Partitioning Global Surface Energy and Their Controlling Factors Based on Machine Learning

As two competitive pathways of surface energy partitioning, latent (LE) and sensible (H) heat fluxes are expected to be strongly influenced by climate change and wide spread of global greening in recent several decades. Quantifying the surface energy fluxes (i.e., LE and H) variations and controlling factors is still a challenge because of the discrepancy in existing models, parameterizations, as well as driven datasets. In this study, we assessed the ability of random forest (RF, a machine learning method) and further predicted the global surface energy fluxes (i.e., LE and H) by combining FLUXNET observations, climate reanalysis and satellite-based leaf area index (LAI). The results show that the surface energy fluxes variations can be highly explained by the established RF models. The coefficient of determination (R2) ranges from 0.66 to 0.89 for the LE, and from 0.53 to 0.90 for the H across 10 plant functional types (PFTs), respectively. Meanwhile, the root mean square error (RMSE) ranges from 12.20 W/m2 to 21.94 W/m2 for the LE and from 12.05 W/m2 to 22.34 W/m2 for the H at a monthly scale, respectively. The important influencing factors in building RF models are divergent with respect to LE and H, but the solar radiation is common to both LE and H and to all 10 PFTs in this study. We also found a contrasting trend of LE and H: a positive trend in LE and a negative trend in H during 1982–2016 and these contrasting trends are dominated by the elevated CO2 concentration level. Our study suggested an important role of the CO2 concentration in determining surface energy partitioning which is needed to be considered in future studies.

[1]  Haishan Chen,et al.  Uncertainty in land surface temperature simulation over China by CMIP3/CMIP5 models , 2014, Theoretical and Applied Climatology.

[2]  A. VanLoocke,et al.  Terrestrial ecosystems in a changing environment: a dominant role for water. , 2015, Annual review of plant biology.

[3]  Nicholas Siler,et al.  Revisiting the surface-energy-flux perspective on the sensitivity of global precipitation to climate change , 2018, Climate Dynamics.

[4]  Nicolas Viovy,et al.  CRUNCEP Version 7 - Atmospheric Forcing Data for the Community Land Model , 2018 .

[5]  Ren Li,et al.  Estimation of surface energy fluxes in the permafrost region of the Tibetan Plateau based on in situ measurements and the surface energy balance system model , 2020, International Journal of Climatology.

[6]  Wayne A. Morrissey.  The National Oceanic and Atmospheric Administration (NOAA) , 2002 .

[7]  P. Jarvis,et al.  The Direct Effects of Increase in the Global Atmospheric CO2 Concentration on Natural and Commercial Temperate Trees and Forests , 1989 .

[8]  Leo Breiman,et al.  Random Forests , 2001, Machine Learning.

[9]  Xingjie Lu,et al.  An efficient method for global parameter sensitivity analysis and its applications to the Australian community land surface model (CABLE) , 2013 .

[10]  Trevor Hastie,et al.  An Introduction to Statistical Learning , 2013, Springer Texts in Statistics.

[11]  Minha Choi,et al.  Developing a soil water index-based Priestley–Taylor algorithm for estimating evapotranspiration over East Asia and Australia , 2019 .

[12]  S. Seneviratne,et al.  Recent decline in the global land evapotranspiration trend due to limited moisture supply , 2010, Nature.

[13]  Jonas Ardö,et al.  The FLUXNET2015 dataset and the ONEFlux processing pipeline for eddy covariance data , 2020, Scientific Data.

[14]  Xiuliang Yuan,et al.  Process refinement contributed more than parameter optimization to improve the CoLM's performance in simulating the carbon and water fluxes in a grassland , 2020 .

[15]  K. Oleson,et al.  Use of FLUXNET in the Community Land Model development , 2008 .

[16]  Ronggao Liu,et al.  Retrospective retrieval of long-term consistent global leaf area index (1981-2011) from combined AVHRR and MODIS data , 2012 .

[17]  Albert Olioso,et al.  Mapping surface fluxes using airborne visible, near infrared, thermal infrared remote sensing data and a spatialized surface energy balance model , 2002 .

[18]  Yi Y. Liu,et al.  Multi-decadal trends in global terrestrial evapotranspiration and its components , 2016, Scientific Reports.

[19]  Max D. Morris,et al.  Factorial sampling plans for preliminary computational experiments , 1991 .

[20]  José M. Bioucas-Dias,et al.  Quantifying the Uncertainty of Land Surface Temperature Retrievals From SEVIRI/Meteosat , 2010, IEEE Transactions on Geoscience and Remote Sensing.

[21]  K. Caldeira,et al.  Combined climate and carbon-cycle effects of large-scale deforestation , 2006, Proceedings of the National Academy of Sciences.

[22]  R. Dickinson,et al.  The Common Land Model , 2003 .

[23]  Y. Hong,et al.  Vegetation Greening and Climate Change Promote Multidecadal Rises of Global Land Evapotranspiration , 2015, Scientific Reports.

[24]  R. Myneni,et al.  Reduced streamflow in water-stressed climates consistent with CO2 effects on vegetation , 2016 .

[25]  Å. Bakketun.  Data Assimilation of Satellite Soil Moisture Retrievals in the Land Surface Model SURFEX , 2017 .

[26]  V. Brovkin,et al.  Combined biogeophysical and biogeochemical effects of large-scale forest cover changes in the MPI earth system model , 2010 .

[27]  M. Anderson,et al.  Impact of air temperature on physically-based maximum precipitation estimation through change in moisture holding capacity of air , 2018 .

[28]  Dan Li,et al.  Attribution of surface temperature anomalies induced by land use and land cover changes , 2017 .

[29]  Wei Gong,et al.  Assessing parameter importance of the Common Land Model based on qualitative and quantitative sensitivity analysis , 2013 .

