The environmental drivers of annual variation in forest greenness are variable in the northern Intermountain West, USA

Abstract

AbstractPeak snowpack in the western USA has decreased 21% since the mid‐20th century, and these trends are forecasted to continue over the next century. In water‐limited systems, forest productivity during the growing season is assumed to be linked with snowpack during the previous winter, but this linkage has proven not to be universal in studies across varying ecosystems. We compared peak Normalized Difference Vegetation Index (NDVI, a proxy for vegetation productivity) at relatively fine scales (30‐m pixel) to point measurements of snow water equivalent (SWE) at 169 locations scattered among the northern Intermountain West for the 25‐yr period between 1993 and 2017. We further compared NDVI to additional environmental variables including topographical variables (elevation, aspect, and slope), and climatic variables (Palmer Drought Severity Index, snowmelt date, and snowmelt length). We hypothesized that peak SWE would be a strong predictor of peak NDVI, but the strength of the SWE–NDVI relationships would vary spatially in our topographically complex study area. We observed weak relationships between snowpack and vegetation productivity, with only 10% of the locations displaying statistically significant correlations between annual peak SWE and peak NDVI. Furthermore, we determined that the response of NDVI to snowpack and environmental variables differed among subregions of our study area. For example, we observed strong SWE–NDVI associations in the continental‐most portion of our study area and weak correlations in other subregions of our study area. Our results highlight that in regions characterized by complex topography, (1) environmental drivers of NDVI are highly variable and (2) fine‐resolution studies more accurately capture the spatial heterogeneity of NDVI compared to course‐resolution studies. However, one consequence of fine‐resolution studies may be that microclimate effects at each site swamp the signal of predictor‐ecological response relationships, whereas these relationships may be more discernable in larger scale remote sensing studies.