Upscaling dryland carbon and water fluxes with artificial neural networks of optical, thermal, and microwave satellite remote sensing

Abstract

Abstract. Earth's drylands are home to more than two billion people, provide key ecosystem services, and exert a large influence on the trends and variability in Earth's carbon cycle. However, modeling dryland carbon and water fluxes with remote sensing suffers from unique challenges not typically encountered in mesic systems, particularly in capturing soil moisture stress. Here, we develop and evaluate an approach for the joint modeling of dryland gross primary production (GPP), net ecosystem exchange (NEE), and evapotranspiration (ET) in the western United States (US) using a suite of AmeriFlux eddy covariance sites spanning major functional types and aridity regimes. We use artificial neural networks (ANNs) to predict dryland ecosystem fluxes by fusing optical vegetation indices, multitemporal thermal observations, and microwave soil moisture and temperature retrievals from the Soil Moisture Active Passive (SMAP) sensor. Our new dryland ANN (DrylANNd) carbon and water flux model explains more than 70 % of monthly variance in GPP and ET, improving upon existing MODIS GPP and ET estimates at most dryland eddy covariance sites. DrylANNd predictions of NEE were considerably worse than its predictions of GPP and ET likely because soil and plant respiratory processes are largely invisible to satellite sensors. Optical vegetation indices, particularly the normalized difference vegetation index (NDVI) and near-infrared reflectance of vegetation (NIRv), were generally the most important variables contributing to model skill. However, daytime and nighttime land surface temperatures and SMAP soil moisture and soil temperature also contributed to model skill, with SMAP especially improving model predictions of shrubland, grassland, and savanna fluxes and land surface temperatures improving predictions in evergreen needleleaf forests. Our results show that a combination of optical vegetation indices and thermal infrared and microwave observations can substantially improve estimates of carbon and water fluxes in drylands, potentially providing the means to better monitor vegetation function and ecosystem services in these important regions that are undergoing rapid hydroclimatic change.

X-BASE: the first terrestrial carbon and water flux products from an extended data-driven scaling framework, FLUXCOM-X

“…2 cannot be quantitatively compared to previous cross validation results in FLUXCOM as the training data are not the same. However, qualitatively the accuracy gradient among fluxes as well as along scales of variability corresponded to patterns identified in FLUXCOM and in comparable empirical modeling activities (Jung et al, 2011;Tramontana et al, 2016;Virkkala et al, 2021;Dannenberg et al, 2023).…”

“…By today, the empirical up-scaling concept has been implemented for a series of regional and global scale applications, each of them adopting disparate and individual methodological choices (e.g. Ichii et al, 2017;Yao et al, 2018;Joiner and Yoshida, 2020;Virkkala et al, 2021;Dannenberg et al, 2023;Burton et al, 2023). These potentially important choices relate to data treatment (quality control, gap-filling, processing pathways), ingestion (sampling, as well as matching EC and spaceborn observations), and methodological configurations (machine learning methods and their training configuration, choice of predictor variables, resolution).…”

CEDAR-GPP: spatiotemporally upscaled estimates of gross primary productivity incorporating CO2 fertilization

“…500m). Alternatively, models could be trained at a high resolution and applied to the coarse resolution to reduce computation and storage requirements (Dannenberg et al, 2023). However, this approach does not address inherent scaling errors in coarse-resolution satellite images (Yan et al, 2016a;Dong et al, 2023).…”

“…Furthermore, information about resource limitations and stress factors can be crucial for certain biomes and/or under specific conditions (Stocker et al, 2018). For example, Dannenberg et al (2023) found that incorporating LST from MODIS and soil moisture from the SMAP satellite datasets improved the machine learning estimation accuracy of GPP in drylands from R 2 ~ 0.4 to 0.7 for dryland sites in North America. CEDAR-GPP integrated multi-source satellite observations (optical, thermal, microwave) as well as climate variables to obtain comprehensive information about GPP dynamics.…”

“…Over the past three decades, global networks of eddy covariance flux towers collected in situ carbon flux measurements that allow for accurate estimates of GPP, providing valuable insights into photosynthesis dynamics under various environmental conditions (Baldocchi, 2020;Beer et al, 2010). To quantify and understand GPP at scales and locations beyond the ~ 1km2 flux tower footprints, machine learning has been employed with gridded satellite and climate datasets to upscale site-based measurements and produce wall-to-wall GPP maps (Yang et al, 2007;Xiao et al, 2008;Jung et al, 2011;Tramontana et al, 2016;Joiner and Yoshida, 2020;Zeng et al, 2020;Dannenberg et al, 2023). This approach provides important observational constraints of global carbon dynamics, complementing process-based and semi-process-based modeling such as Terrestrial biosphere models or the Light Use Efficiency (LUE) models (Beer et al, 2010;Jung et al, 2017;Gampe et al, 2021).…”