Incorporating limited field operability and legacy soil samples in a hypercube sampling design for digital soil mapping

Abstract

Most calibration sampling designs for Digital Soil Mapping (DSM) demarcate spatially distinct sample sites. In practical applications major challenges are often limited field accessibility and the question on how to integrate legacy soil samples to cope with usually scarce resources for field sampling and laboratory analysis. The study focuses on the development and application of an efficiency improved DSM sampling design that (1) applies an optimized sample set size, (2) compensates for limited field accessibility, and (3) enables the integration of legacy soil samples. The proposed sampling design represents a modification of conditioned Latin Hypercube Sampling (cLHS), which originally returns distinct sample sites to optimally cover a soil related covariate space and to preserve the correlation of the covariates in the sample set. The sample set size was determined by comparing multiple sample set sizes of original cLHS sets according to their representation of the covariate space. Limited field accessibility and the integration of legacy samples were incorporated by providing alternative sample sites to replace the original cLHS sites. We applied the modified cLHS design (cLHS adapt ) in a small catchment (4.2 km 2 ) in Central China to model topsoil sand fractions using Random Forest regression (RF). For evaluating the proposed approach, we compared cLHS adapt with the original cLHS design (cLHS orig ). With an optimized sample set size n = 30, the results show a similar representation of the cLHS covariate space between cLHS adapt and cLHS orig , while the correlation between the covariates is preserved (r = 0.40 vs. r = 0.39). Furthermore, we doubled the sample set size of cLHS adapt by adding available legacy samples (cLHS adapt+ ) and compared the prediction accuracies. Based on an external validation set cLHS val (n = 20), the coefficient of determination (R 2 ) of the cLHS adapt predictions range between 0.59 and 0.71 for topsoil sand fractions. The R 2 -values of the RF predictions based on cLHS adapt+ , using additional legacy samples, are marginally increased on average by 5%.

High Resolution Mapping of Soil Properties Using Remote Sensing Variables in South-Western Burkina Faso: A Comparison of Machine Learning and Multiple Linear Regression Models

“…Improving prediction accuracy might require partitioning the landscape into relatively homogeneous areas based on the covariates used as predictors. The latter could be carried out by considering land surface segmentation [100] or the conditioned Latin hypercube (cLHS) [101] using the key variables driving each soil property. Moreover, though the present study considered a vast array of predictors including topographical and spectral data, a single analysis scale was applied as generally carried out in DSM.…”

Mapping soil properties in a poorly-accessible area

“…The approach used in this study to select the sampling points followed the principles proposed by Minasny and McBratney (2006). The points were selected by using cLHS with auxiliary covariates, and considering the access costs (Roudier et al, 2012;Carvalho Júnior et al, 2014;Stumpf et al, 2016). As a constraint, three buffer sizes were created in relation to roads and trails, as proposed by Carvalho Júnior et al (2014).…”

“…To facilitate DSM in poorly-accessible areas, Cambule et al (2013) proposed a methodology of sampling in the area of greater accessibility, which is representative of the total area, and to evaluate the representativeness using, e.g., the similarity between the histogram of the covariates for the total and accessible areas. Other studies considered the costs of accessibility in soil sampling (Roudier et al, 2012;Carvalho Júnior et al, 2014;Stumpf et al, 2016) using a variation/optimization of the method known as conditioned Latin Hypercube Sampling (cLHS), proposed by Minasny and McBratney (2006). The cLHS is a robust tool for the allocation of sampling points by means of a set of auxiliary covariates.…”

“…The Itatiaia National Park (INP) (Brazil) has limited access due to the steep slopes, dense forest cover, rocky outcrops, and altitude vegetation fields in the upper plateau (Barreto et al, 2013), all that making the INP an excellent case study. To obtain a viable result with a low cost, it is important to use DSM tools, ranging from optimization of the sampling site (Minasny and McBratney, 2006;Roudier et al, 2012;Stumpf et al, 2016) to the covariate selection using powerful predictive algorithms (Beguin et al, 2017;Jeong et al, 2017). Based on the above and considering the advancements and challenges of DSM, it is necessary to optimize financial and human resources to produce quality information that can be useful for decision-makers.…”

Rainfall monitoring network design using conditioned Latin hypercube sampling and satellite precipitation estimates: An application in the ungauged Ecuadorian Amazon

“…Conditioned Latin hypercube sampling (cLHS) provides an approach for incorporating prior auxiliary information from remote sensing instruments as well as accessibility restrictions in a sample design. cLHS is a multivariate stratified random strategy (Minasny and McBratney, ) that has been proven to be an efficient sampling method because it captures the marginal variability of several variables using a relatively small sample (Brungard and Boettinger, ; Ramirez‐Lopez et al ., ; Stumpf et al ., ; Domenech et al ., ). Roudier et al .…”