Search

• Multitemporal RS images & digital soil mapping were used to predict soil moisture. • Concept of dynamic and static environmental covariates (ECs) were introduced & used. • Dynamic EC helped capture spatiotemporal variation in soil moisture with confidence. • A general model in predicting soil moisture was proposed using multitemporal images. Soil moisture (SM), a critical component of the global hydrological cycle, is affected by individual or combinations of multiple factors including soil properties, climate, and topography. Despite its importance to many disciplines, predicting SM continuously, accurately, and inexpensively over a large area is a great challenge due to its dynamic nature controlled mostly by the spatial and temporal variability of these factors. Static environmental covariates, such as those derived from a digital elevation model, are commonly used in digital soil mapping (DSM); these are typically less suitable for predicting dynamic properties. Easily available multi-temporal satellite images show strong promise to capture this variability. The objective of this study was to predict SM from multi-temporal satellite images using a DSM approach. Specifically, we examined the feasibility of using dynamic, static, and combinations of environmental covariates (ECs) to predict SM in the Balikhli_Chay watershed in Iran on four separate dates in June, July, August, and September 2018 coincident with satellite overpass. Cubist and random forest (RF) machine learning algorithms (MLAs) were trained for making SM predictions for individual dates, and the data was then compiled without considering the date to create generalized models. The baseline for comparisons were the models developed using only static ECs. For June, July, August, and September, Cubist R2 improvements were 96%, 78%, 185% and 120%, respectively. Using the generalized models, R2 improved by as much as 223% and RMSE decreased by as much as 47% when comparing the best SM prediction model in each month to models developed using only static ECs for that same month using the Cubist model. Similar model improvements were seen for the RF model. The generalized Cubist and RF MLAs performed equally well with concordance of 0.91 and 0.90 for Cubist and RF respectively, and low RMSE of 3.04 and 2.98. The best Cubist and RF MLAs by month were always those developed with dynamic, or satellite-derived, ECs. Based on the variable importance statistics, land surface temperature (LST) was the most important EC. This study showed the strong predictions, and the practical feasibility of using multi-temporal satellite data as a dynamic EC that could help to capture the spatial and temporal variations of soil moisture. This approach could likely be extended to other dynamic soil property (e.g., soil temperature). [ABSTRACT FROM AUTHOR]

Modeling and mapping of soil properties are critical in many environmental, climatic, ecological and hydrological applications. Digital soil mapping (DSM) techniques are now commonly applied to predict soil properties with limited data by developing predictive relationships with environmental covariates. Most studies derive covariates from a digital elevation model (named static covariates). Many works also include single-day remotely sensed satellite imagery. However, multitemporal satellite images can capture information about soil properties over time and bring additional information in predicting soil properties in DSM. We refer to covariates derived from multitemporal satellite images as dynamic covariates. The objective of this study was to assess the performance of DSM when using terrain derivatives (static covariates), single-date remotely sensed satellite indices (limited dynamic covariates), multitemporal satellite indices (dynamic covariates), and combinations of terrain derivatives and satellite indices (covariate fusion) as covariates in predicting soil properties and estimating uncertainty. Three soil properties are considered in this study: organic carbon (OC), sand content, and calcium carbonate equivalent (CCE). Inclusion of single and/or multitemporal remotely sensed satellite indices improved the prediction of soil properties over traditionally used terrain indices. Significant improvements were observed in the prediction of soil properties using two models, Cubist and random forest (RF). The increase in the R2 values for Cubist and RF were 126% and 78% for OC, 110% and 54% for sand, and 87% and 32% for CCE. The RMSE decreased by 34% and 27% for OC, 25% and 12% for sand, and 39% and 19% for CCE, when compared to the terrain indices only model. This also reduced the uncertainty of estimation and mapping. These clearly showed the advantage of using multitemporal satellite data fusion rather than simply using static terrain indices for DSM of soil properties to deliver a great potential in improving soil modeling and mapping for many applications. Unlabelled Image • We introduced dynamic & static environmental covariates (ECs) for digital soil mapping. • Dynamic EC improved soil prediction over static ECs including terrain indices. • Multi-date satellite images captured the variations from change in soil properties. • Multi-date satellite images also reduced the uncertainty in prediction and mapping. • Combination of dynamic and static ECs had a larger influence on soil prediction. [ABSTRACT FROM AUTHOR]