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 #340199
Title: How does spatial and temporal resolution of vegetation index impact crop yield estimation?

Author
item Gao, Feng
item Anderson, Martha

Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 6/5/2017
Publication Date: N/A
Citation: N/A

Interpretive Summary:

Technical Abstract: Timely and accurate estimation of crop yield before harvest is critical for food market and administrative planning. Remote sensing data have long been used in crop yield estimation for decades. The process-based approach uses light use efficiency model to estimate crop yield. Vegetation index (VI) has been found highly correlated to the fraction of Absorbed Photosynthetically Active Radiation (fAPAR) which determines the amount of incoming photosynthetically active radiation absorbed by plants. Thus VI is directly related to crop biomass and yield in the yield estimation model. In addition, most empirical approaches simply build linear or non-linear relationships between yield and VI for yield estimations. These vegetation indices are normally based on a single data source which is limited in either temporal or spatial resolution. The objective for this study is to investigate the impacts of VI at different spatial and temporal resolutions on yield estimation. In this study, we used Landsat, MODIS and the fused Landsat-MODIS data to evaluate their predictability of crop yield. Landsat data have been operationally used in mapping crop types and acreages. However, the use of Landsat is limited for monitoring crop conditions due to the 16-day revisit cycle and cloud coverage. Coarse resolution data such as AVHRR, MODIS and VIIRS have been used to monitor crop condition and yield estimation. However, the 250m to 1km resolution imageries are too coarse to distinguish individual fields. In order to monitor crop condition at field scale, high spatial and temporal resolution remote sensing data are required. The Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM) can be used to generate the daily VI at Landsat spatial resolution using Landsat and MODIS surface reflectance. Our study area locates in a rain-fed agricultural area in central Iowa. Landsat and MODIS surface reflectance from 2001 to 2015 were ordered and processed. Yields of corn and soybeans from 20 counties in the area reported by National Agricultural Statistics Service (NASS) were used. Using the 15 years of Landsat, MODIS, and fused Landsat-MODIS data, the predictability of crop yield from three data sets were assessed. The added values of high temporal and high spatial resolution data were investigated. Results show that the fused Landsat-MODIS data outperform single data source for both corn and soybeans. EVI2 is generally better than NDVI in yield prediction. Maximum VI shows better predictability than cumulative VI in this area. The best time windows of remote sensing data in yield prediction are found from July to August, which is encouraging for yield prediction since harvest is normally occurred after late September. Inter-annual variations related to crop water stress in the yield estimation will be discussed.