基于植被和环境因子的亚高山森林土壤水源涵养功能空间尺度上推模型构建——以岷江上游杂谷脑流域为例

土壤层水源涵养功能是森林水源涵养功能的主体。目前关于森林土壤水源涵养功能的研究主要集中在林地或坡面尺度上。由于流域尺度，尤其是环境空间异质性强的西南亚高山区流域，如何将林地尺度实测结果上推至流域或更大空间尺度仍是生态水文领域面临的巨大挑战之一。以川西岷江上游杂谷脑流域为研究对象，融合多种森林类型样地实测与流域尺度多源遥感数据，构建了基于植被和环境因子的林地-流域森林土壤水源涵养功能尺度转换模型，实现了流域尺度土壤水源涵养功能快速评价及其空间分布预测。样地尺度研究结果表明各类型森林的土壤水文特性各异，总体表现为天然林优于人工林，混交林优于单纯林。林地土壤持水能力受到区域气候、植被、土壤及地形等因子的共同影响，其中风速、NDVI及林龄与土壤最大持水量、毛管持水量及非毛管持水量均呈极显著正相关（P<0.01）。基于关键植被和环境因子构建的林地-流域土壤水源涵养功能尺度上推模型精度较高，土壤最大持水量、土壤毛管持水量和土壤非毛管持水量模型拟合优度R²分别为0.700、0.720和0.908；土壤最大持水量、土壤毛管持水量和土壤非毛管持水量的模型预测值与野外实测值的相关系数介于0.69-0.79之间，平均误差均低于20%，表明模型预测结果可靠。利用构建的土壤水源涵养功能尺度上推模型，估算得出流域尺度森林土壤持水量的空间分布，其结果表明杂谷脑流域森林土壤持水量空间分异明显，海拔较高区域森林土壤持水量最高，其次为距道路和河流有一定距离的缓坡地带，下游干旱河谷地区土壤持水量最低。本研究为亚高山森林生态功能的恢复和提升提供了科学依据和评价工具。

Upscaling subalpine forest soil water-holding capacity based on vegetation and environmental factors: an example of the Zagunao River watershed in the upper reach of the Minjiang River in China

Soil water-holding capacity is the major part of forest water storage. Although extensive studies on forest water-holding capacity at plot or slope scales have been conducted, there remains a lack of effective approaches to extrapolate the forest water-holding capacity from plot to watershed scale due to the spatial heterogeneity of environmental factors especially in subalpine watersheds. Here, we used the Zagunao watershed in the western Sichuan as an example to construct the multivariable linear regression models for upscaling soil water-holding capacity from plot to watershed scales based on vegetation and environmental factors by integrating field experiments at a variety of forest types and remote sensing data at a watershed scale, which eventually enabled a rapid evaluation and spatial prediction of soil water-holding capacity at watershed scale. The forest plot level result showed that the water-holding capacity of natural forest was better than that of the artificial forest with the mixed forests better than monoculture. The correlation analysis suggested that soil water-holding capacity was jointly affected by climate, vegetation, soil, and topography such as wind speed, Normalized Difference Vegetation Index (NDVI), and forest age. The upscaling models of the maximum soil water-holding capacity, capillary water-holding capacity and noncapillary water-holding capacity established based on vegetation and environment factors had good performance with the R² of 0.700, 0.720, and 0.908, respectively. The correlation coefficients and mean relative errors based on observations and predictions of the maximum soil water-holding capacity, capillary water-holding capacity and noncapillary water-holding capacity models were between 0.69 and 0.79, and below 20%, respectively, which indicated the good reliability of the models. By use of the upscaling models, the spatial distributions of forest soil water-holding capacity at the watershed scale were estimated. As suggested by the model prediction, the spatial variations of forest soil water-holding capacity are distinct in the Zagunao watershed, where forests at higher elevations are featured with the highest soil water-holding capacity, followed by forests located at a certain distance from roads and rivers in the lower slopes, and the downstream arid valley area has the lowest soil water-holding capacity. The findings can provide effectively scientific supports and assessment tools for the improvement and restoration of ecological functions in subalpine forests.