Using multi-source satellite image series for a data-informed crop yield estimation at different management scales
Abstract
Timely monitoring of agricultural production and early yield predictions are essential for food security. Remotely sensed time-series could be used to study the variability in crop growth and agricultural production. However, the choice of remotely sensed data and methods is still an issue, as different datasets have different spatiotemporal characteristics. Our primary goal was to test different algorithms and several remotely sensed time-series datasets for yield estimation in U.S. at two spatial scales: regional (county) and field level. For a county-level analysis, MODIS-based surface reflectance, Land Surface Temperature (LST), and Evapotranspiration (ET) time series were used as input datasets. Field-level analysis was carried out using NASA's Harmonized Landsat Sentinel-2 (HLS) product. For this purpose, 3D convolutional neural network (CNN) and CNN followed by long-short term memory (LSTM) were used. The models were trained using time series of yield data at county (2008-2019) and field level (2018,2019). For county-level analysis, the CNN-LSTM model had the highest accuracy, with a mean percentage error of 10.3% for maize and 9.6% for soybean. Integrating several remote sensing time series improved the CNN-LSTM model performance for maize yield estimation. The results showed that when incorporating LST and ET, the highest estimation accuracy could be achieved at county scale with mean percentage error lower than 5%, especially for counties where maize is the dominant crop. This was the case when using data from April till August. This model presented robust results for the year 2012, which is considered a drought year. In the case of field-level analysis, all models achieved accurate results with R2 exceeding 0.8 when data from mid growing season were used. Field level output retained spatial detail, which allowed for intra-field level assessment of yield variability. The integration of Normalized Difference Vegetation Index (NDVI) and Normalized Difference Moisture Index (NDMI) in addition to visible and near-infrared reflectance improved the accuracy of all models at a field level. The results highlight the potential of using satellite data and historical yield observations for yield estimation at different management scales.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2020
- Bibcode:
- 2020AGUFMB013...04G
- Keywords:
-
- 0402 Agricultural systems;
- BIOGEOSCIENCES;
- 0428 Carbon cycling;
- BIOGEOSCIENCES;
- 0495 Water/energy interactions;
- BIOGEOSCIENCES;
- 1843 Land/atmosphere interactions;
- HYDROLOGY