Satellite solar-induced chlorophyll fluorescence and near-infrared reflectance capture complementary aspects of dryland vegetation productivity dynamics

Authors: WANG, X. - University Of Arizona; Biederman, Joel; Knowles, John; Scott, Russell - Russ; TURNER, A.J. - University Of Washington; DANNENBERG, M.P. - University Of Iowa; KOHLER, P - California Institute Of Technology; FRANKENBERG, C. - California Institute Of Technology; LITVAK, M.E. - University Of New Mexico; Flerchinger, Gerald; LAW, B.E. - Oregon State University; KWON, H.J. - Oregon State University; REED, S.C. - Us Geological Survey (USGS); PARTON, W.J. - Colorado State University; BARRON-GAFFORD, G.A. - University Of Arizona; SMITH, W.K. - University Of Arizona

Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 12/18/2021
Publication Date: 1/4/2022
Citation: Wang, X., Biederman, J.A., Knowles, J.F., Scott, R.L., Turner, A., Dannenberg, M., Kohler, P., Frankenberg, C., Litvak, M., Flerchinger, G.N., Law, B., Kwon, H., Reed, S., Parton, W., Barron-Gafford, G., Smith, W. 2022. Satellite solar-induced chlorophyll fluorescence and near-infrared reflectance capture complementary aspects of dryland vegetation productivity dynamics. Remote Sensing of Environment. 270. Article 112858. https://doi.org/10.1016/j.rse.2021.112858.
DOI: https://doi.org/10.1016/j.rse.2021.112858

Interpretive Summary: While the vegetation productivity of drylands is important in the US and globally, the satellite remote sensing tools used to assess productivity are known to work poorly in drylands. Here we compared traditional greenness based satellite estimates of productivity with recently-available estimates from satellites measuring sun-induced fluorescence, a more direct measure of plant photosynthesis. We compared these traditional and new technologies with direct ground-based measurements of photosynthesis at 22 sites in diverse dryland ecosystems of the US Southwest. We found that traditional greenness satellite data work best for assessing productivity in relatively low-productivity deciduous sites, whereas the new fluorescence data offer improved performance for higher-productivity sites with evergreen shrubs and trees. Our findings show that integration of these two remote sensing approaches can lead to better dryland productivity assessment than either one alone.

Technical Abstract: Mounting evidence indicates dryland ecosystems play a dominant role in driving the trend and interannual variability of atmospheric carbon dioxide concentrations. Nevertheless, the seasonal dynamics of dryland ecosystem carbon uptake through photosynthesis [gross primary productivity (GPP)], remain highly uncertain. Remote sensing platforms offer powerful tools for reducing this uncertainty. Here we used in-situ GPP measurements to evaluate MODerate resolution Imaging Spectroradiometer (MODIS) normalized difference vegetation index (NDVI), MODIS near infrared reflectance index (NIRv), and TROPOspheric Monitoring Instrument (TROPOMI) solar-induced chlorophyll fluorescence (SIF) as satellite-based proxies of dryland GPP. As a performance benchmark, we linked our cross-index analysis to a synthesis of GPP estimates from a network of 22 western United States (USA) eddy covariance tower sites that span a representative gradient in dryland ecosystem climate and functional composition. We found that, relative to the commonly-used NDVI, NIRv and SIF represented improved GPP proxies that captured complementary aspects of dryland seasonal GPP dynamics. SIF offered the best performance across high-productivity sites (average GPP > ~1g C m-2 d-1, R2=0.71), whereas NIRv offered the best performance across non-evergreen low-productivity sites (R2=0.59). Notably, vegetation reflectance-based indices (NDVI and NIRv) were found to exhibit pre-/post- peak-of-season bias (hysteresis) with GPP that worsened with the total fraction of woody vegetation, likely due to decoupling between evergreen vegetation reflectance and GPP during seasonal transitions (i.e., the start and end of the growing season) for drought-tolerant dryland shrubs. In general, reflectance-based proxies have limited ability to capture GPP dynamics for regions with significant evergreen components, which account for 26% of the western US, whereas chlorophyll fluorescence-based proxies (SIF) have limited ability to capture GPP dynamics across low- productivity regions, which account for 39% of the western US. Our findings demonstrate that integration of NIRv and SIF observations offers a pathway towards improved representation of dryland dynamics in satellite-based GPP algorithms.