Risks of neglecting phenology when assessing climatic controls of primary production

Abstract

We evaluated the effect that integrating annual aboveground net primary production (ANPP) along different 12-month periods has on temporal modelsof productivity (ANPP as a linear function of annual precipitation). We studied Argentinean Patagonia, which encompasses a variety of climates and biomes. Using MODIS normalized difference vegetation index (NDVI) to estimate green biomass,we assessed the date of maximum annual NDVI for 2000?2016. One quarter of Patagonia (West/South region) exhibited a well-defined seasonality, withmaximum NDVI during spring?summer, whereas the rest (Central/East region) showed a much less well-defined maximum NDVI, generally during fall. Then we calculated temporal models for each pixel, considering both annual and seasonal precipitation (PPT), in two ways: (i) centered models, integrating NDVI for a period centered at the actual growing season, that is, July?June for West/Southregion and January?December for Central/East region, and (ii) displaced models, switching the NDVI integration period. Our results indicate that, with the centered models, 84% of the Central/East region exhibited significant temporal models, but only 52% of the West/South region did. For the displaced models, 60% (40%) of pixels of Central/ East (West/South) region changed their best predictor of ANPP. In general, the best predictor changed from current-year PPT to current-plusprevious- year PPT or from current-year fall to previous-year fall. Our results suggest that more attention must be paid in choosing the integration period for annual ANPP. This is more than a formal matter since the putative best predictor of ANPP can dramatically change depending on the assumed phenology.