Research On Regional Yield Estimation Of Winter Wheat By Remote Sensing Based On Improved CASA Model

The accurate and timely monitoring and evaluation of the regional grain crop yield is more significant for formulating import and export plans of agricultural products,regulating grain markets and adjusting the planting structure.In recent years,crop yield evaluation combined with remote sensing technology has become an important research direction.In this study,an improved Carnegie–Ames–Stanford Approach?CASA?model was coupled with time-series satellite remote sensing images to carry out the research on regional yield estimation of winter wheat.Firstly,in 2009 the entire growing season of winter wheat in the two districts of Tongzhou and Shunyi was divided into 54 stages at 5-day intervals.Net Primary Production?NPP?of winter wheat was estimated by the improved CASA model with HJ-1A/B satellite images from 39 transits.For the 15 stages without HJ-1A/B transit,MOD17A2H data products were interpolated to obtain the spatial distribution of winter wheat NPP at 5-day intervals over the entire growing season of winter wheat.Then,MODIS Normalized Difference Vegetation Index?NDVI?time series data products were adopted to estimate the growth curve of winter wheat,and this curve was subsequently utilized to judge the multiple key phenological stages represented by the flowering stage during the growth stage of winter wheat,using the ratio of the cumulative NPP in the post-flowering stage to that in the pre-flowering stage to build a new estimation model of winter wheat harvest index,and finally the remote sensing estimation of regional harvest index was realized.Then,an NPP-Yield conversion model was utilized to estimate winter wheat biomass and yield in the study area.Finally,the accuracy of the method to estimate winter wheat yield with remote sensing images was verified by comparing its results to the ground-measured yield data.The main contents and conclusions of this thesis are as follows:1.Optimizing parameters of the original CASA model.Five meteorological stations with sunshine hours around the study area were selected.After processing,the solar radiation was calculated day by day,then the total solar radiation?SOL?was accumulated at intervals of 5-day.Based on the NDVI and SR of the two vegetation indexes,₍?1?andwere calculated respectively,and the same weight was given to both to get the estimation value of the fraction of absorbed photosynthetically active radiation?FPAR?.In order to reduce the error,the maximum and minimum values of NDVI and Ratio Vegetation Index?SR?were calculated for each month in the winter wheat growing season to determine the important parameters(1₈?6??(1₈4)9))and₈?6??₈4)9))in the FPAR.Using a simpler and more effective model to estimate the water stress factor((2)in the light use efficiency.And^*was determined according to the environmental conditions of the study area.2.Using MODIS GPP?MOD17A2H?product for NPP interpolation.Based on time series HJ-1A/B satellite remote sensing images,the improved CASA model was used to estimate the NPP of winter wheat at the regional scale.For a few stages for which HJ-1A/B transit images were not available,the MOD17A2H product was used for appropriate interpolation to obtain the spatial information of NPP with 5-day intervals in the whole growth stage of winter wheat.The results showed that due to the influence of environmental factors such as solar radiation and temperature,the monthly accumulated NPP in winter wheat after planting had continuously decreased and reached its lowest value in January.After the wintering stage,the growth activity of winter wheat had accelerated,and NPP had shown an upward trend.In May,NPP was the largest,which was 128 gC m^-2.With the growth and maturity of winter wheat,the NPP accumulation in June continued to decrease until winter wheat was harvested.Overall,the distribution of NPP accumulated in the whole growth season of winter wheat in the study area was relatively average,mainly concentrated at 300-700 gC m^-2,with an average value of 464gC m^-2.3.Based on MODIS NDVI?MOD13Q1 and MYD13Q1?products,the key growth stage of winter wheat was determined.Through NDVI time series data products,the growth curve of winter wheat in the study area was established to simulate the growth characteristics of winter wheat growth season.The Savitzky-Golay?S-G?filtering method was used to smooth the NDVI time series data,so that the filtered NDVI value can better reflect the growth characteristics of winter wheat.According to the judgment method of the key phenological stage,the seedling stage,heading stage,flowering stage and harvest stage of winter wheat were determined.4.Extracting the spatial information of winter wheat harvest index.According to the determined specific time corresponding to the emergence,flowering and harvesting stages of winter wheat,a new parameterthat can effectively evaluate the winter wheat harvest index was constructed by using the ratio of the cumulative NPP in the post-flowering stage to that in the pre-flowering stage,in order to complete the study area extraction of spatial information of winter wheat harvest index.In this thesis,the average value of HI was 0.47,which was mainly concentrated in 0.35-0.55.5.According to the estimated information of NPP and HI,the NPP-Yield conversion model was used to estimate the winter wheat biomass and yield.The results showed that the biomass of winter wheat in the study area was mainly concentrated between 900-1600 g m^-2,with an average value of 1176 g m^-2.In addition,the biomass of Tongzhou District was slightly better than that of Shunyi District as a whole,which was reflected in that the average biomass of 1187g m^-2 in Tongzhou District was higher than that of 1161 g m^-2.The average winter wheat yield in the study area was 5763 kg hm^-2,and the yield was concentrated in the range of 3500-8000kg hm^-2,which exceeded 90%of the area distribution of the entire study area.Among them,the winter wheat yield in Tongzhou District was mainly concentrated in the 4000-8500 kg hm^-2,with an average yield of 5816 kg hm^-2,which was slightly higher than the overall average of the study area.The average winter wheat yield in Shunyi District was 5708 kg hm^-2,slightly lower than that than in Tongzhou District.In 2009,the difference in winter wheat growth between Tongzhou and Shunyi Districts in Beijing was relatively small,showing that the yield in Tongzhou District was slightly higher than that in Shunyi District.In general,the growth of winter wheat in the study area was ideal.6.Using measuring yield data to verify the accuracy,the determination coefficient R² is0.57,and the root mean square error RMSE is 1080 kg hm^-2.Through analysis,the average absolute error is-630 kg hm^-2 and the average relative error is-5.89%,which indicates that the estimated production model proposed in this thesis has good feasibility and application potential on a regional scale.