Spatial prediction of soil water retention in a Páramo landscape: Methodological insight into machine learning using random forest

Blanco, Carlos M. Guio; Martínez-Gómez, Victor; Crespo, Patricio; Ließ, Mareike

doi:10.1016/j.geoderma.2017.12.002

Cited by 88 publications

(55 citation statements)

References 82 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The TOC, sand, silt, and clay are known to influence the SWHC of any kind of soil, and they are commonly used in pedotransfer functions or PTF (Gupta, Dowdy & Larson 1977;Weil & Brady 2017). We included topographic variables as indirect predictors because these parameters have been shown to increase the quality of PTF (Sharma, Mohanty & Zhu 2006;Blanco et al 2018). The elevation, aspect, and convexity were extracted from maps at 5 m resolution.…”

Section: Ensemble Of Small Models Of Swhcmentioning

confidence: 99%

“…Nevertheless, as SWHC is expected to be variably influenced by these two variables in different soils (Brady et al 2008), the use of topographic predictors can help to refine their incorporation into the models or can be used as proxies for other factors that explain SWHC variation (e.g., soil depth, type of pedogenesis) that cannot be measured across large areas. Some studies even used only one predictor derived from the Landsat images and a digital elevation model (Blanco et al 2018) to spatialize the soil water retention capacity. In our study, topographic variables were significantly selected as predictors for all pF values.…”

Section: Ensemble Of Small Modelsmentioning

confidence: 99%

See 1 more Smart Citation

Spatial modelling of soil water holding capacity improves models of plant distributions in mountain landscapes

et al. 2019

View full text Add to dashboard Cite

AimsThe aims of this study were: 1) to test a new methodology to overcome the issue of the predictive capacity of soil water availability in geographic space due to measurement scarcity, 2) to model and generalize soil water availability spatially to a whole region, and 3) to test its predictive capacity in plant SDMs. MethodsFirst, we modelled the measured Soil Water Holding Capacity (SWHC at different pFs) of 24 soils in a focal research area, using a weighted ensemble of small bivariate models (ESM). We then used these models to predict 256 locations of a larger region and used the differences in these pF predictions to calculate three different indices of soil water availability for plants (SWAP. These SWAP variables were added one by one to a set of conventional topo-climatic predictors to model 104 plant species distributions. ResultsWe showed that adding SWAP to the SDMs could improve our ability to predict plant species distributions, and more specifically, pF1.8-pF4.2 became the third most important predictor across all plant models. ConclusionsSoil water availability can contribute a significant increase in the predictive power of plant distribution models, by identifying important additional abiotic information to describe plant ecological niches.

show abstract

Section: Ensemble Of Small Models Of Swhcmentioning

confidence: 99%

Section: Ensemble Of Small Modelsmentioning

confidence: 99%

Spatial modelling of soil water holding capacity improves models of plant distributions in mountain landscapes

et al. 2019

View full text Add to dashboard Cite

show abstract

“…Random forest, a machine learning algorithm (Breiman, 2001), was used to determine which soil or terrain properties were most predictive of LISR within each site‐year. Random forest creates an ensemble of decision trees using a random subset of observations and covariates, preventing overfitting issues that are common in other decision tree algorithms like Classification and Regression Trees (CARTs) and Iterative Dichotomiser 3 (ID3; Grimm, Behrens, Märker, & Elsenbeer, 2008; Guio Blanco, Brito Gomez, Crespo, & Ließ, 2018).…”

Section: Methodsmentioning

confidence: 99%

“…In two separate study years, soil map unit, SOM, STP, and STK had high predictive values in RF models built on data pooled across 11 sites (Smidt et al., 2016). While some farmers use soil properties to implement VRS, many farmers use yield maps from previous years to delineate their management zones (Grisso, Alley, Phillips, & McClellan, 2009). Historical yield may be useful to determine AOSR, since high yielding areas generally have a lower agronomic optimum plant density (AOPD; Carciochi et al., 2019).…”

