Utility of near-surface phenology in estimating productivity and evapotranspiration across diverse ecosystems

Authors: Denham, Sander; Browning, Dawn; Schreiner-Mcgraw, Adam; Scott, Russell; Dalzell, Brent; Flerchinger, Gerald; Clark, Patrick; Goslee, Sarah; Hoover, David; LITVAK, MARCY - University Of New Mexico; MARITZ, MARGUERITE - University Of Texas - El Paso; Huggins, David; Phillips, Claire; Prueger, John; Alfieri, Joseph; BRACHO, ROSVEL - University Of Florida; SILVEIRA, MARIA - University Of Florida; Whippo, Craig

Submitted to: Journal of Environmental Quality
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/29/2025
Publication Date: 6/2/2025
Citation: Denham, S.O., Browning, D.M., Schreiner-Mcgraw, A.P., Scott, R.L., Dalzell, B.J., Flerchinger, G.N., Clark, P., Goslee, S.C., Hoover, D.L., Litvak, M., Maritz, M., Huggins, D.R., Phillips, C.L., Prueger, J.H., Alfieri, J.G., Bracho, R., Silveira, M., Whippo, C.W. 2025. Utility of near-surface phenology in estimating productivity and evapotranspiration across diverse ecosystems. Journal of Environmental Quality. Article e70043. https://doi.org/10.1002/jeq2.70043.
DOI: https://doi.org/10.1002/jeq2.70043

Interpretive Summary: Environmental change is having a growing impact on farming and other managed ecosystems. To understand how these systems are responding, we need to improve how we monitor and model the flow of carbon and water. Models often use satellite-based vegetation indices (VIs) to estimate important processes like plant growth (GPP) and water loss through evaporation and plant transpiration (ET). While VIs are good at predicting these processes, we still don’t fully understand how well they work across different locations and times. Near-surface cameras, like those in the PhenoCam network, are providing more detailed images of vegetation. These images can help us better understand the link between VIs and GPP or ET. In this study, we analyzed long-term data from 15 sites over 76 site-years using green chromatic coordinate (GCC), a type of VI from PhenoCam images, and compared it to GPP and ET data from eddy covariance measurements at Long-Term Agroecosystem Research (LTAR) sites. We found that the strength of the relationship between GCC and GPP or ET varies a lot across different ecosystems (R² from ~0.1 to ~0.9), with shrub-dominated systems showing the most variation. Overall, GCC was a better predictor of productivity (mean R² = 0.64) than ET (mean R² = 0.54) and worked best in grasslands. Using more detailed VIs, which capture small changes in vegetation that satellites might miss—especially in certain types of production systems—will potentially help us predict ecosystem processes more accurately.

Technical Abstract: Agroecosystems, which include row crops, pasture, and grass and shrub grazing lands, are sensitive to changes in management, weather, and genetics. To better understand how these systems are responding to changes, we need to improve monitoring and modeling carbon and water dynamics. Vegetation Indices (VIs) are commonly used to estimate gross primary productivity (GPP) and evapotranspiration (ET), but these empirical relationships are often location and crop specific. There is a need to evaluate if VIs can be effective and, more general, predictors of ecosystem processes through time and across different agroecosystems. Near-surface photographic (red-green-blue) images from PhenoCam can be used to calculate the VI green chromatic coordinate (GCC) and offer a pathway to improve understanding of field-scale relationships between VIs and GPP and ET. We synthesized observations spanning 76 site-years across 15 agroecosystem sites with PhenoCam GCC and GPP or ET estimates from eddy covariance (EC) to quantify interannual variability (IAV) in the relationship between GPP and ET and GCC across. We uncovered a high degree of variability in the strength and slopes of the GCC ~ GPP and ET relationships (R2 = 0.1 - 0.9) within and across production systems. Overall, GCC is a better predictor of GPP than ET (R2 = 0.64 and 0.54, respectively), performing best in croplands (R2 = 0.91). Shrub-dominated systems exhibit the lowest predictive power of GCC for GPP and ET but have less IAV in slope. We propose that PhenoCam estimates of GCC could provide an alternative approach for predictions of ecosystem processes.