Integration of a Crop Growth Model and Deep Learning Methods to Improve Satellite-Based Yield Estimation of Winter Wheat in Henan Province, China

Timely and accurate regional crop-yield estimates are crucial for guiding agronomic practices and policies to improve food security. In this study, a crop-growth model was integrated with time series of remotely sensed data through deep learning (DL) methods to improve the accuracy of regional wheat-yield estimations in Henan Province, China. Firstly, the time series of moderate-resolution imaging spectroradiometer (MODIS) normalized difference vegetation index (NDVI) were input into the long short-term memory network (LSTM) model to identify the wheat-growing region, which was further used to estimate wheat areas at the municipal and county levels. Then, the leaf area index (LAI) and grain-yield time series simulated by the Crop Environment REsource Synthesis for Wheat (CERES-Wheat) model were used to train and evaluate the LSTM, one-dimensional convolutional neural network (1-D CNN) and random forest (RF) models, respectively. Finally, an exponential model of the relationship between the field-measured LAI and MODIS NDVI was applied to obtain the regional LAI, which was input into the trained LSTM, 1-D CNN and RF models to estimate wheat yields within the wheat-growing region. The results showed that the linear correlations between the estimated wheat areas and the statistical areas were significant at both the municipal and county levels. The LSTM model provided more accurate estimates of wheat yields, with higher R2 values and lower root mean square error (RMSE) and mean relative error (MRE) values than the 1-D CNN and RF models. The LSTM model has an inherent advantage in capturing phenological information contained in the time series of the MODIS-derived LAI, which is important for satellite-based crop-yield estimates.

Integration of crop growth model and constraint-based metabolic model predicts metabolic changes over rice plant development under water-limited stress

Rice is a major staple food worldwide and understanding its metabolism is essential for improving crop yield and quality, especially in a changing climate. Constraint-based modelling is an established method for studying metabolism at a systems-level, but one of its limitations is the difficulty in directly integrating certain environmental factors, such as water potential, to the model for predicting metabolic changes in response to environmental changes. Here, we developed a framework to integrate a crop growth model and an upgraded diel multi-organ genome-scale metabolic model of rice to predict the metabolism of rice growth under normal and water-limited conditions. Our model was able to predict distinct metabolic adaptations under water-limited stress compared to normal condition across multiple developmental stages. Our modelling results of dynamic changes in metabolism over the whole plant growth period highlighted key features of rice metabolism under water-limited stress including early leaf senescence, reduction in photosynthesis and significant nitrogen assimilation during grain filling.

May spring seasonal forecasts improve a climate service providing cereal yield predictions?

<p>Following the need of winter cereal farmers from the main producing region (Castilla y León) in Spain to estimate crop yield with at least one season of anticipation, we have developed a climate service based essentially on current and historical meteorological observations, on spring seasonal forecasts from ECMWF System 5 and on the crop growth model AquaCrop. Different experiments have been designed to produce both a synthetic yield database serving as observed truth and three different seasonal forecasting strategies. Calculation of objective verification scores for deterministic and probabilistic crop yield forecasts -including an assessment of their potential economic value- in hindcast mode determines the quality of this service and the differences among forecasting strategies. We demonstrate that the three compared strategies show good skill of wheat yield forecasts at the beginning of July, although the meteorological forcing for Aquacrop simulations between 1st April and 30th June is very different for the three compared strategies. The important role of the memory from previous (autumn and winter) climate conditions carried by the crop growth model is analysed and discussed.  A yearly assessment also allows some preliminary estimation of the value and possible benefits of the service for final users. Finally, we conclude that the simulation synthetically producing the observed truth compares rather well –especially the interannual variability- with other yield data based on surveys and experts estimations although it overestimates yield. Users have played a decisive role in co-design and co-development phases of this climate service. They have also actively intervened in the analysis and evaluation of results.</p><p> </p>