Remote Sensing Estimation Of Winter Wheat Yield In North China Plain Based On Phenological Adjustment

Timely and accurate estimation of grain yield is of great significance to ensure food security and real-time food policy in China.With the development of remote sensing technology,the timeliness of crop yield estimation is improved,and the yield can be estimated in a large range of crops.Due to the influence of heat,geographical location and other factors,Crop Phenology has certain differences in a large range of grain production estimation.Therefore,in this paper,phenology differences are considered in the estimation of grain production.Using MODIS evi time series data,firstly,the winter wheat planting area is extracted according to the characteristics of the winter wheat curve,then the winter wheat rejuvenation period is identified,finally,the phenology of the time series data is adjusted according to the winter wheat rejuvenation period,and the yield is estimated according to the adjusted data.The main research contents and achievements include:(1)Extraction of Winter Wheat in North China Plain.Because MODIS evi data is easy to be affected by cloud,sensor and other factors,the abnormal values appear.In order to reflect the more real growth curve of winter wheat,the original data is filtered by savitzky Golay to remove the abnormal values.After reconstruction,the evi time series curve can keep the trend characteristics of two peaks and one valley of the original curve of winter wheat,which lays a foundation for the follow-up study.Then,according to the reconstructed evi time series data,the second difference calculation is carried out to determine the position of wave crest and wave trough,and the decision rules are made to extract the winter wheat planting area.The results show that the spatial distribution of Winter Wheat in the North China Plain is relatively concentrated,among which the winter wheat in Hebei Province is mainly distributed in the south central part of Hebei Plain,Shandong Province is mainly distributed in the northwest and southwest plain of Shandong Province,Henan Province is mainly distributed in the Middle East Plain and the southwest Nanyang basin.In the accuracy verification,compared with the results of statistical yearbook,the extraction accuracy of Henan Province is more than 80%,while that of Beijing,Tianjin,Hebei and Shandong is less than 80% in only one year,2009 and 2012 respectively,and that of other years is more than 80%.In 2013,the area of Beijing,Tianjin and Hebei was compared with that of the statistical yearbook,and the correlation coefficient was(R~2)0.87,and the root mean square error was 85.1585 × 103hm2,which had a good correlation.The extraction of winter wheat planting area lays a foundation for phenological monitoring and yield estimation.(2)Monitoring of winter wheat returning to green.At first,the 8-day resolution data is interpolated into 1 day by cubic spline interpolation.Then,based on the characteristics of curvature change rate,the return period distribution map of Winter Wheat in North China Plain is obtained by using logistic model.It is found that the beginning time of rejuvenated period in the south of North China Plain is earlier than that in the north,and it is gradually delayed from the south to the north.The overall distribution characteristics are consistent with the geographical location and heat change law.In the analysis of extraction precision,the maximum error between the extracted value and the observed value is 20 days,the minimum error is 0 days,the average error is 11.91 days,and the root mean square error is 12.05 days.The extraction of the beginning time of rejuvenated period is the preparation for the phenology adjustment in the following paper.(3)Estimation of winter wheat yield in North China Plain.In the large-scale planting area of winter wheat,the time series curve is different in the same time period when the phenology is different.In this paper,the rejuvenated period of winter wheat is regarded as the beginning time of phenology after the winter period,and the time series data are adjusted pixel by pixel.Taking the rejuvenated period as the starting time,the time series data of winter wheat after the phenology adjustment are obtained by taking one value every five days.According to the accumulated value of evi in single time phase and four time periods,i.e.green returning period,green jointing period,jointing heading period,and heading milk maturity period,a linear regression was established between the accumulated value of evi and the yield of winter wheat,and the best yield estimation model was selected respectively.In the single time phase yield estimation model,55 days,65 days and 45 days data after the green returning period were selected for yield estimation in Beijing,Tianjin,Hebei,Henan and Shandong respectively.The final yield estimation results and statistical yearbook data of the three regions had first errors of-2.60,-0.80,and-8.80,respectively.The relative errors of all regions were-4.80,and the determination coefficients were all above 0.91.The cumulative values and yield lines in phenological period were above 0.91 In the sex regression model,the best model periods are jointing heading stage,heading milk maturity stage and jointing heading stage respectively.The final relative errors of yield estimation in Beijing,Tianjin,Hebei,Shandong and Henan are-2.52,-3.64,-9.60.The relative errors of the whole region are-6.40,and the determination coefficient(R~2)is above 0.91.