The impact of decadal land cover change on the global warming potential of Beringian Arctic tundra

The goal of this study is to (1) determine how land cover in the Beringian Arctic changed in the last half century; (2) assess what biophysical properties control peak growing season land-atmosphere CO2 and CH4 exchange in multiple landscapes and land cover classes in Beringia; and (3) model how decadal land cover change in Beringia has altered peak growing season CO2 and CH4 exchange and global warming potential. Using a campaign-style, snapshot sampling approach sixteen sites were visited in ten different landscapes throughout the Beringian Arctic between 2005 and 2008. Sites represented a broad range of arctic terrestrial ecosystems, and data collection included CO2 exchange, CH4 exchange, and a number of biophysical and spectral properties for the purpose of spatial scaling and model development. For seven landscapes, ground-truthed land cover maps were created from recent high-resolution Quickbird imagery. Using conservative assumptions regarding land cover change, modern land cover maps were used as baselines for the development of historic high-spatial-resolution land cover maps derived from aerial photography and declassified military imagery dating back to 1948. Using these multi-temporal coverages, trends in decade time scale land cover change were determined for each study landscape. Within Alaska, drier landscapes and open water cover classes expanded whereas wet vegetated land cover classes decreased in area. For Russian landscapes, shrub dominated land cover expanded wherever these were present and land cover generally shifted towards an expanse of wetter landscape vegetation types. Multiple regression models were developed using field data. These were able to effectively predict CO2 and CH4 flux (R 2 = 0.70 and 0.66 respectively) for a range of vegetation types and landscapes at multiple locations in the Beringian Arctic. Originating from measurements taken during the snapshot sampling campaign, the models were relatively simple, spatially scalable models whose input parameters could be derived from automated ground and aerial/satellite based observation platforms. The effectiveness of these models suggests that predicting the GWP of landscapes across Beringia many multiple landscapes may only require the measurement of simple ecosystem measures and given the variety of landscapes within this study, these relationships may extend to other parts of the Arctic as well. Ecosystem fluxes were then spatially extrapolated over the multi-temporal land cover maps to determine the impact of land cover change on CO2 and CH4 flux. Using the global warming potential (GWP) metric, we calculated the global warming potential of these landscapes in CO 2 equivalents (CO2e). Results suggest all landscapes were historic net sinks of carbon and remain net sinks of CO2e. Four of the seven sites appeared to become weaker sinks, while the remaining three sites became stronger sinks of CO2e. Decadal changes in CO 2 and CH4 flux as well as global warming did not appear to have any geographic associations. When the effect of land cover change on NDVI was calculated, most landscapes displayed a change in NDVI consistent with regional aggregations measured at coarse spatial scales (i.e. an increase for Alaskan landscapes, except for the Barrow and Atqasuk sites, and a decrease for Russian landscapes). These findings build on the current understanding of the relationship between ecosystem structural structure and function change and draw attention to the importance of understanding how their spatio-temporal variation can affect global warming potential over decadal time scales. Findings also suggest that using simple set measurements within a network of automated sensors could allow the development of a cost-efficient network for monitoring fluxes. In the process of building ecosystem flux models, the novel snapshot-sampling approach developed in this study demonstrates the capacity for fast and efficient sampling of large areas in combination with remote sensing platforms. The simple models and capacities demonstrated here would benefit the future development of an integrated Arctic observing network. (Abstract shortened by UMI.