Procedures to Simulate Missing Soil Parameters in the Florida Soils Characteristics Database

Abstract

In this study, the Florida Soils Characterization Database (FSCD, http://flsoils.ifas.ufl.edu/) was used as source of physical and chemical soil parameters for the development of a soils characteristics data set to be used for crop growth models. Because this database has missing soil parameters within the various soil horizons, a study was undertaken to fill in missing data using statistical estimation procedures. Three main objectives of this study were to: (i) develop procedures to extract soil characteristics data from available soil data sources; (ii) convert soil profile data into a format usable by simulation models; and (iii) create and evaluate multiple regression models to simulate missing soil parameters. We used the conditional distribution model to investigate empirical relationships within the available measured soil data. A multiregression model is a valuable tool that can be used to fill singular or multiple missing soil data in the incomplete list of the FSCD. Statistical indexes such as the RMSE and the RMSE-observations standard deviation ratio (RSR) were used to test the behavior of a multiregression model in simulating missing soil values in the 0- to 20-cm horizon. The results showed that silt, clay, organic C, cation exchange capacity, bulk density, and sand could be predicted with a model performance that ranged from very good to satisfactory. However, pH and saturated hydraulic conductivity were not simulated with satisfactory performance levels. The approach taken in this study to simulate missing soil parameters is universal, meaning that it can be developed anywhere having available a representative source of measured soil data of the study area.