The Application of Cluster Analysis to Investigate Multivariate Spatial Patterns in Belizean Lowland Savanna Soils

Abstract

Belizean lowland savannas are viewed as acidic, nutrient poor and generally unsuitable for agriculture. Increasing population pressures and emerging technologies may drive agricultural development onto the savannas in the foreseeable future, a process which was witnessed in the Brazilian cerrados. In Belize, the most suitable land management would have to be practised in the correct regions to ensure success. This study identified the most suitable regions for agriculture using cluster analysis. Cluster analysis was performed on a novel dataset, collated from previous Belizean land use studies. Analysis of this dataset found that Belizean lowland savanna soil is generally acidic (mean surface pH=5.45) and nutrient poor (mean surface TEB=3.4cmol/kg). However, the soil properties are spatially heterogeneous. These variations were examined using multivariate cluster analysis, where K-Means clustering produced a lower/better C-Index than Ward’s hierarchical clustering when 5 groups were generated (k-Means=0.61, hierarchical=0.68. Characteristics and spatial distributions of the groups were explained through examination of site characteristics and pedogenisis of Belizean soils. Clay poor groups (K5(3) & K5(4)) occurred due to variations in bedrock geology, and heavy rainfall caused extreme leaching in Southern groups i.e. K5(5). In other groups, (K5(1) & K5(2)) undulating local topography was combined with high clay levels to cause poor drainage. Group K5(2) has the most agriculturally suitable pH range (surface pH=5.3–7.1) and nutrient levels (mean surface OM=8.79%, total N=0.26%, available P=4.12ppm, TEB=17.4cmol/kg). However, high clay levels (mean subsurface clay=45.5%) will impede rooting and cause waterlogging during the wet season. Further investigations are required before solutions can be proposed and the future management of these areas can be recommended.