LSTM变体模型在径流预测中的性能及其可解释性

基于湘江流域1999—2013年实测水文气象数据,采用LSTM模型和其变体模型研究多个预见期下不同输入变量和不同模型结构对径流预测结果的影响,评估LSTM模型及其变体模型在短期径流预测中的性能,基于排列重要性法和积分梯度法探究了LSTM模型对流域径流预测的可解释性。结果表明：在历史径流输入数据的基础上增加有效的水文气象变量输入,可以明显改善模型的预测效果,输入变量的改变比模型结构的差异对预测结果的影响更大;随着预见期的增大,降水数据的加入对预测效果表现出不同程度的提升,预见期为1 d时,预测结果的纳什效率系数(NSE)提升2.0%,预见期为2～4 d时,NSE提升可达13.6%;降水和历史径流在预测中起着重要的作用,而前期湿润条件与降水事件的共同作用是湘江流域洪水的主要诱发因素;LSTM模型可反映两种不同的输入输出关系,这两种关系对应于近期降雨和历史降雨两种洪水诱发机制。     

基于组合赋权模糊物元可拓模型的渭河流域甘肃段河流健康评价

从水文完整性、化学完整性、形态结构完整性、生物完整性及社会服务功能5个准则层出发,构建了渭河流域甘肃段河流健康评价指标体系;采用基于改进层次分析法和Critic法的组合赋权法和模糊物元可拓模型计算河流健康状况评价等级,对渭河流域甘肃段河流健康状况进行了评价,并结合障碍诊断模型,分析了该研究区域河流健康突出问题。结果表明：2017—2019年渭河流域甘肃段干流河流健康等级均为亚健康状态,且呈现小幅好转趋势;5个准则层中化学完整性表现最差,流域治理应重点关注河流水质治理;评价河段的敏感度指标与水文、化学、生物完整性相关,是提高河流健康水平最有效的敏感因子。     

Coupling watershed modeling,public engagement,and soil analysis improves decision making for targeting P retention wetland locations

Constructed and restored wetlands can be effective sinks for particulate and dissolved phosphorus (P) if properly managed, but identifying suitable P retention wetland locations remains challenging. From a landscape perspective, Soil and Water Assessment Tool (SWAT) models identify locations within target watersheds with high nutrient loads that exhibit appropriate site characteristics and hydrodynamics. However, soil properties vary at the field scale, dictating the capacity of wetland systems to remove P and ultimately determining if a given wetland will operate as a sink or source of P over time. Land ownership and site access further complicate identification of P retention wetland locations. As a result, optimization and identification of P retention wetland locations requires analysis at both 1) watershed and 2) field scales, and 3) public engagement. In response, a survey effort linked SWAT model results that identified locations with target watersheds with field soil P storage capacity data and interested landowners. Results suggest that several locations recommended for their high SWAT-predicted P loading and landowner interest were in fact not well suited for project implementation due to soil P saturation and legacy P constraints. These findings highlight the need to couple watershed models with field scale soils analysis to identify locations for P retention wetlands in order to avoid unintended P release. Additionally, increased collaboration with social scientists and others familiar with public engagement strategies is needed to improve outreach activities targeting regional water quality improvements. Practical applications for nutrient retention wetland site selection are also discussed.     

1980～2015年长江流域土地利用变化分析

基于长江流域1980,1990,1995,2000,2005,2010,2015年7期土地利用数据,通过土地利用转移矩阵以及多种指标,分析了1980~2015年间长江流域土地利用的时空变化过程。研究表明：近35 a来耕地和草地呈下降趋势,而林地、水域和城乡用地面积明显增加,未利用地总体呈下降趋势;随着时间的推移,各地类的变化形式由以交换变化量为主导转变为以净变化量为主导,说明各地类逐渐倾向于在原有空间位置发生面积变化,而不是在流域内发生空间位置转移;长江流域土地利用综合动态度从大到小排列依次是1990~1995年、1980~1990年、2005~2010年、1995~2000年、2000~2005年、2010~2015年。1980~1990年和1990~1995年这两个时期的综合土地利用综合动态度明显大于其他各时期,即这两个时期长江流域土地利用变化程度最为剧烈。     

CMADS驱动SWAT模型在水循环模拟中的应用——以洱海流域为例

基于大气数据驱动水文模型的输出结果开展水循环模拟研究是大气和水文学界的研究热点。利用中国大气同化驱动数据集CMADS驱动SWAT模型,模拟2009～2016年期间洱海流域关键水循环要素的时空分布特征。结果表明:①CMADS数据集可很好地驱动SWAT模型,在洱海流域适用性较好,可用于水循环模拟研究。②从时间上看,洱海流域年降水量、实际蒸散和年产水量均呈现出先减少后增加的趋势,分别以4.6,29.3,15.0 mm/a的速率递增,多年平均降水量、多年平均实际蒸散发和多年平均产水量分别为792.8,565.5,286.1 mm。从空间上看,洱海西部降水最丰沛,东部地区次之,北部地区最低;实际蒸散发空间差异性相对较小,高值区主要分布于洱海湖区周围;产水量空间分布差异性较大,洱海西部产水量最大,其次为洱海北部和东部山区。③湿润度呈现出增加的趋势,而产水系数表现出减少的趋势,多年平均湿润度和产水系数分别为0.61和0.43;实际蒸散发与降水变化趋势一致,但与潜在蒸散发变化趋势相反,表明水分条件是限制洱海流域潜热的主要因子。     

