Nonstationary hydrologic behavior in forested watersheds is mediated by climate-induced changes in growing season length and subsequent vegetation growth
Forested watersheds provide important ecosystem services through the provision of high quality freshwater, mitigation of floods, and maintenance of base flows. How alteration of these services under ongoing climate change is mediated by vegetation dynamics is not fully understood. Combining independent remote sensing based vegetation information and distributed hydrological modeling, we investigated the impact of climate-induced vegetation dynamics on long-term non-stationary hydrologic behavior in two forested watersheds in the southern Appalachians. We found significant increases in precipitationrunoff deficit (defined as annual precipitation minus annual runoff), equivalent to annual evapotranspiration plus storage changes, over the last three decades. This non-stationary hydrologic behavior was significantly correlated with long-term and interannual changes in growing season length and subsequent vegetation growth. These patterns in vegetation phenology were attributed primarily to minimum temperature regimes, which showed steeper and more consistent increases than temperature maxima. Using a distributed modeling framework, we also found that the long-term non-stationary hydrologic behavior could not be simulated unless full vegetation dynamics, including vegetation phenology and long-term growth, were incorporated into the model. Incorporating seasonal vegetation dynamics also led to the improved simulation in streamflow dynamics, while its effect spread out through the following dormant seasons. Our study indicates that non-stationary hydrologic behavior has been closely mediated by long-term seasonal and structural forest canopy interaction with climate variables rather than directly driven by climatic variables. This study emphasizes the importance of understanding the ecosystem responses to ongoing climate change for predictions of future freshwater regimes.