Section: Introductionmentioning

confidence: 99%

Soil and terrain properties that predict differences in local ideal seeding rate for soybean

et al. 2020

View full text Add to dashboard Cite

The agronomic optimum seeding rate (AOSR) of soybean [Glycine max (L.) Merr.] varies based on environment. Understanding where AOSR varies within a field is useful for farmers utilizing variable rate seeding technology. An AOSR representing an area smaller than a whole field is referred to as local ideal seeding rate (LISR). The objective of this study was to identify soil and terrain properties that were most predictive of differences in LISR. Seeding rate trials were established at four fields in 2017 and three fields in 2018. Yield data were used to estimate LISR 33-68 times per field.Soil properties were estimated at the same scale as LISR using 0.2-ha grid samples and terrain properties were calculated from 0.76-m digital elevation model developed using light detection and ranging data. Random forest analysis was performed to identify which soil and terrain properties were predictive of LISR within each site-year. At all site-years, terrain properties were generally more predictive of LISR compared to soil properties. Valley depth and general curvature were in the top-five most predictive properties at four of seven site-years. Moving from the lowest valley to the highest ridge was associated with an increase in LISR of 76,000 seeds ha −1 . Moving from the lowest relative slope position to the highest relative slope position was associated with a 38,000 seeds ha −1 increase in LISR. Terrain properties may be appealing to farmers because they relate to LISR, are publicly available data, and stable over time.

show abstract

“…Random Forest (RF) model is a non-parametric technique that has been successfully applied to soil properties prediction (Wiesmeier et al, 2011;Castro Franco et al, 2015;Hengl et al, 2015;Chagas et al, 2016;Yang et al, 2016;Dharumarajan;Hedge;Singh, 2017;Silva et al, 2017;Blanco et al, 2018;Wang et al, 2018a). The model combines a set of decision trees to improve the accuracy of prediction of a given environmental variable, where each tree is generated by bootstrap samples (random sampling with substitution), leaving one-third of training samples, called Out-of-Bag (OOB) data, for using in the model's performance evaluation (Breiman, 2001;Liaw;Wiener, 2002).…”

Section: Machine Learning Algorithms For Ko Predictionmentioning

confidence: 99%

Land-use influence on the soil hydrology: An approach in upper Grande River basin, Southeast Brazil

et al. 2019

View full text Add to dashboard Cite

Brazil is a large country that depends on the hydroelectricity generation hydropower dams. The Upper Grande River Basin (UGRB) is one of the most important Brazilian hydrological regions in terms of water availability and electric energy production. Therefore, studies of water availability are indispensable for a better and more successful decision making in water resources management in the region. This study objective to approach the land-use influence on the soil hydrology in the Upper Grande River Basin, a strategic headwater basin of southeastern Brazil. This study uses hydrological indicators (baseflow/runoff (BF/R) and overland flow/runoff (OF/R)) extracted from eight watersheds, varying the size and localization in the region, to support the results found. Soil saturated hydraulic conductivity (Ko) was determined in situ using a constant flow permeameter, totaling 224 sampled points. Five machine learning algorithms were compared in their performance to predict Ko (Random Forest, Support Vector Machine, Gradient Boosting, Linear Regression, Regularization) using terrain attributes as covariates. The tested methods for predicting Ko resulted in a relatively low coefficient of determination (R2) due to the high spatial variability of this soil hydrologic attribute. The hydrological indicator BF/R was sensitive to land-use changes in the watersheds. The greatest Ko values were associated with native forest and the least values area associated with pasture and rupestrian field.

show abstract

Spatial prediction of soil water retention in a Páramo landscape: Methodological insight into machine learning using random forest

Cited by 88 publications

References 82 publications

Spatial modelling of soil water holding capacity improves models of plant distributions in mountain landscapes

Spatial modelling of soil water holding capacity improves models of plant distributions in mountain landscapes

Soil and terrain properties that predict differences in local ideal seeding rate for soybean

Land-use influence on the soil hydrology: An approach in upper Grande River basin, Southeast Brazil

Contact Info

Product

Resources

About