Local-scale factors matter for tree cover modelling in Mediterranean drylands

Forests contribute directly to ecosystem structure and functioning, maintaining biodiversity, acting as a climate regulator and reducing desertification. To better manage forests, it is essential to have high-resolution forest models and appropriate spatial-explicit variables able to explain tree cover at different scales, including the management scale. Most tree cover models rely only on broad-scale variables (>500 m), such as macroclimate, while only few studies include also local-scale variables (<500 m). This study aimed to identify the importance of local-scale factors relative to broad-scale factors and identify the environmental variables at different scales that explain tree cover in oak woodlands in Mediterranean drylands. Sixty sites previously identified as being covered with Holm oak or Cork oak were stratified by precipitation. Normalized Difference Vegetation Index, used here as a surrogate of tree cover, was modelled using simultaneously broad-scale factors (macroclimate) and local-scale factors (microclimatic and edaphic conditions). The percentage of variance explained by local-and broad-scale factors and the effect size of each environmental variable on tree cover was determined for the study site. It was found that local-scale factors and their interaction with broad-scale factors explained more variance than broad-scale factors alone. The most important local-scale factors explaining tree cover were elevation, potential solar radiation, used as a surrogate of microclimatic conditions, and wetness evaluated terrain used as an indicator of water flow accumulation. The main broad-scale factors were related to temperature and precipitation. The effect of some local-scale variables in tree cover seems to increase in areas where water as a limiting factor is more important. This study demonstrates the critical importance of including local scale factors in multi-scale modelling of tree cover to obtain better predictions. These models will support well-suited forest management decisions, such as reforestation and afforestation plans to reverse evergreen oaks decline in Mediterranean drylands.