The Potato Yield Estimation And It’s Accuracy Assessment Using Multi-source Remote Sensing Data

Potato is a kind of economic crops with high grain yield and high addedvalue. The potato planting area and output of Inner Mongolia AutonomousRegion is rank former in each province and municipality; it located in the northand suitable for planting one year season of spring potato. The Potato industryis one of the six pillar industries of the Inner Mongolia autonomous region, itdevelops toward industrialize, the planting region is widely distributed and theplanting area just after corn. Based on the remote sensing technology can bereal-time, accurate, efficient and comprehensive grasp in all aspects of potatoplanting production and yield information.The Midwest of Inner Mongolia potato planting area as topical region,Qahar Right Wing Front Banner and Jining city as study region are studied.Remote sensing image are used MODIS TM5(HJ1B replacement) of potatogrowth period from May to September. The2010mid July is potato’s growingseason, to investigate the spatial distribution of potato planting, establish ofinterpretation signs and obtain the spatial distribution map. In2010mid-lateSeptember is potato’s harvest period, take potato sample and yield survey. Thecollected effective sample are69, The collection is separate into household andthe company’s mode of operation, upland and irrigated land types, sample landdistance from road one kilometer, distribution in each township. Establishregression model of59measured sample’s yield data with different growthperiod of vegetation index, back substitution test the10experimental samples.Variables of estimate model are used9(5kinds of single growth periodand different accumulation growth period) different periods NDVI and EVIfrom MODIS and TM5, establish linear, logarithmic, two items, power, S typeand index six kinds of216Single-element regression model to comparativeanalysis. Result shows that, MODIS of from June to September, tuber formation period, tuber expansion period, the starch accumulation stage, themature period cumulative NDVI as variables, R2as0.671linear regressionmodel is better than other types of regression model. In the correlation F testand T test, were both reach significant level. The MODIS images of monthlyaverage value cumulated NDVI value from May to September into the linearregression model, result shows that study region potato planting area is209063acres; the total yield is413062.65kg, yield1976kg per acre. The standard errorof back substitution test is140, the mean absolute error of sample point withregression prediction reached-86kg/acre, the average value of relative errorreached-4.51%. The error value is in controllable range, means that it isfeasible to based on remote sensing technique to yield estimation, and thetomato total output and yield of study region potato is statistical significanceand reliable.