1. Temporal and Spatial Pattern of Thermokarst Lake Area Change at Yukon Flats, Alaska Min Chen(min@lanl.gov), Joel C. Rowland, Cathy J. Wilson, Garrett L. Altmann, Steven P. Brumby Division of Earth and Environmental Sciences, Los Alamos National Laboratory, Los Alamos, NM Introduction Data and Methods Temporal Trend Spatial Pattern Conclusion The development, expansion and drainage of thermokarst lakes depend on the lateral and vertical degradation of permafrost (Hinzman et al.,2005). Consequently, areal changes in thermokarst lakes can reflect changes in the spatial distribution and depth of permafrost to certain degree. However, permafrost degradation is not the only factor that impacts thermokarst lakes, other factors also have significant impacts on lake areas, including precipitation, evaporation (Bowling et al.,2003), connectivity to the rivers (Anderson et al.,2007), floods (Lesack and Marsh,2007). Because of limited availability of remote sensing images and intense work involved in extracting lakes from those images, most investigations on lake area change were conducted by directly comparing lake areas over two to four time periods, without consideration of seasonal and inter-annual variability in lake areas that might be caused by other impacting factors. Lack of consideration of seasonal and inter-annual variability can thus limit our ability to infer causal mechanisms of lake area change and prevent us from separating long-term trends from inter-annual variability (Arp et al.,2011). Besides the long-term trend in lake area change at regional scale, spatial heterogeneity in lake behaviors has also been of increasing interest. The inter-lake variation can mask or skew detection of total lake area change at regional scale (Arp et al.,2011). Investigation of inter-lake variation in area change can help us better understand the hydrologic and geomorphic processes within a region (Arp et al.,2011). In order to better understand the linkage among lake area, permafrost and seasonal and inter-annual variability in climate, we selected a 422,382 ha study area southwest of the Yukon River to explore the temporal and spatial pattern in lake area changes from 1984 to 2009. The goal of our study was to detect whether there was statistically significant long term trend in lake area change and whether the lakes with similar change trend were clustered at certain locations or randomly distributed. If there were significant temporal trend and spatial patterns, we sought to identify key drivers for the temporal and spatial patterns. Fig. 1. Location and Landsat imagery (August 16, 2000) of the study area. Area: 422,382 ha Elevation: 88-150 m Permafrost: Discontinuous Annual Precip: 26.72 cm Annual PET: 48.2 cm Study Area Periods Dates I: 1984-1986 August 12, 1984 July 30, 1985 June 15, 1986 II: 1994 September 9, 1994 III: 1999-2002 June 26, 28, 1999 August 16, 22, 1999 September 6, 8, 1999 June 4, 6, 2000 June 13, 2000 July 6, 8 2000 August 16, 2000 June 16, 2001 September 20, 2001 July 21, 2002 August 6, 2002 IV: 2009 July 16, 2009 August 17, 2009 Table 1 Landsat Image Data (USGS) Table 2 Other Data Sources StationData National Climatic Data Center (daily, monthly) Fairbanks INTL ARPT Precipitation Potential Evapotranspiration * Derived using Priestley-Taylor equation Air Temperature Alaska-Pacific River Forecast Center Beaver, Fort YukonIce-jam flooding NCAR/EOLYukon BridgeBorehole data Data Analysis  Temporal Trend Statistical modeling in R TLA=a+b 1 LWB+b 2 MDT+b 3i PRD+e (eq. 1) TLA — total area of closed basin thermokarst lakes (ha); LWB — local water balance (cm), P-PET since preceding October; MDT — mean daily air temperature (°C), the average of daily mean temperature over the period from May 1 st to the date when Landsat image was acquired; PRD — a dummy variable, represents different time periods, including 1984-1986, 1992, 1999-2002, 2009; a, b — intercept and coefficients, iindicates different time periods; e — the error term  Spatial Pattern