干旱区三种典型地貌下电磁感应式土壤盐分协同解译模型

为明确不同地貌类型下土壤理化性质对电磁感应式表观电导率测量精度的影响,该研究以新疆玛纳斯河流域3种典型地貌类型(冲积洪积扇缘、冲积平原、干三角洲)为研究对象,运用电磁感应仪EM38结合土壤采样室内测定方法,分析土壤剖面(100 cm)每20 cm土层的土壤性质对不同高度(130、110、90、70、50 cm)所测表观电导率的影响程度和贡献率,通过引入对表观电导率贡献率较高的非盐分因子作为辅助变量,利用多元线性回归方法,建立土壤盐分多因素协同解译模型。结果表明:3种地貌类型中,土壤盐分含量是影响表观土壤电导率贡献率最大的作用因子,不同地貌类型下各土层影响表观电导率的因子存在明显的差异,冲积洪积扇缘地貌主要表现为上层(0～60cm)土壤含水率和底层(60～100)土壤阳离子交换量和有机碳含量对表观电导率贡献较高,冲积平原地貌则是表层(0～20 cm)和底层(40～100cm)的土壤含水率以及20～40cm土层的黏粒含量和阳离子交换量对表观电导率表现出较高的贡献率,干三角洲地貌下上层土壤(0～60cm)阳离子交换量和下层(60～100cm)土壤有机碳含量能够对表观电导率产生较为明显的影响。通过引入对表观电导率影响较大的作用因子,建立了针对不同地貌类型下分层土壤盐分协同解译模型,与仅以表观电导率解译土壤盐分含量相比,冲积洪积扇缘、冲积平原、干三角洲地貌类型下0～100cm土层盐分预测模型校正决定系数分别由0.81～0.86、0.55～0.87、0.25～0.56提高至0.83～0.91、0.63～0.93、0.48～0.70,多因素协同解译模型有效提高了土壤盐分解译模型精度。研究结果可为盐渍化土壤的快速准确监测提供可靠的理论依据和技术方法。

Synergistic interpretation model for soil salinity by electromagnetic induction under three typical landforms in arid areas

Abstract: Accurate and rapid assessment and measurement of soil salt accumulation and spatial distribution changes are essential for preventing land degradation and improving the ecological environment. The soil Apparent Conductivity (ECa) obtained by electromagnetic induction technology can be more effective and faster to obtain soil salinity information, which helps to overcome some challenges in traditional sampling methods and reduce costs. However, the differences in soil properties among different geomorphologic types may lead to the decrease in the accuracy of EM38 in predicting soil salinity. In order to clarify the effect of soil properties on apparent conductivity under different geomorphologic types, three typical landforms (alluvial-proluvial fan edge, alluvial plain and dry delta) in Manas River Basin of Xinjiang were taken as the research objects. The apparent conductivity data of two measurement modes (horizontal mode EMh and vertical mode EMv) were obtained by using EM38 at different heights from the ground, i.e., 30 cm, 50 cm, 70 cm, 90 cm, 110 cm and 130 cm, respectively. Moreover, in each landform, 20 representative sampling points were selected for soil sample collection, with sampling depth of 0-20, 20-40, 40-60, 60-80, 80-100 cm. Soil salt content, soil moisture content, soil clay mass fraction, soil Cation Exchange Capacity (CEC) and soil organic carbon content (SOC) were determined. Firstly, path analysis method was used to analyze the influence degree and contribution rate of salt content, soil clay mass fraction, soil CEC and SOC at different depths on apparent conductivity (ECa) measured at different heights. Next, by selecting non-salinity factors with high contribution rate of ECa as auxiliary variables, a multi-factor collaborative interpretation model of soil salinity was established, and compared with the model established only with ECa as independent variable. Finally, the optimal interpretation model of soil salt content was established and the accuracy of the model was evaluated. The results showed that among the three types of landforms, soil salt content was the most important factor affecting the contribution rate of ECa, and there were significant differences in the factors affecting ECa of each soil layer under different landforms. Water content of the upper soil (0-60 cm) contributed most to ECa, whereas soil CEC content and organic carbon content of the bottom soil (60-100 cm) in the alluvial-proluvial fan edge. A high contribution was made by soil salt content at 0-20 cm layer and 40-80 cm layer, whereas CEC and clay mass fraction for 20-40 cm layer in the alluvial plain. And ECa can be significantly affected by CEC content of the upper layer (0-60 cm) and soil organic carbon content of the lower layer (60-100 cm) in the dry delta. According to the accumulative contribution rate of more than 80%, non-salt factors were selected to establish the synergistic interpretation model of layered soil salt under different landforms. The R2adj of different soil layers under three types of landforms increased from 0.81-0.86, 0.57-0.87 and 0.27-0.47 to 0.83-0.91, 0.63-0.93 and 0.48-0.70, respectively. The accuracy verification results showed that the R2 of salt prediction models for different soil layers under three types of landforms were 0.61-0.81, 0.48-0.85 and 0.35-0.66, respectively. The research results can provide reliable theoretical basis and technical methods for rapid and accurate monitoring of saline soil.