Estimating canopy dynamics and productivity of upper Midwestern agroecosystems using remotely sensed observations

Background/Question/Methods

When validated against field measurements, remotely sensed data can estimate seasonal canopy dynamics, crop productivity and carbon cycling in terrestrial ecosystems over large areas. Regional crop yield maps can then be generated from these calculations to estimate carbon dynamics in agroecosystems. This research employed a method for estimating canopy dynamics and crop yields from satellite observations in corn and soybean agroecosystems in Wisconsin. Field measurements of continuous fraction of photosynthetically active radiation (fPAR) were made using Li-COR light bar sensors in corn and soybean agroecosystems over the 2010 and 2011 growing seasons. Using harvest indices with the fPAR measurements, crop yields were calculated.

Moderate Resolution Imaging Spectroradiometer (MODIS) Normalized Difference Vegetation Index (NDVI) data obtained for the same time period were used to model the dynamics of fPAR in the same locations. These values, in combination with harvest indices for corn and soybeans, were used to estimate crop yields. These estimates were compared to farmer reported and field-derived yields as a means to validate this methodology of using remotely sensed data to accurately estimate crop productivity.

Results/Conclusions

MODIS NDVI provided an accurate estimate of canopy dynamics and crop yields for the first field season. Average daily fPAR during the growing season (DOY 133 - 260) ranged from 1 to 98% light interception in field measurements and from 55 to 95% interception in MODIS derived measurements. This represented an accurate capture, by MODIS NDVI, of seasonal canopy dynamics of emergence, leaf expansion, canopy closure, and crop senescence.

Results also demonstrated reasonable agreement among the MODIS derived, field derived, and farmer reported yields. MODIS and field derived yield estimates differed by <10%, with field-derived yields averaging 13% above farmer reported yields (mean field, reported yields = 0.97, 0.86 kg/m²). MODIS derived yields overestimated farmer reported yields by 22% (mean MODIS, reported yields = 1.04, 0.86 kg/m²). We speculate this approach can be improved by using remotely sensed products that have a smaller spatial resolution, such as 30 meter LANDSAT. Crop maps developed from remotely sensed data provide powerful tools for monitoring agroecosystem dynamics on multiple spatial scales. Combining such tools with climate data in spatially based ecological process models can improve understanding of carbon and nutrient dynamics and crop productivity on multiple spatial scales.