Quantifying precipitation and land-surface temperature controls on the vegetation dynamics of the semi-arid tropical high Andes based on remote sensing data

Knowledge of how abiotic environmental controls like precipitation and temperature are related to dynamics of vegetation is important for improved predictions of effects of climate variability on biodiversity reduction and habitat degradation. Vegetation dynamics in arid and semi-arid regions of the world are highly dependent on precipitation, in particular the processes of recruitment, growth and reproduction, nutrient cycling and net ecosystem productivity. However, results from research studies of high-latitude mountain regions also indicate that temperature exerts a strong control on vegetation. The southern tropical Andes is a region of both semi-arid monomodal austral summer precipitation and of strong temperature gradients related to strong elevation gradients. We investigate the relationship between vegetation dynamics, precipitation and land surface temperature along a selected transect across the southern tropical Andes. Vegetation data are determined from the Normalized Difference Vegetation Index (NDVI), which is derived from Moderate Resolution Imaging Spectroradiometer (MODIS) data at a spatial resolution of 250 m and a temporal resolution of 16 days. Land surface temperature is derived from the MODIS eight-day land- surface temperature (LST) product that has a spatial resolution of one square kilometre. Precipitation data is computed from the Tropical Rain Measurement Mission (TRMM) data, available at three-hourly intervals and on a 0.25-degree geographic grid. As a first step in our analysis, we apply a simple land cover (LC) classification scheme using thresholds derived from NDVI-based phenologic metrics over ten hydrological years (2001-2010). We delineate areas in which no LC change is detected (i.e., stable LC classes). We then use the stable LC classes for spatial aggregation of mean NDVI values for each TRMM pixel. Finally, we aggregate precipitation data to annual precipitation sums and compute LST nocturnal minima and diurnal maxima for each TRMM pixel. The results of our study show a moderate correlation (r² = 0.49) between annual mean precipitation and annual mean NDVI for all stable LC classes over the entire range of values of annual mean precipitation from 2001 to 2010. However, when annual mean precipitation ranges between 60 – 600 mm per year, the explained variance (r² = 0.79) is significantly higher, a result that indicates a strong relationship between precipitation and vegetation dynamics within the study region. Multivariate regression analyses of mean NDVI to mean annual precipitation and mean annual difference of minimum nocturnal to maximum diurnal LST per TRMM pixel (mean annual LST amplitude) shows a very good coefficient of determination (r² = 0.73). This result implies that mean annual LST amplitude is the second major control on vegetation dynamics after precipitation within the tropical high Andes. Our findings indicate that changes in LST cause potential elevation changes of LC within the study region.