Interactions and influences of savanna fire across multiple extents

Access full-text files




Marden, Alexander William

Journal Title

Journal ISSN

Volume Title



The spatiotemporal interactions and feedbacks of fire and vegetation in savanna systems are a key component of savanna structure, function, and can give insight into woody plant encroachment. The increased accessibility to remote sensing data and software via Google Earth Engine for fire analysis facilitates integrated savanna assessments for local-to-regional analyses and management. This study assesses spatial and temporal fire dynamics in the Botswana Kalahari at different extents and resolutions to test the hypothesis that fire and vegetation patterns interact via distinct processes at different extents and resolutions in savanna systems. MODIS burn products (MCD64A1 500meters) and in situ vegetation and fire measurements were used to test the affects of variables at different extents and resolutions on savanna fire and vegetation patterns. Spatiotemporal analysis at the extent of the Botswana Kalahari explored 1) the spatial autocorrelation of fire return time using a bivariate Moran's I analysis to observe how large patterns of fuel dynamics affect neighboring areas across time and space and 2) how variability of variables over small spaces influence larger patterns of fire occurrence using Geographically Weighted Regression or GWR. In situ grass and woody vegetation characteristics were measured and combined with MODIS fire history data to observe how vegetation at the square meter scale is influenced by regional affects of fire occurrence, grazing, and human influence using OLS regression. Fire intensity and post-fire vegetation mortality assessments were performed on sites that burned approximately two months after initial measurements were recorded to observe vegetation affects on fire intensity and the initial affects of fire on vegetation. Woody plant populations were projected across a number of fires under divergent fuel conditions to observe how fine resolution fire and vegetation patterns influence larger patterns.

In the Botswana Kalahari study area, fire occurrence over time was heavily affected by fire presence in neighboring 500m pixels (first and second order), an indication that spatiotemporal patterns of fire are affected by patterns in neighboring areas. The temporal and spatial patterns observed suggest that temporally dynamic neighboring fuel conditions impact fire in a given location.

Large scale patterns were observed between variables and fire occurrence using an OLS regression, and spatial variability of local coefficients was observed using a GWR model. Seasonality of precipitation, boreholes, and EVI had negative significant coefficients. Soil moisture, drought severity, and herbaceous cover had positive significant coefficients, but when examined locally using GWR there were high amounts of spatial variation -- every variable ranged from positive to negative significant local coefficients except for seasonality of precipitation. Explanatory power of the variables was significantly improved by the GWR model. The variation in the local coefficients present from the GWR maps indicate that larger patterns of fire presence are influenced by locally specific contexts, while the relative consistency of coefficient sign and high coverage of significance of EVI and herbaceous cover variables signals that fuel conditions are important across the area.

The In situ OLS model showed that fire along with regional patterns of herbivory and human influence were highly impactful on grass biomass. Past fire presence was significant and highly correlated with grass biomass. Woody plant canopy cover and regionally specific anthropogenic influence via a road/river with an associated grazing intensity gradient had negative associations with grass biomass.

The In situ pre and post-burn vegetation measurements show that distinct fuel conditions caused fire intensity to vary dramatically over a small area on the same day, and the projected woody plant population model suggests that continued disparate fuel conditions would cause significantly different vegetation conditions. Post burn vegetation measurements indicated that the plot with high grass biomass and low woody biomass had high fire intensity while the plot with low grass biomass and high woody biomass had a low intensity fire. Long term modeling of the two fuel scenarios projected high amounts of woody plant growth with high recruitment and little mortality in the low fuel scenario, while the high fuel scenario projected a stable woody plant population. These findings support the idea that fine resolution variability of fire presence affects larger patterns of vegetation uniformity/non-uniformity.

Taken together, the findings suggest that different specific factors interact between extents and resolutions of fire patterns -- land use at large extents affect fire presence in a given area, and fine resolution vegetation conditions affect directionality and spread of fire that influence larger patterns of fire. However, fuel connectivity is a common variable that connects these interactions, and since fuel conditions are affected by past fire presence and intensity, fire history is a common factor that is important across scales. With these interactions in mind, the observed vegetation heterogeneity in the study is likely caused in part by the diversity of land uses in the surrounding areas.



LCSH Subject Headings