内陆干旱区实施最严格水资源管理的关键技术研究——以新疆呼图壁河流域为典型示范区

以最严格水资源管理理念为前提,从最严格水资源管理下的区域水资源综合模拟、水资源管理分配,以及典型研究区流域水资源监测框架核心技术及考核指标的评价体系这4个方面,研究内陆干旱区实施最严格水资源管理的关键技术,且以天山北坡中段呼图壁河流域为典型示范区,揭示流域内水资源的供需规律,建立适应内陆干旱区的最严格水资源管理制度,为实现区域水资源的可持续利用以及支撑区域经济社会的可持续发展提供科学依据和决策建议,为干旱区贯彻和落实最严格水资源管理制度提供科技支撑。     

奎屯河流域健康评价

以新疆典型的干旱内陆河流域--奎屯河流域为研究对象,在确定流域健康评价时空尺度的基础上,提出涵盖流域水资源健康、涉水生态环境健康、社会经济健康的评价指标体系,构建基于组合赋权与协调发展度的流域健康评价模型。结果表明:现状年(2015年)处于亚健康状态的分区有:子区1、子区2、子区3、子区6,综合评价指数分别为:0.419、0.404、0.408、0.401,协调发展类型为临界协调发展;处于不健康状态的分区有:子区4和子区5,综合评价指数为0.358和0.343,协调发展类型属于中度不协调衰退型。     

博斯腾湖流域水资源管理决策支持系统设计与实现

为了提高博斯腾湖流域水资源管理水平,设计开发了针对该流域的水资源管理决策支持系统。系统基于Microsoft.NET平台,采用Visual Studio、SQL Server、ArcGIS工具集开发,在流域“地-云-空”一体化监测体系和系统架构的基础上,构建了地图要素管理、基本水文信息、流域生态流量等主要模块,实现了流域空间可视化查询、实时流量动态模拟、流域生态基流计算以及水资源管理配置等功能。结合博斯腾湖流域1956—2019年水文资料及目前数据获取情况,系统可为不同发展模式下水资源管理提供多方面参考信息和技术支撑,能有效提高水资源管理部门决策效率。     

Restoring Murray River floodplain wetlands: Does the sediment record inform on watering regime?

The floodplain wetlands of the southern Murray Darling Basin (MDB) have been subject to the impacts of catchment and water resource development for more than a century. Their current degraded state is attributed to the regulation of the rivers and abstraction of water volume for irrigation. The MDB Plan is to return at least 2,750 Gl of mean annual flow to the system to restore the condition of waterways. Considerable recent investment in infrastructure enables water to be released into the basin's floodplain wetlands. The proposed watering regime is underpinned by modelling that suggests that, before regulation, overbank flows would have occurred regularly as discharge peaked in winter and spring. Sediment cores have been extracted from over 50 floodplain wetlands of the southern Murray Basin. Those from several, large meander wavelength billabongs extend for 1,000–5,000 years suggesting that these sites were permanently inundated over that time. Others extend to ~200 years and are presumed not to have accumulated sediment until more recently. The records of most wetlands, however, only extend to the onset of river regulation in the 1920s, suggesting that before then they were not inundated for sufficient duration for net accumulation to occur. Preserved diatoms suggest that the shallow, plant‐dominated wetlands of the past have transitioned to deep, turbid water systems today. As rivers are identified as a source of sediment to wetlands, less regular inundation, rather than more, is a viable option in restoring the ecological function of these floodplain wetlands and in slowing sediment infill.     

雅鲁藏布江流域枯水期化学风化及 CO2 消耗

基于雅鲁藏布江流域枯水期干支流地表径流、泉水、地下水以及降水样品,对主要离子、pH、TDS和CO_2消耗进行分析,研究枯水期水化学特征及成因,并分析流域化学风化速率及CO_2消耗通量。结果表明:流域水体总体呈现弱碱性,地表径流TDS均值约为世界河流的2.3倍,和长江流域相似,但远低于黄河流域。岩石风化是地表径流离子来源的主要影响因素。碳酸盐风化、硅酸盐风化和蒸发盐风化分别占离子来源的46.3%、28.6%和21.3%,碳酸盐风化速率和硅酸盐风化速率分别为1.43t/(km~2·a)和1.11t/(km~2·a),对应的CO_2消耗速率为0.80×10~5mol/(km~2·a)和0.55×10~5 mol/(km~2·a),枯水期化学风化速率和CO_2消耗速率较小。