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

Using remote sensing technology to estimate yield has become a mainstream mean in the field of crop yield estimation. Remote sensing with real-time, dynamic, comprehensive features, it can monitor crop dynamically, in real-time, all-weather, also simulate process of crop growth by extracting remote sensing parameters. NPP?Net Primary Productivity? is the first step for yield estimation, the article selects the CASA?Carnegie-Ames-Stanford Approach? model, one of the most popular yield estimation model, for NPP estimation of winter wheat, in order to realize the winter wheat yield estimation. This paper from four aspects to analyze the shortage of the original CASA model, including regional scale and parameter calculating method, maximum light energy utilization value and calculation model of the actual light energy utilization, and then improved the calculating method of some parameters in the original CASA model to get the new estimation model. First of all, using HJ-1A/B satellite products which spatial resolution is 30 m to extract NDVI?Normalized Difference Vegetation Index?; Next, improved the calculation method of NDVImax, NDVImin and FPAR, to calculate the winter wheat NPP; And then, we selects accumulation NPP from March to May to estimate biomass; In the end, by combining with the advantages of MODIS NDVI and HJ-1A/B, established a winter wheat harvest index regression model to obtain HI, thus realized the yield estimation. The main research contents and results are as follows:1.Amendment original model parameters to achieve the estimation of study area's winter wheat NPP. Through the extraction of quantile fractile with winter wheat NDVI maximum probability distribution to determine NDVImax and NDVImin, and combining previous algorithm of improved FPAR with a correction factor to determine the FPAR algorithm of this paper; Utilization of solar radiation?SOL? around the area of the site data for the interpolation by natural neighbor spatial interpolation method; According to the temperature, precipitation and other meteorological datas in the study area to calculate the real light utilization energy???; At the last,we take the above parameters into NPP estimation model for winter wheat NPP of study area. Results show that the average NPP in March, April, May were 78gC·m^-2, 297gC·m^-2 and 320gC·m^-2 respectively, This difference is caused by growth characteristics of winter wheat in different periods. In March, the wheat in the green up period, wheat's leaf area increased gradually. Into April, winter wheat into exuberant growth period, leaf area continue to increase, the NPP also will increase. In May, the wheat gradually into flowering, grain filling, and milk stage etc, during the time most parts of NPP more than 250gC·m^-2, Which are consistent with wheat physiological characteristics in different periods, wheat grows well. In addition, this paper selects yield formation of the critical period's NPP cumulative value to estimate yield of winter wheat in the study area. The proportions of winter wheat NPP in cumulative NPP are 11.23%, 42.71% and 46.06% respectively in March, April and May, which shows that winter wheat in good condition during the yield formation process.2. Using MODIS NDVI time series to establish winter wheat growth curve, and then extracts the key phenological period of winter wheat. In order to make the curve more real reflect wheat phenological period, through the Savitzky-Golay?S-G? filter method to smooth the original curve. By piecewise fitting wheat growth curve, determine the piecewise function. Based on key phenological period of wheat growth characteristics, calculate the curve inflection point, extreme value point, such as feature points, as a key phenological period. The results show that the key phenological period of winter wheat in the study area?green stage, heading stage, anthesis and ratooning buds? were February 28 th, April 15, April 28, May 23 respectively, which lay the foundation for the establishment of winter wheat HI model.3.Using the measured data and the ten-day HJ-1A/B NDVI to establish winter wheat HI model for the wheat HI space distribution of study area. This paper references Jianqiang Ren's method to calculate the HI, by extracting parameter HINDVI_SUM based on the data of GuanTao winter wheat time-series NDVI in 2014, and HINDVI_SUM is equal to the crop's reproductive growth key stages?flowering period to ratooning buds? and vegetative growth key stage?green period to flowering period? of the accumulated NDVI ratio. Then established the relationship between Guantao's actual measurement HI and Guantao HINDVI_SUM, as the harvest index model, through calculating the study area HINDVI_SUM, and then plug in the regression equation to get the spatial distribution of winter wheat harvest index of the whole study area. The result shows that the winter wheat HI of the whole study area on the spatial difference is not obvious, the study area HI predicted is between 0.3 and 0.56, the average is 0.44.4.Based on formation mechanism of winter wheat yield, firstly, transformed winter wheat NPP into winter wheat dry weight. Secondly, introduced winter wheat HI. Finally, we will get the spatial distribution of winter wheat yield of our whole study area. The result shows that the average yield of the whole study area is 7127.4kg·hm^-2, Quzhou, Qiuxian, and Guantao average yield are 6886.65kg·hm^-2, 7094.83kg·hm^-2, 7519.92kg·hm^-2, respectively. Predicted average yield of Guantao significantly is higher than the others, this may be associated with the standardization of the farming system of Guantao, in recent years, and the fine varieties of plant, especially the black wheat planting increasingly have a certain relationship. Consider from space, the north of Guantao, Qiuxian northeast, south of quzhou, yield mainly between 8000-9000kg·hm^-2, superior to other fields in the study area.5.By using two kinds of authentication method verifies the accuracy of the results. One is measured data, another one is statistical yearbook data. The measured data shows that estimated yield of Guantao winter wheat in 2014 slightly is higher than the actual measurements', the average relative error is 1.54%,the absolute error is 105.18kg·hm^-2. Statistical yearbook data validation result shows that the average forecasting yield of the whole study area in 2014 is a bit below than statistical values', the relative error of 1.03%, the absolute error is 74.07kg·hm^-2, but predictive value of Guantao is higher than statistical yearbooks', the relative error is 3.73%. The validation result is basically identical with the former.