Mapping gamma radiation and its uncertainty from weathering products in a Tasmanian landscape with a proximal sensor and random forest kriging

Abstract

The radionuclides of potassium (40 K), uranium (238U) and thorium (232Th) emit from the land surface gamma radiation that is characteristic of the underlying rocks and the distribution of their weathering products in the landscape. We measured the radiation along widely separated transects using a mobile proximal sensor over a 10 000‐ha region of Tasmania. We supplemented the transect data with information from soil and geological maps and dense data from LandSat and SPOT imagery, a digital elevation model and terrain attributes on a grid at 30‐m intervals so as to map the radionuclides. We used a sequence of steps, starting with a spatial bootstrap and random forests to predict emissions across the study area and at sampling points excluded from the bootstrap samples. The predictions at the sampling points were compared with the observed values to obtain residuals, which were then used to krige them at all points on the 30 m grid. We combined the random forest and kriging predictions on the 30‐m grid to obtain our random forest kriging predictions. Repeating the procedure 100 times provided confidence limits on our results and predictions. The resulting maps of the radionuclides accord well with what we know of the soil, lithology and topography of the region from other sources. Alluvial deposits with large amounts of potassium extend from the foot slopes of the Great Western Tiers and along the flood plains of the Meander River, providing evidence of previous widespread weathering and deposition of the material. The fertile Red Ferrosols (roughly equivalent to Ferralic Nitisols in the World reference base (WRB) classification) on the extensive Tertiary basalt plateau emit little gamma radiation as a result of deep weathering and of potassium movement and accumulation down its talus slopes. The maps show the complexity of the region in terms of soil, lithology and terrain, and they show the merits of gamma radiometry for mapping and understanding the distribution of materials in such regions. Copyright © 2013 John Wiley & Sons, Ltd.

Local-scale spatial modelling for interpolating climatic temperature variables to predict agricultural plant suitability

“…The algorithm as summarized by Viscarra-Rossel et al (2013) functions by drawing Tv, v=1, 2,…, V new bootstrap samples from the data that become 'trees in the forest'. This consists of both 'in-the-bag' (training) and 'out-of-bag' (validation) samples.…”

Mapping Spatial Variations of Structure and Function Parameters for Forest Condition Assessment of the Changbai Mountain National Nature Reserve

“…Random forest kriging (RFK) was the extension of RF, which integrated RF prediction values and estimation of the residuals by ordinary kriging (OK) using Equation (3) [101]. It considered spatial parametric non-stationarity with the effects of environmental variables derived from the benefits of RF [102,103]. RFK also added the spatial dependence of the residuals interpolated through OK to the estimated trend, as part of the spatial autocorrelation.…”

“…It considered spatial parametric non-stationarity with the effects of environmental variables derived from the benefits of Remote Sens. 2019, 11, 3004 9 of 25 RF [102,103]. RFK also added the spatial dependence of the residuals interpolated through OK to the estimated trend, as part of the spatial autocorrelation.…”

Multi-Sensor Prediction of Stand Volume by a Hybrid Model of Support Vector Machine for Regression Kriging

“…After the re-design to predict quantitative outputs and solve regression problems, this algorithm came to be the support vector machine for regression (SVR) and acquired wide successes in stand volume modeling [19,20]. Hybrid approaches involve either the statistical regression or machine learning model between the target variable and remote sensing predictors, interpolating residuals of predictions by kriging, and combining them [21][22][23]. Those two-step approaches both consider the spatial heterogeneity conveyed by remote sensing predictors and autocorrelation of neighboring observed data [24,25].…”