[30]  Andy Liaw,et al.  Classification and Regression by randomForest , 2007 .

[31]  P. Ciais,et al.  Characteristics, drivers and feedbacks of global greening , 2019, Nature Reviews Earth & Environment.

[32]  Alessandro Anav,et al.  Global Data Sets of Vegetation Leaf Area Index (LAI)3g and Fraction of Photosynthetically Active Radiation (FPAR)3g Derived from Global Inventory Modeling and Mapping Studies (GIMMS) Normalized Difference Vegetation Index (NDVI3g) for the Period 1981 to 2011 , 2013, Remote. Sens..

[33]  Andrea Saltelli,et al.  An effective screening design for sensitivity analysis of large models , 2007, Environ. Model. Softw..

[34]  L. H. Allen,et al.  Transpiration responses to vapor pressure deficit in well watered ‘slow-wilting’ and commercial soybean , 2007 .

[35]  Ke Zhang,et al.  Increased control of vegetation on global terrestrial energy fluxes , 2020, Nature Climate Change.

[36]  D. Liu,et al.  Quantifying the impacts of pre-occurred ENSO signals on wheat yield variation using machine learning in Australia , 2020, Agricultural and Forest Meteorology.

[37]  K. Moffett,et al.  Remote Sens , 2015 .

[38]  Arden L. Buck,et al.  New Equations for Computing Vapor Pressure and Enhancement Factor , 1981 .

[39]  J. Muller,et al.  Influences of leaf area index and albedo on estimating energy fluxes with HOLAPS framework , 2020, Journal of Hydrology.

[40]  Thomas Nauss,et al.  Importance of spatial predictor variable selection in machine learning applications - Moving from data reproduction to spatial prediction , 2019, Ecological Modelling.

[41]  J. Goodier.  The Concise Encyclopedia of Statistics , 2009 .

[42]  S. Seneviratne,et al.  Global intercomparison of 12 land surface heat flux estimates , 2011 .

[43]  Martin Jung,et al.  The FLUXCOM ensemble of global land-atmosphere energy fluxes , 2018, Scientific Data.

[44]  J. Canadell,et al.  Recent increases in terrestrial carbon uptake at little cost to the water cycle , 2017, Nature Communications.

[45]  Nannan Zhang,et al.  Validation and Application of the Modified Satellite-Based Priestley-Taylor Algorithm for Mapping Terrestrial Evapotranspiration , 2014, Remote. Sens..

[46]  A. Dai,et al.  The uncertainties and causes of the recent changes in global evapotranspiration from 1982 to 2010 , 2017, Climate Dynamics.

[47]  A. Kurban,et al.  The sensitivity of global surface air temperature to vegetation greenness , 2020, International Journal of Climatology.

[48]  Z. Su.  The Surface Energy Balance System (SEBS) for estimation of turbulent heat fluxes , 2002 .

[49]  Shuangcheng Li,et al.  Local cooling and warming effects of forests based on satellite observations , 2015, Nature Communications.

[50]  Zhanqing Li,et al.  A simple method to estimate actual evapotranspiration from a combination of net radiation, vegetation index, and temperature , 2007 .

[51]  A. Luckman,et al.  The Impact of Föhn Winds on Surface Energy Balance During the 2010–2011 Melt Season Over Larsen C Ice Shelf, Antarctica , 2017 .

[52]  Elfatih A. B. Eltahir,et al.  A Soil Moisture–Rainfall Feedback Mechanism: 1. Theory and observations , 1998 .

[53]  H. Tian,et al.  Responses of global terrestrial evapotranspiration to climate change and increasing atmospheric CO2 in the 21st century , 2015 .

[54]  Jie He,et al.  Improving land surface temperature modeling for dry land of China , 2011 .

[55]  Jindi Wang,et al.  Long-Time-Series Global Land Surface Satellite Leaf Area Index Product Derived From MODIS and AVHRR Surface Reflectance , 2016, IEEE Transactions on Geoscience and Remote Sensing.

[56]  Jiyuan Liu,et al.  The moderating or amplifying biophysical effects of afforestation on CO2-induced cooling depend on the local background climate regimes in China , 2018, Agricultural and Forest Meteorology.

[57]  Shilong Piao,et al.  Response of terrestrial evapotranspiration to Earth's greening , 2018, Current Opinion in Environmental Sustainability.

[58]  Maosheng Zhao,et al.  Improvements to a MODIS global terrestrial evapotranspiration algorithm , 2011 .

[59]  Atul K. Jain,et al.  Increased atmospheric vapor pressure deficit reduces global vegetation growth , 2019, Science Advances.

[60]  Y. Lü,et al.  Revegetation in China’s Loess Plateau is approaching sustainable water resource limits , 2016 .

[61]  Pedro M. Domingos.  A Unifeid Bias-Variance Decomposition and its Applications , 2000, ICML.

[62]  N. Verhoest,et al.  GLEAM v3: satellite-based land evaporation and root-zone soil moisture , 2016 .

[63]  W. Oechel,et al.  Energy partitioning between latent and sensible heat flux during the warm season at FLUXNET sites , 2002 .

[64]  Richard G. Allen,et al.  Intercomparison of surface energy fluxes, soil moisture, and evapotranspiration from eddy covariance, large-aperture scintillometer, and modeling across three ecosystems in a semiarid climate , 2018 .

[65]  D. Medvigy,et al.  Strong control of surface roughness variations on the simulated dry season regional atmospheric response to contemporary deforestation in Rondônia, Brazil , 2014 .

[66]  Lisheng Song,et al.  Evaluating Different Machine Learning Methods for Upscaling Evapotranspiration from Flux Towers to the Regional Scale , 2018, Journal of Geophysical Research: Atmospheres.