Skip to main content
ARS Home » Plains Area » Fort Collins, Colorado » Center for Agricultural Resources Research » Soil Management and Sugarbeet Research » Research » Publications at this Location » Publication #345757

Research Project: Management Practices for Long Term Productivity of Great Plains Agriculture

Location: Soil Management and Sugarbeet Research

Title: Simple models to predict grassland ecosystem C exchange and actual evapotranspiration using NDVI and environmental variables

Author
item Del Grosso, Stephen - Steve
item Parton, William - Colorado State University
item Derner, Justin
item Chen, Maosi - Colorado State University

Submitted to: Agricultural and Forest Meteorology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/5/2017
Publication Date: 11/20/2017
Citation: Del Grosso, S.J., Parton, W.J., Derner, J.D., Chen, M. 2017. Simple models to predict grassland ecosystem C exchange and actual evapotranspiration using NDVI and environmental variables. Agricultural and Forest Meteorology. 249:1–10, doi.org/10.1016/j.agrformet.2017.11.007.

Interpretive Summary: Semiarid grasslands contribute significantly to net terrestrial carbon flux as plant productivity and heterotrophic respiration in these moisture-limited systems are correlated with metrics related to water availability (e.g., precipitation, Actual EvapoTranspiration or AET). These factors are also correlated with remotely sensed metrics such as the Normalized Difference Vegetation Index (NDVI). We used measurements of growing season net ecosystem exchange of carbon (NEE), NDVI, precipitation, and volumetric soil water content (VSWC) from grazed pastures in the Colorado, shortgrass steppe to quantify the correlation of NEE with these driving variables. NDVI explained 60 and 40% of the variability in daytime and nighttime NEE, respectively, on non-rain days; these correlations were reduced to 41 and 15%, respectively, on rain days. Daytime NEE was almost always negative (sink) on non-rain days but positive on most rain days. In contrast, nighttime NEE was always positive (source), across rain and non-rain days. A model based on NDVI, VSWC, daytime vs. nighttime, and rain vs. non-rain days explained 48% of observed variability in NEE at a daily scale; this increased to 62% and 77%, respectively, at the weekly and monthly scales. NDVI explained 50-52% of the variability in AET regardless of rain or non-rain days. A model based on NDVI, VSWC, Potential EvapoTranspiration (or PET), and rain vs. non-rain days explained 70% of the observed variability in AET at a daily scale; this increased to 90 and 96%, respectively, at weekly and monthly scales. We conclude that remotely-sensed NDVI is a robust tool, when combined with VSWC and knowledge of rain events, for predicting NEE and AET across multiple temporal scales in semiarid grasslands

Technical Abstract: Semiarid grasslands contribute significantly to net terrestrial carbon flux as plant productivity and heterotrophic respiration in these moisture-limited systems are correlated with metrics related to water availability (e.g., precipitation, Actual EvapoTranspiration or AET). These variables are also correlated with remotely sensed metrics such as the Normalized Difference Vegetation Index (NDVI). We used measurements of growing season net ecosystem exchange of carbon (NEE), NDVI from MODIS and AVHRR, precipitation, and volumetric soil water content (VSWC) from grazed pastures in the semiarid, shortgrass steppe to quantify the correlation of NEE with these driving variables. MODIS NDVI explained 60 and 40% of the variability in daytime and nighttime NEE, respectively, on non-rain days; these correlations were reduced to 41 and 15%, respectively, on rain days. Daytime NEE was almost always negative (sink) on non-rain days but positive on most rain days. In contrast, nighttime NEE was always positive (source), across rain and non-rain days. A model based on MODIS NDVI, VSWC, daytime vs. nighttime, and rain vs. non-rain days explained 48% of observed variability in NEE at a daily scale; this increased to 62% and 77%, respectively, at the weekly and monthly scales. MODIS NDVI explained 50-52% of the variability in AET regardless of rain or non-rain days. A model based on MODIS NDVI, VSWC, Potential EvapoTranspiration (or PET), and rain vs. non-rain days explained 70% of the observed variability in AET at a daily scale; this increased to 90 and 96%, respectively, at weekly and monthly scales. Models based on AVHRR NDVI showed similar patterns as those using MODIS, but correlations with observations were lower. We conclude that remotely-sensed NDVI is a robust tool, when combined with VSWC and knowledge of rain events, for predicting NEE and AET across multiple temporal scales (day to season) in semiarid grasslands