Satellite remote sensing provides an alternative to these two approaches to mitigate the effects of limited data availability because it makes field scale measurements with global coverage. Further, even this calibration with regional crop yields is not always performed, likely because the stresses imposed on crop growth, especially in developing regions, are highly variable and dependent on unknown field-scale management decisions.Gridded crop models perform significantly worse in developing regions. For example, in, two unknown G × E × M factors (the planting date and planting density) and three crop growth model coefficients (the biomass to energy ratio, the harvest index, and the potential heat units) are calibrated based only on goodness-of-fit criteria with United State Department of Agriculture National Agricultural Statistics Service (USDA-NASS) county-level maize yields. Unfortunately, these studies have been limited by generally only using regional yields for calibration. Gridded modeling strategies can potentially reduce the effects of the limited data availability by calibrating to identify locally optimal crop growth model parameters on a regional scale. Upscaling strategies and gridded modeling strategies have been developed to address the limitation on the availability of data however, significant uncertainties remain, especially due to limited agromanagement information. Unfortunately, the detailed agromanagement information to run field-scale crop models is, in general, unavailable at a national or global scale, introducing significant uncertainty into the model predicted effect of soil variability, weather variability, irrigation changes, or fertilization changes on the attainable crop yield. This is because they seek to physically model the major genotype, environment, and management (G × E × M) interactions that affect the individual components of the crop-soil system and, ultimately, the yield. Mechanistic crop growth models are well-suited for the task of analyzing the effect that changing a particular factor will have while keeping the other factors constant. In order to isolate one of these factors, it is necessary to have good estimates of the other factors in the agricultural system being studied. Several applications-such as adaptation to climate change, optimizing agricultural policies, supporting precision agriculture, and reducing yield gaps -require isolating the effect of a particular variable from the other factors affecting crop growth. Understanding the effect that environmental and agromanagement factors-such as weather, soil, and fertilization-have on crop growth is a critical question in agronomy-related fields. We believe that a potential application of this methodology is to develop satellite products to monitor in-season field-scale crop growth on a global scale by reproducing the methodology with field-scale crop growth model simulations (utilizing farmer-recorded field-scale agromanagement data) and collocated high-resolution satellite data (fused with moderate-resolution satellite data). In addition, by comparing to survey-based United State Department of Agriculture (USDA) ground truth data, we showed that the BLSTMs are able to predict actual county-level yields with R 2 values between 0.45 and 0.6 and actual state-level phenological dates (emergence, silking, and maturity) with R 2 values between 0.75 and 0.85. Several other plant, soil, and phenological in-season state variables were also evaluated in the study for their retrievability via k-fold cross validation.
Using k-fold cross validation, we showed that three distinct in-season maize state variables (leaf area index, aboveground biomass, and specific leaf area) can be retrieved with cross-validated R 2 values ranging from 0.4 to 0.8 for significant portions of the season. We evaluated the performance of the BLSTMs through both k-fold cross validation and comparison to regional scale ground-truth yields and phenology.
#APSIM GENERIC CROP TEMPLATE SIMULATOR#
Specifically, we showed that bidirectional long short-term memory networks (BLSTMs) can be trained to predict the in-season state variables and yields of Agricultural Production Systems sIMulator (APSIM) maize crop growth model simulations from collocated Moderate Resolution Imaging Spectroradiometer (MODIS) 500-m satellite measurements over the United States Corn Belt at a regional scale.
In this study, we presented a novel approach to training agronomic satellite retrieval algorithms by utilizing collocated crop growth model simulations and solar-reflective satellite measurements. Due to its worldwide coverage and high revisit time, satellite-based remote sensing provides the ability to monitor in-season crop state variables and yields globally.