Research On Real-time Corn Phenology Estimation And Yield Forecast Based On Multi-source Data

Corn is one of the three most important food crops in the world,and China is the second largest corn-growing country after the United States.Timely and accurate grasp of corn planting and yield information is of great significance to strengthening the effective allocation of agricultural resources,crop production management,grain market regulation and even ensuring my China’s food security.The formation and accumulation of crop yield is a complex process,involving multiple physiological and biochemical processes,and is affected by many factors(such as meteorological conditions,soil conditions,field management,etc.).The key to understanding the process of yield accumulation is to clarify the growth stage of crops,the relationship between crop parameters and yield at different growth stages,and the response relationship between crop growth,yield accumulation and environmental factors.These all depend on the accurate grasp of the physiology-related phenological of the specific crop.The physiology-related phenological period of crops is divided according to the physiological process characteristics of specific crops,that is,the time when crops reach a certain growth stage,the growth process of different crops is different,and the phenological period is also different.Crops have relatively stable physiological process characteristics(such as photosynthesis,dry matter distribution)in different phenological stages,so crop physiology-related phenological stages have clear physiological significance,which is conducive to understanding the process of crop growth and yield accumulation at a finer scale.Remote sensing technology has the characteristics of large-area detection and short period,and is a key tool for crop regional-scale agricultural management.At present,there have been many useful explorations on the after-season non-real-time estimation of crop physiology-related phenology;however,the real-time estimation of phenological period can only rely on the incomplete time series vegetation index curve,and the uncertainty of its change trend has great influence on crop physiology-related phenology within growing season.The real-time estimation method of remote sensing in the future presents new challenges and needs to be further studied.Therefore,combined with the phenological survey data of satellite remote sensing and ground statistics,this paper proposes a real-time estimation method of corn physiology-related phenology within growing season,and based on the real-time phenological estimation results and the deep learning method,a near-realtime yield prediction model within growing season is proposed.The main research contents include:(1)A real-time estimation method of physiology-related phenological period within growing season is proposed,which combines satellite remote sensing and ground statistical survey data,which can estimate the key phenological period of crops in near real-time with high accuracy.In this paper,the U.S.Corn Belt from 2001 to 2019 was used as the research area to analyze and verify the accuracy of the within-season estimation method.The results show that high real-time estimation accuracy of corn physiology-related phenology period(RMSE from 0.20 days to 2.64 days)can be obtained by combining real-time statistical report data;Significantly higher than the accuracy of existing after-season phenology products.(2)Near real time yield prediction within growing season is one of the hot spots and difficulties in remote sensing of agriculture applications and researches.In this paper,a near-real-time corn yield prediction model based on phenological process and crop mechanism is proposed to achieve accurate estimation of corn yield in key phenological periods within growing season.In this paper,the main corn-producing areas in the Yangtze River Basin of China and the US Corn Belt are used as the research areas to analyze and verify the accuracy of the yield estimation and prediction models.Based on the crop phenology and physiological process information from 2007 to 2018 in the research area of the Yangtze River Basin in China,the accuracy of the corn yield estimation model steadily improved after the silking period.In the US study area,the accuracy of the yield prediction model was verified using the county-level yield statistics of the US Corn Belt from 2001 to 2019.The model performed well in the early reproductive stage(prediction accuracy RMSE was 0.93Mg/ha),and the accuracy of corn yield prediction improved significantly from silking stage to dough stage.The paper also discusses the uncertainty introduced by the real-time phenology estimation and the real-time acreage extraction results to the yield prediction model.In general,the crop physiology-related phenology estimation method and yield prediction model proposed in this paper can achieve near-real-time within-season phenology estimation and yield prediction of maize at the regional scale with high accuracy.The model construction framework proposed in this paper also provides theoretical basis and ideas for real-time estimation of phenological stages and yield prediction of other crop types(such as wheat and rice).