基于多源数据与多模型集成的流域人为蒸散发变异评估

Assessing human-induced evapotranspiration change based on multi-source data and Bayesian model averaging at the basin scale

人类活动对流域蒸散发的干扰日益显著,然而实测蒸散发数据稀少,且尺度差异与空间异质性等问题限制了大尺度陆面模型与遥感产品在人为蒸散发评估中的适用性与可靠性。本文提出了一种基于多源数据与贝叶斯模型平均的人为蒸散发变异评估框架,并应用于珠江流域。结果表明,通过综合利用地面观测(降水、径流、潜热通量)、社会统计(水资源开发利用)、卫星遥感(GRACE卫星)、与陆面模型(Noah、CLSM、VIC、ERA-Interim)等数据集,并以概率方式集成多个优选模型的模拟结果,可有效提高评估结果的可靠性。2003—2016年,人类活动导致珠江流域蒸散发增大21%(137 mm/a),人为变异的年内最大值与最小值分别出现在5月(+38 mm/月)和7月(-27 mm/月),月际振荡可能主要受灌溉耗水、水库调蓄等因素影响。

Anthropogenic disturbance on terrestrial evapotranspiration intensifies with socioeconomic develop-ments.However,on-site evapotranspiration measurement is not readily available in many areas,and the ap-plicability of data from large-scale land surface models and remote-sensing products is restrained by chal-lenges in resolution gaps and spatial heterogeneity.In this paper,an evaluation framework of human-in-duced evapotranspiration change based on multi-source data and the Bayesian model averaging method is proposed.The case study in the Pearl River basin of China indicates that the reliability of assessment can be significantly improved by assimilating datasets derived from ground observations(precipitation,stream-flow),water census,remote-sensing(GRACE satellite),and land surface models(Noah,CLSM,VIC,ERA-Interim),and synthesizing outputs from individual models in a probabilistic way.The results show that mean annual evapotranspiration was enhanced by 21%(137 mm/a)by human activities during 2003-2016,with the maximum and minimum impact found in May(+38 mm/month)and July(-27 mm/month),respectively.Irrigation water consumption and reservoir regulation were likely to be the major fac-tors that dominated monthly variations in human-induced evapotranspiration change.