Location: Dale Bumpers Small Farms Research CenterTitle: Prediction of topsoil texture through regression trees and multiple linear regressions
|PINHEIRO, HELENA - Universidade Federal Do Rio De Janeiro|
|JUNIOR, WALDIR - Embrapa|
|CHAGAS, CESAR - Embrapa|
|ANJOS, HELENA - Universidade Federal Do Rio De Janeiro|
Submitted to: Revista Brasileira De Ciencia Do Solo
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/17/2017
Publication Date: 4/1/2018
Citation: Pinheiro, H.S., Junior, W.C., Chagas, C.S., Anjos, H.C., Owens, P.R. 2018. Prediction of topsoil texture through regression trees and multiple linear regressions. Revista Brasileira De Ciencia Do Solo. https://doi.org/10.1590/18069657rbcs20170167.
Interpretive Summary: Soil properties and function are necessary for land management and long-term sustainability. Many nations lack the detailed soil information to manage land and make scientific-based agronomic or environmental recommendations. In areas with limited information, modern technology, data and new digital mapping tools can facilitate creation of soil maps. This research utilized satellite imagery/land surface elevation, soil sample data and mathematics to create soil maps for the Guapi-Macacu watershed in Brazil. The digital soil mapping methods with mathematical approaches demonstrated good predictions of soil texture, carbon and depth. These soil properties are crucial for proper agronomic and environmental recommendations. The method also presented ways to estimate the likelihood of correct predictions. Advancement in these types of tools help small farmers in the USA and globally.
Technical Abstract: Users of soil survey products are mostly interested in understanding how soil properties vary in the space and time. In order to support decisions makers, digital soil mapping (DSM) aims to represent the spatial variability of soil properties quantitatively. The goal of this study is to evaluate DSM techniques (Regression Trees- RT, and Multiple Linear Regressions- MLR) and the ability of these tools to predict the mineral fraction content under wide variability of landscapes. The study site is the entire Guapi-Macacu watershed (1250.78 km²) which is located in Rio de Janeiro State in the Southeast region of Brazil. To develop the explanatory variables, terrain attributes and remote sensing data, with 30 m of spatial resolution, were used to represent landscape co-variables selected as an input in predictive models. The selection of sample sites is based on the Latin Hypercube algorithm. A representative set of one hundred points with feasible field access were chosen. Different input databases were tested for the prediction of the mineral fraction content (harmonized and original data). The harmonization of data according to GlobalSoil.Net consortium standards was performed through the Spline algorithm. The results showed better performance from the RT models, using input from six covariates on average; while the simplest MLR model use twice as many input variables which created more complex models without gaining precision. Furthermore, better R² values were obtained using RT models which was independent of the harmonization of soil data. The harmonized dataset at 0-5 cm and 5-15 cm depths, in general, presented better results for the attributes of clay and silt, with R2 values of 0.52 (0-5 cm) and 0.69 (5-15 cm), respectively. The prediction of sand content showed better results when the original depth data was used as input, although all regression tree models had R2 values greater than 0.52. The RT models presented a better statistical index than MLR for all predicted attributes; however, the variance between models suggests similarity of performance. Regarding the harmonization of soil data, both input databases (harmonized or not) can be used to predict soil properties, since the variance of model performance was low and the generalization of soil maps presented similar trends. The products obtained from digital soil mapping approach make it possible the knowledge of uncertainties, providing easier interpretation to the soil management and for land use decisions.