The FLUXCOM ensemble of global land-atmosphere energy fluxes (1812.04951v1)

Abstract: Although a key driver of Earth's climate system, global land-atmosphere energy fluxes are poorly constrained. Here we use machine learning to merge energy flux measurements from FLUXNET eddy covariance towers with remote sensing and meteorological data to estimate net radiation, latent and sensible heat and their uncertainties. The resulting FLUXCOM database comprises 147 global gridded products in two setups: (1) 0.0833${\deg}$ resolution using MODIS remote sensing data (RS) and (2) 0.5${\deg}$ resolution using remote sensing and meteorological data (RS+METEO). Within each setup we use a full factorial design across machine learning methods, forcing datasets and energy balance closure corrections. For RS and RS+METEO setups respectively, we estimate 2001-2013 global (${\pm}$ 1 standard deviation) net radiation as 75.8${\pm}$1.4 ${W\ m^{-2}}$ and 77.6${\pm}$2 ${W\ m^{-2}}$, sensible heat as 33${\pm}$4 ${W\ m^{-2}}$ and 36${\pm}$5 ${W\ m^{-2}}$, and evapotranspiration as 75.6${\pm}$10 ${\times}$ 10$^3$ ${km^3\ yr^{-1}}$ and 76${\pm}$6 ${\times}$ 10$^3$ ${km^3\ yr^{-1}}$. FLUXCOM products are suitable to quantify global land-atmosphere interactions and benchmark land surface model simulations.

• The paper demonstrates that machine learning techniques reliably integrate FLUXNET measurements with remote sensing and meteorological inputs to estimate energy fluxes.
• It employs dual setups—RS and RS+METEO—with multiple algorithms and energy balance corrections, yielding consistent net radiation, sensible heat, and evapotranspiration estimates.
• The findings validate the FLUXCOM ensemble as a robust benchmark for land surface models and offer pathways to reduce uncertainties in global flux estimations.

Machine Learning Estimation of Global Land-Atmosphere Energy Fluxes: An Examination of the FLUXCOM Ensemble

The paper discusses the FLUXCOM initiative's comprehensive approach to estimating global land-atmosphere energy fluxes using machine learning techniques. By integrating FLUXNET eddy covariance tower measurements with remote sensing and meteorological data, this work addresses significant uncertainties in the quantification of net radiation, latent and sensible heat fluxes, and their associated uncertainties.

Methodology

The FLUXCOM ensemble is segmented into two primary setups for generating global products:

1. RS (Remote Sensing) Setup: Utilizes MODIS satellite data with a spatial resolution of 0.0833° and temporal resolution of 8 days.
2. RS+METEO (Remote Sensing plus Meteorological Data) Setup: Combines remote sensing data with meteorological inputs, providing daily updates at 0.5° resolution.

The paper employs a factorial design, encompassing nine machine learning methods within the RS setup and three methods within the RS+METEO setup. Additionally, it considers various energy balance closure corrections— a crucial aspect due to the lack of closure observed in FLUXNET data. For the RS+METEO setup, four global climate forcing datasets are utilized to facilitate comparisons with LSM simulations on identical forcing grounds.

Results

The paper reveals strong numerical results for the global energy fluxes from 2001-2013:

• Net Radiation: 75.8±1.4 W/m² (RS) and 77.6±2 W/m² (RS+METEO).
• Sensible Heat: 33±4 W/m² (RS) and 36±5 W/m² (RS+METEO).
• Evapotranspiration (ET): 75.6±10 × 10³ km²/yr (RS) and 76±6 × 10³ km²/yr (RS+METEO).

These results depict consistent estimates with existing climate products, useful for benchmarking land surface models and advancing empirical upscaling.

Implications and Future Work

By showcasing the robustness of machine learning-based flux products against the uncertainties of LSMs, the paper provides a significant tool for improving global estimations of land-atmosphere interactions. The robust extraction of flux variation patterns across different methods underscores the potential for ML to refine the cross-comparison of model outputs.

The FLUXCOM ensemble's distinctive handling of empirical uncertainties paves the way for refined global simulations, essential for climate modeling and ecosystem monitoring. The temporal irrelevance in the interannual signal invites further research on incorporating dynamic predictors to potentially close gaps that remain due to fixed land surface properties.

Future work could leverage this comprehensive database to explore the effects of climatic anomalies and modifications in land-atmosphere dynamics across temporal scales. The inclusion of rising CO₂ effects and phenology variations could render these models even more pertinent in understanding complex biosphere-atmosphere feedback mechanisms.

Conclusion

The FLUXCOM initiative offers a substantial advancement in the empirical estimation of global land-atmosphere energy fluxes through sophisticated machine learning approaches. It serves as a groundwork for further inquiries into improving the reliability and completeness of global energy and water budget studies, thereby enriching both theoretical understanding and practical applications within the field of climatology and environmental science.