Department of Tropical Plant and Soil Sciences, University of Hawai'i at Mānoa, 3190 Maile Way, St. John Plant Science Lab, Room 102, Honolulu, HI 96822, United States of America.
Department of Educational Administration, University of Hawai'i at Mānoa, 1776 University Avenue, Honolulu, HI 96822, United States of America.
Sci Total Environ. 2023 Apr 20;870:161900. doi: 10.1016/j.scitotenv.2023.161900. Epub 2023 Jan 30.
Soil health is recognized as an important ecosystem property sensitive to human impact. As a concept, soil health cannot be directly measured, and so assessment and modeling efforts largely rely upon key biological, chemical, and physical indicators. Efforts to develop an overall soil health index are largely lacking due to significant statistical challenges and the necessity for regional calibration. Taken from the field of educational research, structural equation modeling (SEM) is an attractive approach to enhance the reliability and validity of soil health scoring. Therefore, SEM may be utilized to advance research efforts to understand management practices impacts on soil health. Our objectives were to develop a robust scoring function that (i) captures the concept of soil health and latent variables, (ii) adjusts scores by inherent soil properties and legacy of intensive land use to adequately reflect our regional conditions and contemporary land management, and (iii) meets the diverse practitioner needs. Through this process, we refined our minimum dataset of soil health indicators and reconceptualized soil health indicators into functional properties. Our results support the development of a robust single level or a multilevel SEM model-depending on the practitioner's goals-that accounts for repeated sampling or pseudoreplication. While the SEM scoring functions were highly related to the conventional scoring approach, SEM outperformed the conventional methods in terms of its wider distribution of scores-and thus enhanced discriminatory power on the lower and higher range of scores. We also confirmed that the SEM scoring function that includes adjustments for mineralogy and legacy of intensive land use successfully differentiates among contemporary management practices and land use. Therefore, we have confidence that the tool is reliable and appropriate to further examine more nuanced impacts of land use change and management practices within a given land use across time and space covering a diversity of soils. (300 words).
土壤健康被认为是对人类影响敏感的重要生态系统特性。作为一个概念，土壤健康无法直接测量，因此评估和建模工作主要依赖于关键的生物、化学和物理指标。由于存在重大的统计挑战和区域校准的必要性，开发综合土壤健康指数的努力在很大程度上仍付诸阙如。结构方程模型（SEM）源自教育研究领域，是提高土壤健康评分可靠性和有效性的一种有吸引力的方法。因此，SEM 可用于推进研究工作，以了解管理实践对土壤健康的影响。我们的目标是开发一种稳健的评分函数，（i）捕捉土壤健康和潜在变量的概念，（ii）通过固有土壤特性和集约化土地利用的遗留物调整分数，以充分反映我们的区域条件和当代土地管理，以及（iii）满足不同从业者的需求。通过这个过程，我们改进了我们的土壤健康指标最小数据集，并将土壤健康指标重新概念化为功能特性。我们的结果支持开发稳健的单级或多级 SEM 模型，具体取决于从业者的目标，该模型考虑了重复采样或伪重复。虽然 SEM 评分函数与传统评分方法高度相关，但 SEM 在分数分布更广泛方面优于传统方法，从而在分数的较低和较高范围内提高了区分能力。我们还证实，包含对矿物学和集约化土地利用遗留物进行调整的 SEM 评分函数能够成功区分当代管理实践和土地利用。因此，我们有信心该工具是可靠的，适合进一步研究在给定土地利用中，随着时间和空间的推移，土地利用变化和管理实践更细微的影响，涵盖了多种土壤。（300 字）
