Selection of Regional Crop Yield Predictors from Earth Observation Products - Assessment with an operational Canadian Crop Yield Forecaster
Abstract
Crop health indices derived from Earth Observation (EO) satellites have long been used to predict crop yield. Among the various EO-based indices, the Normalized Difference Vegetation Index (NDVI), especially those obtained from the Advanced Very High Resolution Radiometer (AVHRR) or from the MODerate-resolution Imaging Spectroradiometer (MODIS) are the most widely used for operational crop yield forecasting, due to their relatively longer history and better temporal and spatial coverage. The Canadian Crop Yield Forecaster (CCYF) is a statistically-based crop yield modelling tool which provides regional crop yield outlooks during the growing season. The model currently uses NDVI from AVHRR and station-based agroclimate indices as potential predictors, and is further built with robust regression and cross-validation algorithms. Although AVHRR NDVI is frequently selected as the primary yield predictor for many crops, there are still crops at certain regions where yield forecasting mainly relies on ground based climate information, sometimes with limited forecasting skill. Recent studies have suggested that some newly developed and readily available EO based crop stress or soil moisture indices have strong correlations with crop yield, and thus could potentially be used as yield predictors. This study uses the AVHRR NDVI and Agroclimate indices as the baseline indices in the CCYF modelling platform, against which, three additional EO products were assessed: the NDVI and an Evaporative Stress Index (ESI) both derived from MODIS and a composite satellite soil moisture index developed by the European Space Agency Climate Change Initiative (ESA-CCI). Five major crops (spring wheat, canola, barley, corn for grain and soybeans) in Canada were used as testing crops. The preliminary results showed that no single predictor prevailed over all the crops across all the regions. However, improvements were found for many crops in certain regions with selected EO indicators or their combinations. Introducing these EO indices to the CCYF improved the overall forecasting skill of all five tested crops and increased the portion of EO based predictors over the station-based agroclimate predictors. This is important since reduced reliance on station-based data can improve the yield prediction at finer spatial scales.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2018
- Bibcode:
- 2018AGUFMGC44B..04Z
- Keywords:
-
- 0402 Agricultural systems;
- BIOGEOSCIENCESDE: 1630 Impacts of global change;
- GLOBAL CHANGEDE: 1640 Remote sensing;
- GLOBAL CHANGEDE: 6309 Decision making under uncertainty;
- POLICY SCIENCES