[A-HW22] Advancements in quantification of natural and anthropogenic influences on terrestrial water cycle

[AHW22-P03] Assessment of the Role of Hydroclimatic Factors on India's Evapotranspiration: A Performance Weightage Ensemble Approach

*Abhishek Chakraborty¹、Sekhar Muddu^1,2、Lakshminarayana Rao³ (1.Interdisciplinary Centre for Water Research (ICWaR), Indian Institute of Science (IISc) Bangalore, Bengaluru, Karnataka, India、2.Department of Civil Engineering, Indian Institute of Science (IISc) Bangalore, Bengaluru, Karnataka, India、3.Centre for Sustainable Technologies (CST), Indian Institute of Science (IISc) Bangalore, Bengaluru, Karnataka, India)

キーワード：Evapotranspiration, Performance weightage ensembles, Sequential learning algorithms, Uncertainty

Evapotranspiration (ET) acts as one of the most significant components of the terrestrial water cycle. ET is primarily mediated by biotic-abiotic factors like plants and soil and controlled by the interplay of hydroclimatic-anthropogenic perturbations. On a multidecadal scale, the primary drivers of ET are climate change (CC), land-use change (LUC), elevated CO₂ concentrations (eCO₂), and nitrogen deposition (Ndep). The impacts of these drivers on ET are usually simulated through terrestrial ecosystem models (TEMs). The knowledge of these impacts becomes essential, especially in the case of India, which has undergone a drastic transformation in the past few decades due to population expansion, land use change primarily into croplands, rapid urbanization, intensification of irrigation, and groundwater exploitation. The simulation of ET from multi-model ensembles is preferred over a single model as they can help reduce uncertainty (Schwalm et al., 2015, Geophys Res Lett). Most TEM-based studies have employed equal weightage ensembles (ensembles prepared by an arithmetic average of models) (Jia et al., 2020, Earth Syst Dynam); however, not all models perform equally. This leads to performance weightage (PW) ensembles used thoroughly for climate models (Knutti et al., 2017, Geophys Res Lett). However, the usage of PW ensembles is still limited in simulating ecosystem water fluxes. SLAs have been used extensively to successfully simulate climate parameters (Strobach and Bel, 2020, Nat Commun) and can help increase the accuracy and reliability of ensemble predictions of ET over India. This study uses SLA-based PW schemes to assess the performance, impacts of drivers, and the associated uncertainties in the simulation of the spatiotemporal variability of ET across India and its major biomes (croplands, forests, grasslands, and shrublands) over the period 1950-2010.

We utilized thirteen TEMs from the Multi-scale Synthesis and Terrestrial Model Intercomparison Project (MsTMIP). We considered the raw as well as bias-corrected ET values of the models for four different forcings: CC, CC+LUC, CC+LUC+eCO₂, and CC+LUC+eCO₂+Ndep. We further utilized exponentiated weighted and exponentiated gradient averages as the two SLA schemes on both the raw and bias-corrected models. We varied the learning period from 5 to 30 years over 1950-1980 and validated the resultant SLA-based PW ensemble results over 1980-2010 through FLUXCOM RS+METEO observational setup (Jung et al., 2019, Sci Data). The results indicated improvement in performance and a reduction in uncertainty through the usage of such PW ensemble schemes. We also observed varying roles of hydroclimatic parameters over the decades on India's ET on an annual and seasonal scale for the different homogeneous meteorological regions. Through this study, an improved understanding of the spatiotemporal variability and responses to hydroclimatic factors is expected, which could help policymakers devise plans to combat global warming.