vegetation indexes which distilled from remote sensing images directly and the cotton yield, the relativity is poor. In this research, calibrated the normalized difference vegetation index with linearity model, and established remote sensing model for cotton yield estimation.2. The correlation analysis of the vegetation indexes and the leaf area index(LAI) in cotton all growing period of Nanjiang and Beijiang, which indicated the relativity is very well, and reached to significance level in full bloom and bulling stage, and the two stages are the optimum temporal for estimating LAI. The regression analysis of LAI and vegetation indexes, attained the best estimate model of LAI, and SPSS software used to fit and test the models, and the fitting results are satisfied.3. With the regressive analysis of the vegetation index NDVI in every growing period and cotton yield, obtained remote sensing model for cotton yield estimation. The optimum temporal for estimating yield of Nanjiang long-fiber cotton Xinhai-14 is full bloom stage, and the exponential function equation fit well. The best model for estimating yield of hybrid cotton is linear equation full bloom stage, and then is the cubic multinomial equation of full bloom stage and boll opening stage. The best model of"China cotton 35"is cubic multinomial equation of full bloom stage.134-1 and 18-3 cotton's the best model is the exponential function equation at the full bulling stage.297-5 cotton's the best model is the exponential function equation in bolls setting stage.4. The vegetation index and yield in optimum temporal yield estimation of test regions'breed in 2006 is used as a testing sample, to check the model and precision analyze. In the optimum temporal ,the models precision reach to 98.46% except long-fiber cotton Xinhai-14 ,its precision is 84.6%, especially 297-5 cotton ,which is new demanded for Shihezi, attaches to 98.46%.5. For only using remote sensing factor to estimate cotton yield ,the precision stability is poor. With LAI and NDVI and yield of every growing stages to regressive analyze, the precisions are over 90%, 297-5 cotton's yield estimation compound model is the highest, attached to 98.46%.All compound models precisions are higher than single factorial estimation model. It indicated that modeling multiple complex with LAI and NDVI is an available approach to increase the precision of cotton yield estimation .