农田-防护林-荒漠复合系统土壤水盐运移规律及耦合模型建立

为探讨节水灌溉条件下干旱内陆区不同景观单元土壤水盐动态规律及水盐通量变化特征,以新疆三工河流域绿洲-荒漠过渡带典型景观格局农田-防护林-荒漠为研究对象,利用2018年4月—9月连续定位观测数据资料,分析各景观单元作物生育期(4月1日—6月28日)和非生育期(6月29日—9月15日)土壤水盐动态规律及其变异性、土壤水盐通量变化特征及影响因素,构建农田-防护林-荒漠复合系统BP神经网络土壤水盐耦合模型,并对所建模型参数敏感性及应用可行性进行探讨。结果表明,各景观单元作物生育期和非生育期土壤含水率、电导率均具有较明显的垂直分层、水平递变和季节波动特征;按变异性可划分为3个典型土层:活跃层(0～40 cm)、次活跃层(40～140 cm)和相对稳定层(140cm);距防护林越近,农田土壤含水率和电导率分别呈降低和升高趋势,荒漠均呈升高趋势;单次降水和灌溉事件后各景观单元各典型土层土壤含水率和电导率随时间分别均呈负指数函数和三次函数变化趋势。土壤控制体(单位面积深140 cm土柱)内,生育期农田和防护林均为向下水分通量,非生育期均为向上水分通量,荒漠两时期均为向下水分通量;农田和防护林土壤贮水量与土壤积盐量随地下水位下降、蒸散发量增大均呈递减趋势;荒漠土壤水盐通量对各因素及其交互效应响应较微弱;生育期最后1次充分灌溉的淋洗作用可使该系统土壤积盐量趋于平衡状态。拓扑结构为32-36-6的BP神经网络土壤水盐耦合模型具有较高的模拟精度;灌溉和地下水位是影响该系统土壤水盐动态的关键因素。研究结果可为节水灌溉条件下绿洲-荒漠共生系统寻求生产和生态之间的平衡机制提供理论依据。

Soil water and salt dynamics and its coupling model at cropland-treebelt-desert compound system

Abstract: Water shortage and soil salinization are two of the most concerned problems in the arid inland river basins of Northwest China with very low precipitation and extremely high evaporative demand. Water-saving irrigation has been popularized within the arid regions of China to enhance the efficient use of water and fertilizer and improve economic efficiency. However, this has triggered a series of evident/potential eco-environment issues due to regional water and salt imbalance, including land deterioration, groundwater level descending, and natural vegetation degradation, etc. Thus, identifying the spatial-temporal dynamic, variability, and distribution characteristics of soil water and salt in arid ecosystem is of great significance to regional water resources management and vegetation conservation. This study aimed to investigate the dynamics of soil water and salt and the variation of water and salt flux within and between different landscape units in arid areas under water-saving irrigation conditions. To achieve the objective, the soil water content (SWC), soil electrical conductivity (SEC), groundwater and vegetation dynamics were continuously monitored along a cropland-treebelt-desert compound system at the oasis-desert ecotone in the Sangong River Basin, Northwest China, from April to September 2018. A coupling model of soil water and salt in the cropland-treebelt-desert compound system based on BP neural network algorithm (BPNNA) was proposed, and the sensitivity of parameters of the coupling model and its application feasibility were discussed. The results showed that: 1) SWC and SEC of each landscape unit had obvious vertical stratification, horizontal progressive and seasonal fluctuation characteristics during the growth period (from April 1 to June 28) and non-growth period (from June 29 to September 15). Based on the coefficients of variation of soil water and salt, a 0-220/300 cm soil profile could be divided into three typical layers: active layer (0-40 cm), sub-active layer (>40-140 cm) and relatively steady layer (>140 cm). With increasing proximity to the treebelt, the SWC and SEC of the cropland showed a decreasing and increasing trend respectively, while the desert showed an increasing trend; the time-varying process in SWC and SEC of each typical soil layer of each landscape unit after precipitation and irrigation event could be well described by exponential decay function and three quadratic function, respectively. 2) Within the active or sub-active layer (140 cm in depth), a downward water flux was observed at the cropland and treebelt during the growth period, and a upward water flux was observed during the non-growth period, while at the desert a downward water flux was observed at the two periods. Soil water storage and accumulated soil salt of cropland and treebelt showed a decrease trend with the decrease of water table depth and/or the increase of evapotranspiration. Soil water and salt flux of desert had weak response to various influencing factors and their interaction effects, the leaching effect of the last adequate irrigation at the growth period could make the accumulated soil salt among cropland-treebelt-desert tend to be in a relatively balanced state. 3) The coupling model of soil water-salt using BPNNA with the 32-36-6 structure had a high simulation accuracy and applicability in the cropland-treebelt-desert compound system. Irrigation and water table depth were the main factors affecting soil water and salt dynamics in this system. The results could provide an insight for seeking a balance mechanism between production demand and ecological protection in oasis-desert systems under water-saving irrigation conditions.