Skip to main content
ARS Home » Northeast Area » Beltsville, Maryland (BARC) » Beltsville Agricultural Research Center » Hydrology and Remote Sensing Laboratory » Research » Publications at this Location » Publication #353111

Research Project: Integrating Remote Sensing, Measurements and Modeling for Multi-Scale Assessment of Water Availability, Use, and Quality in Agroecosystems

Location: Hydrology and Remote Sensing Laboratory

Title: Assessing variability of corn and soybean yields in central Iowa using high spatiotemporal resolution multi-satellite imagery

Author
item Gao, Feng
item Anderson, Martha
item Daughtry, Craig
item Johnson, D. - National Agricultural Statistical Service (NASS, USDA)

Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 9/14/2018
Publication Date: 9/18/2018
Citation: Gao, F.N., Anderson, M.C., Daughtry, C.S., Johnson, D. 2018. Assessing variability of corn and soybean yields in central Iowa using high spatiotemporal resolution multi-satellite imagery. Remote Sensing of Environment. 10:1489. https://doi.org/10.3390/rs10091489.
DOI: https://doi.org/10.3390/rs10091489

Interpretive Summary: Accurate estimation of crop yield is critical for sustaining agricultural markets and ensuring food security. Remote sensing data have been used to estimate crop yield for decades. However, the value of high spatial and temporal resolution remote sensing data for yield estimation has not been thoroughly investigated due to the lack of these kinds of data sources. This paper evaluates the added value of high temporal and spatial resolution data generated using Landsat, Sentinel-2, MODIS (Moderate Resolution Imaging Spectroradiometer) and multi-sensor data fusion. Our results demonstrate the importance of high temporal and spatial resolution remote sensing data and support the development of new medium resolution sensors for crop yield estimation as conducted by the National Agricultural Statistics Service and Foreign Agricultural Service.

Technical Abstract: Remote sensing data have been used to estimate crop yield for decades. The utility of remote sensing derived vegetation indices (VIs) in crop yield modeling has been generally evaluated at the regional or state level using moderate to coarse resolution (a few hundred to kilo-meters) remote sensing data. Although better able to resolve individual fields, use of medium-resolution data (10-100m) for yield estimation has been more limited due to the low temporal sampling frequency typically provided by these satellite systems (e.g., Landsat). With the launch of Sentinel-2A in 2015 and Sentinel-2B in 2017, however, temporal sampling at this critical field scale is now significantly higher when both systems are used in combination. Furthermore, data fusion approaches have been developed to blend data from different spatial and temporal resolutions, facilitating near daily monitoring of crop conditions at field scale. This paper investigates the impacts of improved temporal sampling afforded by multi-source medium resolution VI datasets on our ability to explain spatial and temporal variability in crop yields reported by the USDA National Agricultural Statistics Service (NASS). The study compares performance of several VI metrics based on both Normalized Difference Vegetation Index (NDVI) and the two-band Enhanced Vegetation Index (EVI2) timeseries, including peak VI and cumulative values. Dataset performance was evaluated in a rain-fed agricultural area in central Iowa (a part of U.S. Corn Belt) using Landsat-MODIS data fusion results (daily 30m resolution) from 2001 to 2015 and the Harmonized Landsat and Sentinel-2 (HLS) data from 2016 and 2017. In addition, Landsat-8, Sentinel-2 and MODIS data from 2016 and 2017 were used to evaluate yield variability over a large area covering 10 agricultural states in the United States in the Google Earth Engine (GEE). Results show that the fused Landsat-MODIS results and the combination of Landsat-8 and Sentinel-2 explain yield variability better than using any single data source alone. EVI2 performs marginally better than NDVI when derived from surface reflectance. The maximum NDVI and EVI2 describe yield variability better than their cumulative values in the central Iowa study area. Results from ten states in 2016 and 2017 further reveal the importance of high frequent observations in explaining yield spatial variability. Even though NDVI and EVI2 are effective in explaining yield variability within the season, the inter-annual variability is more complex and may depend on other environmental factors and management approaches. These findings emphasize the importance of high temporal and spatial resolution remote sensing data and support the development of new medium resolution sensors for agricultural applications.