Floods have a devastating impact on infrastructure, agriculture, and the natural landscape,
costing an average of $8.83 billion per year in the United States since 2007. Floods mainly
occur in response to extreme precipitation events. For example, Hurricane Irene caused an
estimated $14.2 billion in damage in the eastern United States over a 6-day period in August
2011. The Northeast has experienced the largest increase in extreme precipitation in the
nation, with a 53% increase in 1996-2014 compared to the 1901-1995 average. However, the
impact of increases in total and extreme precipitation on flooding is unclear.
To better understand the connections between precipitation and flooding in the Northeast,
we developed a random forest model to identify key meteorological and land surface drivers
of river discharge in three different watersheds across a range of latitudes. Our developed
model reasonably captures daily flows, with larger errors in simulating 95th percentile high
flows and 2-year high flows. We find that three day antecedent precipitation is the most
important predictor of high flows for the two southern (less snow-dominated) systems,
while antecedent soil moisture was the most important high-flow predictor in the
northernmost watershed. Although snow melt moderately contributes to model prediction
accuracy in the winter, precipitation and soil moisture remain the dominant factors in flow
prediction throughout the year for all three watersheds. We further quantify the effects of
climate change on flooding in the Northeast by forcing our developed model with climate
change projections. Specifically, we analyze the difference in simulated flow using future
(2070-2099) and historical (1976-2005) regional climate model output.