Potential of remote sensing surface temperature- and evapotranspiration-based land-atmosphere coupling metrics for land surface model calibration

Authors: ZHOU, J. - Tsinghua University; YANG, K. - Tsinghua University; Crow, Wade; DING, J. - Tianjin University; ZHAO, L. - Southwest University; FENNG, H. - Hohai University; ZOU, M. - Shanghai Normal University; LU, H. - Tsinghua University; TANG, R. - University Of Chinese Academy Of Sciences

Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/26/2023
Publication Date: 4/1/2023
Citation: Zhou, J., Yang, K., Crow, W.T., Ding, J., Zhao, L., Fenng, H., Zou, M., Lu, H., Tang, R. 2023. Potential of remote sensing surface temperature- and evapotranspiration-based land-atmosphere coupling metrics for land surface model calibration. Remote Sensing of Environment. 291. Article 113557. https://doi.org/10.1016/j.rse.2023.113557.
DOI: https://doi.org/10.1016/j.rse.2023.113557

Interpretive Summary: Land surface models attempt to represent the correct relationship between land surface states (e.g., soil moisture and temperature) and land surface water fluxes (e.g., runoff and evapotranspiration). If land models can successfully reproduce these relationships, they can contribute significantly to the forecasting skill of numerical weather prediction models and the ability of climate models to capture future trends in the terrestrial water cycle. However, recent evidence suggests that land surface models typically do a poor job of capturing these state/flux relationships. This paper presents a new strategy based on the combined use of microwave and thermal-infrared remote sensing that allows us to globally assess how accurately land surface models are capturing the relationship between soil moisture and evapotranspiration. Results show that the application of the approach can improve the ability of a land model to estimate evapotranspiration during summertime conditions in the United States. As such, this approach provides a potentially valuable pathway for improving the accuracy of land surface models - and thus numerical weather forecasts and climate assessments that depend, in part, on these models.

Technical Abstract: Imperfect land physics introduces significant levels of uncertainty into current land surface models (LSMs) and can cause bias in the representation of land-atmosphere coupling strength (rho). When LSMs are coupled with atmospheric prediction models, such errors will eventually degrade the accuracy of lower atmosphere forecasts. Here, we investigate the potential of two remote sensing (RS)-based references rho for addressing LSM rho bias. Both rho references are based on a newly proposed two-system approach for eliminating the impact of random in RS retrievals and quantified using the temporal correlations of soil moisture (SM) versus both evapotranspiration (ET) and the diurnal amplitude of surface temperature (dT), respectively. Experiments are conducted to calibrate an off-line LSM individually against each resulting rho reference and using a combination of both dT- and ET-represented rho references. The resulting calibrated LSM is further evaluated using independent ground-based ET observations and RS dT retrievals. Results show that although dT- and ET-represented references rho are physically consistent across space, model calibration results based on them are quite different. Specifically, the calibration experiment targets on ET-represented rho outperforms that targeting dT-represented rho in ET and dT modeling. Diagnostic results indicate that the failure of dT-based calibration experiments is due to confounding impacts of transpiration/evapotranspiration partitioning error and large dT uncertainties in model. However, results also confirm the potential of both dT- and ET-represented references rho for jointly diagnosing and understanding LSM rho bias. As a result, we suggest diagnosing LSM rho bias using both ET- and dT-represented references rho – but calibrating LSM using only ET-represented reference rho data.