Spatial soil properties distribution in “Hoya del río Suárez” region in Colombia.

There is a spatial soil properties distribution surface in raster format. This is the result of a study about digital soil mapping. Implementing geoestatistics (regression kriging RK) and machine learning algorithms (random forest RF, support vector machines SVM, and ensemble models), the study found the best performance for 5 soil properties (clay fraction, bulk density, total porosity, pH, and cation exchange capacity). The study was located in a region named “Hoya del río Suárez”, which is the main sugarcane-producing region in Colombia, and its land area is around 47000 hectares. Those raster surfaces were carried out in 2021, and the database used for doing this study was compiled between 2015 and 2016.

本数据集为栅格格式的土壤属性空间分布曲面，系一项数字土壤制图研究的成果。该研究采用地理统计学（geoestatistics，RK，即回归克里金法）与机器学习算法（随机森林RF、支持向量机SVM及集成模型）开展建模实验，最终针对5种土壤属性实现了最优预测性能，涉及的属性包括黏粒含量、土壤容重、总孔隙度、pH值与阳离子交换量。研究区域为哥伦比亚核心甘蔗产区苏亚雷斯河谷（Hoya del río Suárez），该区域土地总面积约4.7万公顷。上述栅格分布曲面生成于2021年，本研究所使用的数据库于2015至2016年间汇编完成。