Projecting potential evapotranspiration change and quantifying its uncertainty under future climate scenarios: A case study in southeastern Australia

JOURNAL OF HYDROLOGY

卷:584

Article

Shi, Lijie^1,2  Feng, Puyu^1,2  Wang, Bin²  Liu, De Li^2,3,4  Cleverly, James¹  Fang, Quanxiao⁵  Yu, Qiang^1,5,6 

[1] Univ Technol Sydney, Fac Sci, Sch Life Sci, POB 123, Sydney, NSW 2007, Australia

[2] Wagga Wagga Agr Inst, NSW Dept Primary Ind, Wagga Wagga, NSW 2650, Australia

[3] Univ New South Wales, Climate Change Res Ctr, Sydney, NSW 2052, Australia

[4] Univ New South Wales, ARC Ctr Excellence Climate Syst Sci, Sydney, NSW 2052, Australia

[5] Northwest A&F Univ, State Key Lab Soil Eros & Dryland Farming Loess P, Yangling 712100, Shaanxi, Peoples R China

[6] Univ Chinese Acad Sci, Coll Resources & Environm, Beijing 100049, Peoples R China

关键词: Potential evapotranspiration;    Empirical ETp models;    Random forest;    Uncertainty;    Climate change;    Southeastern Australia;   

DOI  :  10.1016/j.jhydrol.2020.124756

来源: Elsevier

【 摘 要 】

Projecting the likely change of potential evapotranspiration (ETp) under future climate scenarios is crucial for quantifying the impacts of climate change on the hydrologic cycle and aridity conditions. However, there are different sources of uncertainty in projecting future ETp that may arise from global climate models (GCMs), emission scenarios, and multiple ETp models used. In this study, we developed three random forest-based (RF-based) ETp models with solar radiation and air temperature at eight climatic stations in southeastern Australia. With Penman model as the benchmark, their performance was firstly compared with four empirical models (Jensen-Haise, Makkink, Abtew, and Hargreaves), which requires the same meteorological inputs. In general, the RF-based ETp models showed better performance in ETp estimates across all stations, with coefficients of determination (R-2) ranging from 0.68 to 0.92, root mean square errors (RMSE) ranging from 0.58 mm day(-1) to 1.46 mm day(-1), and relative mean bias errors (rMBE) ranging from -16.10% to 9.73%. The RF-based and empirical models were then used to project future ETp for the eight stations based on statistically downscaled daily climatic data from 34 GCMs under two different representative concentration pathways (RCP4.5 and RCP8.5). All models indicated that ETp was likely to increase at the eight stations. The ensemble increases of mean ETp across eight stations ranged from 33 mm year(-1) (2.1%, 2040s) to 129 mm year(-1) (9.2%, 2090s) and from 43 mm year(-1) (2.8%, 2040s) to 248 mm year(-1) (17.6%, 2090s) under RCP4.5 and under RCP8.5, respectively. In addition, we also quantified uncertainties in ETp projections originating from ETp models, GCMs, RCPs, and their combined effects using the analysis of variance (ANOVA) method. Results showed that RCP-related uncertainty contributed the most to projected ETp uncertainty (around 40% for most stations) while GCM-related and ETp model-related uncertainties accounted for roughly equal amounts of projected ETp uncertainty (10%-30%). This study demonstrated the better performance of RF-based ETp models. It is advisable to use multiple ETp models driven by various GCMs under different RCPs to produce reliable projections of future ETp.

【 授权许可】

Free