Regional Productivity Forecast And Panicle Nitrogen Fertilization Based On Integration Of Remote Sensing And Rice Model

Rice is one of the most important food crops in China.Accurate prediction of rice regional productivity and precise nitrogen management can provide important support for national food security.The coupling technology of crop modeling and remote sensing combines the advantages of the mechanism and continuity of crop model and the real-time and regional characteristics of remote sensing,and becomes an important tool for crop regional productivity prediction and production management decision-making.However,the research on the regional scale has problems such as large amount of calculation,low calculation efficiency,and difficulty in actual production and application.Taking rice as the research object and using the initialization/parameterization method,this study compared the calculation efficiency of different artificial intelligence optimization algorithms,and thentook actual parcel of rice fields as the coupling unit to develop a rice productivity prediction technology based on the coupling of multi-source remote sensing and crop growth model Rice Grow.Then,based on historical meteorological data,short-term climate predictions were made.Based on the management zoning idea,the soil nutrient and vegetation index were clustered and partitioned,and the overlay analysis technology was used to obtain field information in the early stage of the forecast.The study area was divided into multiple sub-areas.According to the production forecast of each sub-region,the suitable nitrogen application plan for panicle sprout was proposed.The multi-source satellite remote sensing data(Sentinel-1,Sentinel-2,PLANET)obtained during the key growth period of rice were inverted,and the inversion values of the leaf area index and the above-ground biomass of rice growth parameters were obtained.The initialization/parameterization method was chosen as the coupling method of remote sensing and crop growth model.Fuzzy particle swarm algorithm combining the advantages of particle swarm algorithm and simulated annealing algorithm was used as an assimilation algorithm.The remote sensing inversion result was used as the coupling parameter.The initial parameter value of the crop simulation model Rice Grow is optimized through the assimilation algorithm.Taking the parcel of rice field extrated by UAV image data as the coupling unit,the prediction of rice productivity at the regional scale was realized.The NRMSE of the simulated and measured output values in 2019 and 2020 are 13.4%and16.9%,respectively,indicating that the coupled system based on the fuzzy particle swarm algorithm can realize the regional-scale rice productivity prediction.Combined with the method of simulated zoning,the fuzzy C-means clustering partitioning method based on attribute weights was selected to analyze the soil nutrient data(total nitrogen,organic matter,nitrate nitrogen,ammonium nitrogen and p H value)and the vegetation index EVI of the study area obtained before panicle nitrogen application.Then overlay analysis on the field information obtained by the UAV image dataand the clustering division results were performed,and the research area was then divided into multiple sub-areas with uniform crop growth status and growth environment.The results showed that the study area wasdivided into 41 sub-areas in 2019 and 10 sub-areas in 2020.The variability of each index of each sub-area is less than or equal to that of the whole study area,indicating that the zoning has a good effect.Comparing the methods of weather data generation for short-term weather forecasting,the simulation results based on the weather data generated by the average value and variance of historical weather data for two years were the best.The error of the output simulation results in 2019 and 2020 is 1.6%and 12.2%for NRMSE,and R~2 is 0.99 and 0.96,indicating that the method of generating meteorological data based on the average and variance of historical meteorological data could be used for short-term meteorological data prediction.Based on the results of management divisions and short-term weather forecasts,the fuzzy particle swarm algorithm was used to optimize the panicle nitrogen application amount and time by the target yield and the target nitrogen partial productivity and the nitrogen management recommendations for each sub-region were obtained.The recommended panicle nitrogen application rate can be reduced by 33.02%and 45.53%in 2019 and 2020,respectively.