摘要:
土壤水是衔接大气、地表、土壤和地下的水分转换和循环的核心,土壤湿度是全球气候观测系统的基本气候变量之一,在区域尺度的水分和能量交换中起着重要作用。根区土壤湿度的估算和时空变化特征的获取,对农业产量评估、洪水和干旱预测、水土保持等均具有重要意义。以西辽河流域作为研究区,基于人工神经网络,以遥感表层土壤湿度、累积降水量、累积日最高温、累积日最低温、相对湿度、日照时长、云覆盖度、风速、土壤属性、归一化植被指数、实际蒸散发量等作为解释变量,以站点实测的根区土壤湿度作为目标变量,采用2013—2018年的数据训练模型,估算研究区内2019—2020年每天的根区土壤湿度。结果表明,基于人工神经网络的根区土壤湿度估算值与站点实测根区土壤湿度之间的平均均方根误差为0.056 7 m3/m3,平均相关系数为0.611 7,表明人工神经网络模型能够有效地估算西辽河流域内的根区土壤湿度。研究发现土壤湿度的变化量与降水量密切相关。
Abstract:
Soil moisture is the core of water conversion and circulation that connects the atmosphere, surface, soil, and subsurface. As a basic climate variable of the global climate observing system, soil moisture plays a vital role in regional-scale water and energy exchange. The estimation of root zone soil moisture (RZSM) and the analysis of its spatio-temporal variations are of great significance for crop yield assessment, flood and drought prediction, and soil and water conservation. Based on the artificial neural network (ANN), this study estimated the daily RZSM in the Western Liaohe River basin during 2019—2020 with remote sensing image-based surface soil moisture, cumulative precipitation, cumulative daily maximum and minimum temperatures, relative humidity, sunshine duration, cloud coverage, wind speed, soil attributes, normalized difference vegetation index, and actual evapotranspiration as explanatory variables, the in-situ measured RZSM as the target variable, and the 2013—2018 data used for model training. The estimated results show that the average RMSE and average R between the RZSM estimated based on ANN and the in-situ measured RZSM were 0.056 7 m3/m3 and 0.611 7, respectively. Therefore, the ANN can effectively estimate the RZSM in the Western Liaohe River basin. In addition, this study shows that the variation in the soil moisture is closely related to precipitation.
下载: