A remote-sensing approach to studying drought resistance in Swedish old-growth and production forests

Abstract: Boreal forest ecosystems are predicted to experience hotter and drier summers due to climate change, leading to more frequently reduced soil water availability and an increased risk of droughts, as was already the case in Northern Europe during the severe summer drought of 2018. Prolonged water stress in forest ecosystems can lead to impacts ranging from reduced photosynthesis to forest dieback. However, it is currently unknown whether characteristics associated with undisturbed old-growth forests, such as a natural age structure, and the presence of understory and dead wood, are linked to increased drought resistance. Therefore, this study investigated 2018 drought impacts of over 300 Swedish forest pairs through a comparative analysis of spatially proximate old-growth and production forest stands, and their associated Landsat EVI2 Z-scores. Prior to this, several candidate satellite vegetation indices (Landsat EVI2, MODIS EVI2, MODIS NDWI, MODIS CCI) were evaluated on their ability to capture anomalies in gross primary production (GPP) measured at different Eddy-Covariance ecosystem monitoring stations. The candidate indices were also compared at the landscape level across a highly heterogenous ‘case study’ area in northern Sweden. While neither of the indices consistently captured GPP anomalies at the tested stations, the increasingly degraded spatial resolution associated with the MODIS vegetation indices proved to cause severe smoothing effects when investigating impacts across a topographically varied landscape; in contrast, Landsat EVI2 captured negative anomalies along drier slopes, while moister valleys were associated with positive Z-scores. It was therefore chosen as the most suitable index, and for further analysis, forested areas were separated into five distinct topographical soil moisture classes. Despite their older age, old-growth forests were on average associated with significantly higher Z-scores than their surrounding production forests. Overall Z-scores increased with increasing soil moisture, whereas relative differences (old-growth – production Z-score per forest pair) decreased. When normalized by the frequency distribution of the soil moisture classes, the mean difference amounted to +0.13. Furthermore, variability in drought impacts was higher for old-growth than production forests, and consequently, both the stands associated with the highest positive and negative anomalies were old-growth. The largest negative Z-score differences can be reconciled by the fact that those old-growth forests were on significantly steeper slopes and had a higher fraction of purely deciduous stands. Apart from that, however, relative differences in drought impacts were not associated with a clear spatial pattern, and no general relationship was found with drought severity, nor relative differences in age, elevation, and slope. Differences were therefore likely caused by other factors not considered.