Evaluation of Integrative Hierarchical Stepwise Sampling for Digital Soil Mapping

Abstract

This paper presents an integrative hierarchical stepwise sampling (IHS) method and two case studies to compare it with stratified random sampling (SRS) and conditioned Latin hypercube sampling (cLHS). The first comparison between IHS and SRS was conducted for mapping sand content of two soil layers in a study area in Anhui Province, China. Two sample sets of the same sample size were collected in the field based on IHS and SRS. The second case study is a simulation study, where we compared IHS and cLHS for mapping soil series in the Raffelson watershed in Wisconsin (USA). The study used an accurate and detailed soil series map produced previously as a proxy of the real soil distribution. Virtual samples with nine sample sizes designed by IHS and cLHS were collected on the soil map. For both case studies, an individual predictive soil mapping method was employed and independent validation samples were used to evaluate the mapping accuracies. Results indicate that IHS generally performs better than SRS for capturing distributions of the environmental variables. It obtained higher mapping accuracies than SRS at different sample sizes. On the other hand, cLHS appears to provide a better representation for distributions of the environmental variables than IHS, but the mapping accuracies with IHS are higher than those with cLHS at nearly all sample sizes. Finally, both case studies showed that IHS provides valuable information on representativeness of the samples.Abbreviations: FCM, fuzzy c-means; IHS, integrative hierarchical stepwise sampling; cLHS, conditioned Latin hypercube sampling; SRS, stratified random sampling. S oil sampling plays an essential role in soil survey. A sampling design specifies where and how many to sample in the field. Different sampling design strategies will produce different soil distribution maps and greatly affect the mapping accuracies (Brus and de Gruijter, 1997;van Groenigen et al., 2000;de Gruijter et al., 2006;Brus et al., 2006;Gregoire and Valentine, 2007;Heim et al., 2009). Because the collection of soil samples in the field is often time-consuming and resource-intensive, it is important to apply efficient sampling design strategies.When prior knowledge on an area's local soil variation is limited, simple random sampling has been a commonly used sampling strategy (Oliver and Webster, 1986;Grinand et al., 2008;Walvoort et al., 2010 • Evaluation of an integrative hierarchical stepwise (IHS) sampling method by comparing it with stratified random sampling (SRS) and conditioned Latin hypercube sampling (cLHS) through two case studies.• IHS obtained higher mapping accuracies than SRS and cLHS at nearly all sample sizes.• IHS provides valuable information on the representativeness of samples.• SRS and cLHS were found to generate unstable results on sample sets and soil maps.

Soil Property Variation within Taxa of ST at University of Nigeria, Nsukka

“…Result got can as well serve as indicating mapping accuracy since profiles were sited at locations most typical of soil class. Yang et al (2016) noted that one effective way to capture the greatest soil variability is by sampling at those locations where the soils are most typical of the class. The results of this study will be standard for comparisons of subsequent characterization and classification of the soils.…”

PyCLiPSM: Harnessing heterogeneous computing resources on CPUs and GPUs for accelerated digital soil mapping

“…iPSM (Zhu et al., 2015) is an algorithm specially designed for DSM and has been used in a wide range of DSM studies (An et al., 2018; Yang, Qi, Zhu, Shi, & An, 2016; Zeng et al., 2016; Zhang & Zhu, 2019; S.‐J. Zhang, Zhu, et al., 2016).…”

“…Readers interested in this topic are referred to Yang et al. (2016), Zeng et al. (2016), Zhang, Huang, Zhu, and Keel (2016), Zhu et al.…”

Peer Review #1 of "An improved similarity-based approach to predicting and mapping soil organic carbon and soil total nitrogen in a coastal region of northeastern China (v0.1)"

“…Although soil sampling in this study considered climate and vegetation properties of the landscape for clustering, it was further verified by the local soil experts to determine whether the units are typical and are representative of the study area. This data sampling was not based on probability sampling strategy, thus it might underestimate or overestimate the SOC and STN distribution in the region as reported in previous studies (Zhu, 1997;Yang L et al, 2016). However, it was probably the best sampling option given that sampling was done in such a densely forested terrain and with limited resources.…”