Moran’s I in ArcGIS Chi Squired Test in R Fig. 2. Total actual areas of closed basin thermokarst lakes over the study period. Different colors represent different time periods. For 2,280 (26.6% of all lakes) closed basin thermokarst lakes, with a total area of 19, 264 ha (47.9% of total lake area within study area) Predictors 1 Estimates of coefficients 2 P value Intercept23050 (a)<0.0001 Local water balance (LWB)87 (b 1 )0.04 Mean daily temperature (MDT)-780 (b 2 )0.0003 Period (PRD) 19941701(b 32 )0.08 1999-2002-1106 (b 33 )0.05 2009-936 (b 34 )0.15 Table 3 Regression Analysis of Lake Area and Local Water Balance, Summer Mean Daily Temperature and Time Periods Note: 1. LWB, MDT, PRD are predictors and coefficients specified for regression model (eq. 1). 2. a, b 1, b 2, b 32, b 33,b 34 are coefficients specified for regression model (eq. 2). Variability in Lake Area Variability in Climate Factors Fig. 3. Local water balance and summer mean daily temperature over the study period. Red bars indicate summer mean temperature and blue bars represent local water balance. Temporal Trend in Lake Area Change Regression analysis showed that local water balance, summer mean daily temperature and time period explained 94.1% of total variance in lake areas and they were all significant at significance level of 0.05. Lake area increased with local water balance and decreased with summer mean daily temperature (i.e. decreased with active layer depth). Local water balance and summer mean daily temperature together explained 82.1% of total variance in lake areas, and time period accounted for another 12.0%. Compared to lake area (12,296 ha) in 1984, lake area increased by 1,701 ha (13.8%) in 1994, but decreased by 1,106 ha (9.0%) during 1999-2002 and 936 ha (7.6%) in 2009. a b Fig. 4. Lake clusters with different changing trends (a) and their surrounding surficial geology (b) Observed Frequency Expected Frequency Adjusted Standardized Residual Lake Changing Trend DecreaseNo ChangeIncreaseRow Total Deposit Types Alluvial Fan 37 29 1.6 151 152 -0.2 2 9 -2.4 190 Alluvial Terrace 226 197 3.4 1020 1027 -0.7 36 58 -4.5 1282 Floodplain 86 123 -4.5 650 641 0.9 65 36 6.1 801 Column Total 3491821103 2273 Pearson’s Chi-squared test: chi squared=53.9, df=4, p=5.6×10 -11 Table 5 Association between lake changing trends and deposit types Possible Drivers  NO Significant trend in air temperature, precipitation, permafrost temperature.  Change in ice-jam flooding frequency coincided with lake area change at each time period. Periods Ice-jam Flooding Frequency Average Winter Snowfall (cm) Number of years Fort Yukon Beaver Village Total Average (per year) 1979-19868303 0.3754.1 1987-19948415 0.6254.4 1995-20028000 03.2 2003-20097202 0.2863.3 Table 4 Ice-jam Flooding Frequency and Average Winter Snowfall Possible Drivers  Ice-jam flooding frequency  Vertical permafrost degradation caused by heating effect of lakes Fig. 5. Sub-permafrost Groundwater flow in discontinuous permafrost region. Lake A recharges groundwater; Lake B has no connection to groundwater; Lake C is recharged by groundwater. River D is recharged by groundwater.  Taking the lake area (12,296 ha) in 1984 as a baseline, lake area increased by 13.8% in 1994, but decreased by 9.0% and 7.6% during 1999-2002 and in 2009, respectively.  Among the 2,280 closed basin thermokarst lakes, 350 lakes showed an area decrease and 103 lakes showed an increase between 1984-1986 (period I) and 1999-2002 (period III).  The expanding lakes were mainly distributed along the floodplain of Yukon River and its tributaries, while the shrinking lakes were located away from rivers or on alluvial terraces.  Fluctuating ice-jam flooding frequency might be the main driver for the observed temporal lake area change pattern.  Two mechanisms, decreasing ice-jam flooding frequency and local permafrost degradation due to heating effect of water bodies, might be driving the spatial pattern of individual lake area changes. Poster ID: C21B-0468 LA-UR: 11-11849