A two-scale framework for mapping site productivity of Eucalyptus globulus Labill. plantations in northern Spain in the context of climate change and using spatially explicit environmental variables as predictors

This research aimed to obtain accurate estimates of the productivity of eucalyptus plantations under different climate change scenarios without the need for additional fieldwork. Thus, we used tree growth data from 1,102 research plots, existing spatially continuous environmental data, and the random forest (RF) algorithm to construct raster-based models. We constructed models to predict site index (SI) at landscape scale (250 m·pixel^-1), which is useful for planning purposes and for analyzing the effect of climate change on productivity, and at forest plot scale (resolutions of 10, 25, 50, and 100 m·pixel^-1), which is essential for predicting plantation yields. All models explained ~50% of site index variability, as is usual in this type of study. We found that the different spatial resolutions of predictor variables did not affect the amount of variability explained. This finding may be due to two opposing effects on the explained variability at finer scales: a positive effect, as finer scales enable capture of microscale landform variability through a high-resolution digital elevation model (DEM), and a negative effect due to the introduction of “noise” when downscaling the climatic and lithological information from coarser scales. Elevation and the climatic variables (mainly temperature) were the most important predictor variables: For every 100 m-increase in elevation, the productivity decreased by on average 0.3–0.9 m of site index (1–1.3 m³·ha^-1·year^-1 of maximum mean annual increment in volume) and for each degree-Celsius-increase in annual mean temperature, productivity increased by about 2.2 m in site index (3 m³·ha^-1·year^-1 of maximum mean annual increment in volume). Due to the forecasted increase in temperatures under climate change, productivity is expected to increase significantly in Eucalyptus globulus plantations in northern Spain in the coming decades, by between 1.68% and 3.38% of the current average site index under the most pessimistic climate change scenario and between 1.79% and 2.48% of the current average site index for the moderate scenario. We conclude that currently available spatially continuous environmental data can be used to develop accurate raster data models for predicting site productivity for E. globulus without the need for fieldwork. The spatially explicit maps produced in the study provide support to forest planners, forest managers, private landowners and politicians, enabling well-founded decisions to be made regarding selection of the best sites for afforestation and providing accurate yield predictions for the plantations.