Optimal interpolation methods for farmland soil organic matter in various landforms of a complex topography. (March 2020)

• Record Type: Journal Article
• Authors: Long, Jun; Liu, Yaling; Xing, Shihe; Zhang, Liming; Qu, Mingkai; Qiu, Longxia; Huang, Qian; Zhou, Biqing; Shen, Jinquan
• Is Part Of: Ecological indicators. Volume 110(2020)
• Publication Date: 2020-03
• Subjects: Spatial interpolation -- Prediction accuracy -- Complex topography -- Landforms -- Farmland soil organic matter; Environmental monitoring -- Periodicals; Environmental management -- Periodicals; Environmental impact analysis -- Periodicals; Environmental risk assessment -- Periodicals; Sustainable development -- Periodicals; 333.71405
• DOI: 10.1016/j.ecolind.2019.105926
• Languages: English
• ISSNs: 1470-160X

Abstract

Highlights: Samples from 188, 247 sampling sites were interpolated across various landforms. SK performed the best in SOM interpolation over the entire region of complex topography. SK was the best for hill-mountain and valley-basin, but it was IDW for plain-platform. Geo-statistical models were more desirable than deterministic interpolations. Environmental variables were effective to improve the interpolation accuracy. Abstract: Interpolation plays an important role in revealing the spatial variations of soil organic matter (SOM). However, selecting an optimal one from multiple interpolations is challenging, particularly in the complex topography which is characterized by slope-rich terrain. In this work, the study area is a large region of slope-rich and complex topography, encompassing 9 cities and including 3 landforms of plain-platform, valley-basin, and hill-mountain. Based on 188, 247 sampling sites, 12 widely used deterministic and geo-statistical interpolations were performed for farmland SOM estimation. Moreover, ancillary variables such as terrain, climate and soil cover patterns were combined with the optimal models of the above interpolations to examine the effects of incorporation of environmental variables on interpolation accuracy. The results showed that interpolation accuracy varied from landform to landform. The prediction accuracy of geo-statistical interpolations was generally higher than that of deterministic methods, and ancillary variables were generally effective to improve the interpolation accuracy. Specifically, for the entire region, simple kriging (SK) was the most accurate of these 12 interpolations, with a root-mean-square prediction error (RMSE) of 6.75 g kg −1 and a correlation coefficient (r) of 0.79 between the predicted and measured values ( p < 0.01). Furthermore, when SK was combined with terrain, climate and soil cover patterns for interpolation (SK_T, SK_C, and SK_S), they performed better than SK. In terms of various landforms, such as hill-mountain and valley-basin, SK was still the best performing model, with RMSEs of 6.91 g kg −1 and 7.03 g kg −1, respectively. Further comparisons showed that SK_T, SK_C, and SK_S performed better than SK in hill-mountain and valley-basin. In plain-platform, the use of inverse distance weighting (IDW) was more desirable, with RMSE of 5.96 g kg −1, and RMSEs of methods where IDW was combined with terrain, climate and soil cover patterns (IDW_T, IDW_C, and IDW_S) followed the order of IDW_S > IDW > IDW_T > IDW_C. Our findings could inform interpolation of many environmental variables, not tied to SOM, in similar areas characterized by various landforms of a complex topography.