The Interactions Between Fire and Hydroclimate Over Seasonal Timescales 543 views

Fire is a ubiquitous component of the Earth system and in African drylands in particular. It represents a dramatic and instantaneous change at the land surface by affecting plant communities, hydrologic cycling and the energy balance. The aim of this dissertation is to build understanding of how fire is affected by, and in turn affects, regional climate. This dissertation consists of research on two fronts. In the first I develop a novel methodology to explore the hypothesis that strong seasonality enhances burned area. Using monthly, global, gridded temperature and precipitation data I derived seasonality metrics that can be used to describe a periodic seasonal cycle. Using just three such metrics and a random forest model, I explained 66% of the variance in global burned area, on par with significantly more complex models that are limited to a regional scope. A more complex random forest model that included all 9 metrics correctly predicted 87% of the variability in global burned area. These findings confirm that seasonality plays a large role in determining global burned area and suggest mechanism by which this occurs. The methodology developed in this Chapter will be useful for other researchers wishing to describe seasonality using standardized, interpretable metrics. The second research area surrounds the hypothesis that fire can influence rainfall on seasonal timescales. First, I demonstrated a relationship between dry season fires and subsequent with a statistical model and observational data. I find that more extensive and later dry season fires account for reductions of up to 30 mm of rainfall (~%10 of average yearly totals) in the subsequent dry season. The observed effect is strongest in regions that are already water limited. This could potentially lead to a negative feedback represented by an interannual oscillation in rainfall and fire activity, an effect observed in actual rainfall records. I then used a simple physically-based boundary layer model to evaluate how the land surface could contribute to these observed rainfall deficits over the Kalahari region of southern Africa. Using simple, but realistic parameterizations of fire at the land surface, I showed that positive albedo anomalies (brightening) or increases in latent heat flux after fire could explain observed rainfall reduction. This is in large part by less vigorous boundary layer growth and a reduced probability of the boundary layer exceeding the lifting condensation layer. I also showed new satellite-based evidence that brightening does indeed occur after fires over the Kalahari transect in regions receiving less that 850 mm of rainfall annually. This finding challenges the idea that immediate darkening is the only meaningful albedo change after dryland fires and supports the idea that brightening is responsible for observed rainfall deficits after fire. Finally, I extended this observational approach to the whole continent of Africa. I applied a pixel grouping technique to label satellite burned area data into individual fire events and compared the albedo following fire to a surrounding unburnt reference region. On average albedo was +2.71 x 10-4 higher in burn scars in the five years following fire, representing a statistically significant negative forcing on a continental scale. These findings build new understanding of the land surface effect of fire and the potential for interactions with regional hydroclimate.

PHD (Doctor of Philosophy)

fire; land-atmosphere interactions; Africa; precipitation

English

All rights reserved (no additional license for public reuse)

2018-05-08