The Study On Rice Planting Area Extraction And Yield Estimation Model In Liaoning Province Based On MODIS Image

In recent years, remote sensing technology is increasingly applied in the agricultural production, it saves time and cost, so the research of extraction and yield estimation on the rice using satellite images has become widely. Using the MODIS data in Liaoning province from 2006 to 2013, this paper study the extraction of aera and yield estimation of the rice, accuracy is reliable and stable, which provide reference and guidance for promoting the development of the Northeast rice yield estimation with satellite.In order to set up the optimal fitting model, this paper extracts the rice planting area and estimates the rice yield using the moderate resolution imaging spectroradiometer (MODIS) MOD09 and MOD 13 data in Liaoning province. Extract the rice planting area in Liaoning province using threshold method and comparison method, which inversing the normalized difference vegetation index (NDVI), enhanced vegetation index (EVI) and land surface water index (LSWI) of Liaoning province, combined with the growth period of rice in the main stage, such as transplanting stage, heading stage, mature stage of spectral characteristics, comprehensive considering suitable slope for rice planting, the result is comparison method has a higher precision 90%.Liaoning province is divided in three dimensions as district> region and provincial level, establish rice yield estimation model using NDVI and EVI and LSWI in transplanting stage, heading stage and mature stage.(1) Combing a single phenological phase and multiple phenological phases respectively with a single vegetation index and multiple vegetation indexes, make correlation analysis with the statistic yield, the correlation of multiple vegetation indexes and a single phenological phase is higher than a single vegetation index and a single phenological phase, the correlation of a single vegetation index and multiple phenological phases is higher than a single phenological phase, overall performance is district level> region level> province level.(2) Select the combinations of higher correlation and set up fitting models in SPASS,choose the highest as optimal fitting models of yield estimate. Finally calculate rice yields with the vegetation indexes in phenological phases, make residual analysis with statistics, the accuracy is above 80%. The model residual of multiple vegetation indexes combining with a single phenological phase is smaller than a single vegetation index combining with a single phenological phase, The model residual of a single vegetation index combining multiple phenological phases is smaller than a single phenological phase, the overall accuracy is district level> region level> province level.