Constraining spatial variability of methane ebullition seeps in thermokarst lakes using point process models

Ebullition is an important but highly heterogeneous mode of methane emission in lakes. Variability in both spatial distribution and temporal flux creates difficulty in constraining uncertainties in whole lake emission estimates. Analysis of short- and long-term flux measurements on 162 ebullition seeps in 24 panarctic lakes confirmed that seep classes, identified a priori according to bubble patterns in winter lake ice, have distinct associated fluxes irrespective of lake or region. To understand the drivers of ebullition's spatial variability and uncover ways to better quantify ebullition in field work, we combined point-process modeling with field measurements of 2679 GPS-marked and classified ebullition seeps in three Alaskan thermokarst (thaw) lakes that varied by region, permafrost type, and seep distribution. Spatial analysis of field data revealed that seeps cluster above thawed permafrost soil mounds in lake bottoms. Seep density and clustering, determined from field observations, were used as parameters in a Poisson cluster process model to simulate seeps across entire lake surfaces. Sampling results indicated that (1) applying seep-class mean flux values to unmeasured seeps counted on ice-bubble surveys does not compromise accuracy of whole lake flux estimates; (2) three distributed 50 m(2) ice-bubble survey transects more accurately estimate mean lake ebullition than 17 dispersed 0.2 m(2) bubble traps; and (3) the uncertainty associated with whole lake mean ebullition estimated by lake-ice survey transects is inversely related to seep density. Findings suggest that transect field data collected on a large number of widely distributed lakes can be combined to provide a well-constrained, bottom-up estimate of regional lake ebullition.