Global ecosystem-scale plant hydraulic traits retrieved using model-data fusion

Droughts are expected to become more frequent and severe under climate change, increasing the need for accurate predictions of plant drought response. This response varies substantially, depending on plant properties that regulate water transport and storage within plants, i.e., plant hydraulic traits. It is, therefore, crucial to map plant hydraulic traits at a large scale to better assess drought impacts. Improved understanding of global variations in plant hydraulic traits is also needed for parameterizing the latest generation of land surface models, many of which explicitly simulate plant hydraulic processes for the first time. Here, we use a model-data fusion approach to evaluate the spatial pattern of plant hydraulic traits across the globe. This approach integrates a plant hydraulic model with data sets derived from microwave remote sensing that inform ecosystem-scale plant water regulation. In particular, we use both surface soil moisture and vegetation optical depth (VOD) derived from the X-band Japan Aerospace Exploration Agency (JAXA) Advanced Microwave Scanning Radiometer for Earth Observing System (EOS; collectively AMSR-E). VOD is proportional to vegetation water content and, therefore, closely related to leaf water potential. In addition, evapotranspiration (ET) from the Atmosphere-Land Exchange Inverse (ALEXI) model is also used as a constraint to derive plant hydraulic traits. The derived traits are compared to independent data sources based on ground measurements. Using the K-means clustering method, we build six hydraulic functional types (HFTs) with distinct trait combinations - mathematically tractable alternatives to the common approach of assigning plant hydraulic values based on plant functional types. Using traits averaged by HFTs rather than by plant functional types (PFTs) improves VOD and ET estimation accuracies in the majority of areas across the globe. The use of HFTs and/or plant hydraulic traits derived from model-data fusion in this study will contribute to improved parameterization of plant hydraulics in large-scale models and the prediction of ecosystem drought response.