多源土壤湿度产品在黄土高原地区适用性分析与评价

Analysis and Evaluation of Applicability of Multi-source Soil Moisture Products on Loess Plateau

摘要:

高精度的土壤湿度数据对于区域水资源管理和农业生产具有重要意义。在利用现有土壤湿度产品前，需对不同土壤湿度产品进行全面精度适用性分析与评价，以选择目标研究区内综合精度表现最高产品作为后续研究的基础数据。本研究利用黄土高原地区不同站点土壤湿度实测数据，综合分析与评估ERA5、GLDAS、SMCI和SoMo 4种土壤湿度产品在黄土高原的精度，包括从空间角度评估不同地区精度、从定量角度评估年尺度和干湿季尺度精度。此外，还评估了不同土壤湿度产品在不同土地覆盖类型下综合精度表现，最终通过组合多种土壤湿度产品来提高黄土高原土壤湿度产品整体精度。结果表明：ERA5、GLDAS和SoMo 3种土壤湿度产品高估了黄土高原土壤湿度，约有85%站点偏差（Bias）在0~0.3 m^3/m^3之间，而GLDAS产品在各站点的多年平均土壤湿度与站点实测值较为一致，超过96.5%站点Bias在-0.15~0.15 m^3/m^3之间，中位数为0 m^3/m^3。年尺度、干季、湿季评估结果显示，GLDAS产品整体精度表现较好，ERA5产品整体精度表现较差，SMCI与SoMo产品精度表现无明显差异。GLDAS产品在草地、森林、灌溉耕地等土地覆盖类型整体精度表现较好，而SoMo产品在非灌溉耕地整体精度表现较好。通过集成GLDAS与SoMo 2种产品得到黄土高原土壤湿度产品组合数据集，能在保持土壤湿度较小误差情况下，增强其对黄土高原土壤湿度动态变化的反映能力。研究所得黄土高原土壤湿度产品组合数据集与站点实测值间的相关系数（R）、均方根误差（RMSE）、偏差（Bias）和无偏均方根误差（ubRMSE）中位数分别为0.66、0.06 m^3/m^3、0.02 m^3/m^3、0.04 m^3/m^3。总体而言，GLDAS产品可作为黄土高原的优选土壤湿度产品，用于该地区生态、农业、水文等领域研究。

Abstract:

Accurate soil moisture data are essential for regional water resource management and agricultural production. Thus, a comprehensive accuracy analysis and assessment is necessary to identify the soil moisture product with the optimal overall performance in the target region before applying the existing products, which can then serve as the foundational dataset for subsequent research. The accuracy of four soil moisture products (ERA5, GLDAS, SMCI, and SoMo) was evaluated based on the soil moisture data obtained from different observation sites across Loess Plateau of China. The evaluation included spatial accuracy assessment across different regions, quantitative performance estimation at annual and seasonal (dry/wet) scales, and accuracy appraisal under various land cover types. Based on the evaluation results above, a fusion of multiple products was conducted to enhance the overall accuracy of soil moisture estimation on Loess Plateau. The results showed that ERA5, GLDAS, and SoMo products tended to overestimate the actual soil moisture on the Loess Plateau, with Bias values between 0 m^3/m^3 and 0.3 m^3/m^3 at about 85% of the total sites investigated. The GLDAS product had the highest consistency with the observed multi-year average soil moisture, with Bias values of -0.15 m^3/m^3~0.15 m^3/m^3 at more than 96.5% of the total sites and with a median Bias of 0 m^3/m^3. At annual scale and during both dry and wet seasons, the GLDAS product showed the best overall accuracy, while ERA5 performed the worst. The SMCI and SoMo products had comparable performances. The GLDAS product performed better under the land cover types of grassland, forest, and irrigated cropland, while SoMo product performed best under non-irrigated cropland. Through combining the GLDAS and SoMo products, an integrated optimal soil moisture product was obtained for Loess Plateau, which could enhance the dataset's ability to reflect the dynamic changes in soil moisture on Loess Plateau while maintained a small error in values of soil moisture. The median values of the correlation coefficient (R), root mean square error (RMSE), bias (Bias), and unbiased root mean square error (ubRMSE) between the integrated soil moisture product and site measurements on Loess Plateau were 0.66, 0.06 m^3/m^3, 0.02 m^3/m^3, and 0.04 m^3/m^3, respectively. In general, the GLDAS product was recommended as the optimal soil moisture product for ecological, agricultural, and hydrological studies on Loess Plateau.