“Water has become a highly precious resource. There are some places where a barrel of water costs more than a barrel of oil.”
Dead spruce near Steamboat Springs, Colorado
Remote Sensing/Statistical Modeling
By combining 30m global forest disturbance datasets with long-term, ongoing USGS water gage data via Google Earth Engine, we are able to accurately monitor when and where disturbance occurs, and tie that to disruptions in the water outputs of critical watersheds across the nation. Currently, we're looking at the impacts to over 700 watersheds, from southern Florida to northwest Washington.
Most people assume disturbance = more water. Not so, when you start looking at the literature. It turns out there is a substantial amount of watersheds that have lower water post-disturbance. This is significant if you're a land manager who cares about their water supply and sees a wildfire cresting the ridge, as well as a home owner concerned about flooding post-mountain pine beetle outbreak.
Collaborative Hydrological Modeling
We are fusing the data on forest canopy damage with a high resolution hydrology model (DHSVM) to estimate the impact of these large disturbances on water yield now and in the future. These outputs are critical to land managers who are attempting to supply thirsty, populous regions.
One potential strategy that we've explored (see publications) is the possibility of planting more climatically suitable species post-disturbance, essentially using natural disturbances to kick start migration (semi-assisted migration, if you will). It appears to have some effect, but only in some types of forests/watersheds. Figuring out when and where such management tools are effective is a significant challenge, but important, for future forest management.
There are many forested watersheds (and non-forested) across the country that are monitored for water. We can use these to get at what exactly makes a watershed produce more water after a disturbance, or less, or change the timing of the delivery of that water. Details: (a) Percent forest loss in the largest disturbance event (2001-2010) and the original percent forested for the undammed watersheds, with a range of disturbance percentages, (b) post-disturbance response group distribution in terms of deviations in flow timing, and (c) deviations in water yield. Response groups are similar watersheds that had similar behavior in regards to either total water yield or timing (treated independently). Boxplots show 25th and 75th percentiles, whiskers extent to 1.5 the interquartile range beyond those percentiles. Outliers shown as points.
The most explanatory variables for post-disturbance watershed response, with predictor variables standardized and binned into response groups (y-axis shows standard deviations from the mean). Differences among groups illustrate their associative values relative to watershed response in terms of water yield (top) or streamflow timing (bottom). Large values of percent forest disturbed were strongly associated with observed increases in water yield. High snow percentage was associated with insensitivity to disturbances in terms of streamflow timing. Numbers indicate significantly different groups (p < 0.05, pairwise Kruskal-Wallis rank sum test). Boxplots show 25th and 75th percentiles, whiskers extent to 1.5 the interquartile range beyond those percentiles. Outliers shown as points.
Predicted water yield response to an imposed 20 % disturbance per watershed (total n = 671). Dammed watersheds are shown in faded colors. Above local-scale variability, the Great Plains and Southeast exhibit increased water yield post-disturbance, whereas the Southwest, Northeast and central Rockies are relatively insensitive, while the northern Rockies exhibit water yield decreases. The Pacific Northwest is highly variable.
