Most dryland ecosystems show high landscape heterogeneity that can influence soil fertility, although the underlying processes are still poorly understood. Furthermore, our understanding of the same could be affected by the scale dependency of the landscape representation. Here, we study the relationships between soil and landscape attributes at different spatial resolutions in a semiarid area, to better understand which landscape processes control soil fertility and whether such control is affected by the resolution of landscape representation.
Materials and methods
A stratified sampling plan based on topography, vegetation type and lithology was carried out in a semiarid catchment of south-east Spain to select the soil sampling sites. Furthermore, the landscape attributes were resampled at five different resolutions (5, 10, 20, 40 and 80 m) from digital elevation model and remote sensing data. Principal Component Analyses (PCA) for soil and landscape attributes were performed to ascertain the main trends of variation of each set. Then, correlations between the soil and landscape principal component scores at each resolution were calculated to determine the relationships between the two sets of variables.
Results and discussion
The first component of soil PCA (S-PC1) was related with a number of properties linked to soil fertility: soil organic carbon, nutrient availability, cation exchange capacity, pH and water-holding capacity. The S-PC1 was correlated positively with the first component of landscape PCA (L-PC1), which represents areas of dense vegetation cover associated with topographic convergence and/or low solar radiation. Besides, S-PC1 was also correlated negatively with L-PC2, which represents areas with intense channel formation on easily erodible lithologies and scarce vegetation cover. As the resolution of landscape representation decreased, the explained variance increased faster in L-PC2 than in L-PC1, showing that the representation of each underlying landscape process was more evident at a particular resolution. The relationship between soil fertility and landscape also seemed to change with spatial resolution, as S-PC1 was correlated best with L-PC1 at finer resolutions while the correlation with L-PC2 increased at coarser resolutions.
The relationship of S-PC1 with L-PC1 and L-PC2 suggests the presence of a soil fertility gradient mainly ‘driven’ by water availability derived from local topographic conditions against gully erosion processes arising from geomorphological features. Each landscape process ‘controlling’ soil fertility is better represented at different resolutions: conditions that improve water availability were more evident at finer resolutions (5–10 m), while erosion was more evident at medium-coarse resolutions (20–40 m).