1. Who Pulled the Plug? The Mysterious Case of Disappearing Siberian Lakes T.J. Bohn 1, R. Schroeder 2, E. Podest 2, N. Pinto 2, K.C. McDonald 2, and D.P. Lettenmaier 1 1 Dept. of Civil and Environmental Engineering, University of Washington, Seattle, WA, USA 2 JPL-NASA, Pasadena, CA, USA, UW-UBC Hydrology Symposium Vancouver, BC, 2011-Sep-30

2. Disappearing Arctic Lakes! (Smith et al., 2005) Smith et al. (2005) Science Compared LANDSAT imagery over W. Siberia from 1973 with 1997/1998 Found 6% decrease in total lake area 10% of lakes shrank to below 40ha 1% of lakes completely disappeared Hypothesized thawing permafrost as the driver This has been cited many times as evidence of dramatic climate change

3. Why we should care Approximately the same amount of carbon is stored in the world’s permafrost as is currently in the atmosphere (> 500Gt, Zimov et al, 2006) If permafrost thaws, soil carbon will decompose to CO2 (or CH4) and escape into the atmosphere Potential feedback to global warming

4. How are lakes involved? Much of the Arctic land surface is covered by lakes, ponds, and wetlands –Permafrost can impede drainage of water through soil –Low summer ET rates –Flat topography –Snowmelt can inundate large areas, seasonally

5. Breach Permafrost is sensitive to lakes As permafrost thaws, melt from excess ground ice collects in ponds/lakes – increase in lake area At same time, warming climate warms the lakes and thaw bulbs get deeper Eventually, thaw bulbs may extend completely through permafrost If permafrost was impeding lake drainage, lake can now drain – decrease in lake area –Lakes have low albedo –Ice floats and insulates the lake in winter –Lake bottoms tend to resist freezing: “thaw bulb” or “talik” Lakes are sensitive to permafrost Ice Cover (in winter) Liquid Water Thaw Bulb Permafrost Unfrozen Soil/Rock Lakes are indicators of permafrost health

6. Smith et al’s Hypothesis Lake area changes showed distinct geographic pattern: In continuous permafrost zone, 12% increase in total lake area –Melting of excess ground ice? (first phase of permafrost degradation) In discontinuous, isolated, and sporadic permafrost zones, 5% to 9% decrease in total lake area –Lakes thawing through permafrost? (second phase of permafrost degradation) (Smith et al., 2005)

7. Are *all* Arctic lakes thawing through permafrost? Lakes actively thawing through permafrost has big implications for climate change projections Liberation of massive amounts of carbon stored in frozen soil, as CH4 (e.g. Walter et al. (2006)) Problems with Smith et al’s hypothesis: Only 1% of the lakes examined were found to vanish completely –Can we assume that the other lakes are going to vanish? Observations had poor temporal coverage –All LANDSAT images were from “summer” – presumably JJA –Where multiple observations existed, the minimum lake area was taken to be the “true” lake area (all extra assumed to be seasonal) –Images only came from three years: 1973, 1997, 1998 Can we draw conclusions based on a few snapshots? Could some other process have caused lake areas to change?

8. Modeling Framework VIC hydrology model –Large, “flat” grid cells (e.g. 100x100 km) –On hourly time step, simulate: Soil T profile Water table depth Z WT NPP Soil Respiration Other hydrologic variables… Link to CH4 emissions model (Walter & Heimann 2000)

9. VIC Dynamic Lake/Wetland Model Water & energy balance model Includes mixing, ice cover Dynamic area based on bathymetry Can flood surrounding wetlands based on topography Bowling and Lettenmaier, 2010 Special application: treat all lakes, ponds, and inundated wetland area as a single “lake”

10. Lake Bathymetry/Topography Lake depths from literature SRTM and ASTER DEMs for surrounding topography Lake size histograms from GLWD (Lehner and Doll, 2004) and LANDSAT Lake storage-area relationship LANDSAT courtesy of E. Podest and N. Pinto of NASA/JPL ILEC Arctic Thaw Lakes Bog Pools Depth(m) Cumulative Area (km 2 ) 10 0100 0 5

11. Arctic Application Model simulates permafrost –Validation of active layer depths & soil temperatures is underway Model simulates heat flux below lake, and reproduces thaw bulb under lake However, lake does *not* require presence of permafrost to exist –Soil under lake is not permeable enough for lake to drain through soil very quickly –Soil under lake = wetland soil = peat –Soil’s baseflow rate was calibrated – we might experiment with this Lake dynamics dominated by snowmelt, ET, and surface drainage (via outlet channel) Outlet channel width calibrated to match remotely- sensed inundation extent product

12. AMSR/QSCAT-Derived Inundation Courtesy R. Schroeder, NASA/JPL Annual Max – Min Fractional Inundation Daily, for snow- free days 2002-2009 25km resolution Max % - Min % > 15 % < 3%

13. Comparison with AMSR For each grid cell, calibration iterates over lake outlet width until annual average simulated lake area is unbiased with respect to AMSR inundation product Calibrations do this iteration for a range of values of wetland microtopography and soil baseflow parameters The combination having lowest MSE/VAR metric is taken ( = 1 – Nash-Sutcliffe Eff.) MSE is of monthly simulated lake area vs. AMSR/Quickscat product, 2002-2007 VAR is variance of monthly AMSR/Quickscat product, 2002-2007 MSE/VAR of 1.0 can be achieved with a straight line through the mean, i.e. we want to get scores substantially closer to 0.0 than to 1.0 for a good fit Tundra / Cont. Permafrost Uvaly Hills Irrigated Crops

14. Simulated Changes in Lake Area -30 -20 -10 0 10 20 30 Percent change in lake area 1973-06 to 1997-06 % change Simulation gives similar magnitude & pattern of change, for the same approximate snapshot as Smith et al. (2005) Break into 4 latitude bands: –67-72 N (continuous permafrost) –64-67 N (discontinuous) –61-64 N (isol./sporadic) –57-61 N (non-permafrost) –All bands cover 65-90E

15. 60-year Timeseries Average monthly lake fraction from: –Continuous permafrost (67- 72N) –Discont. permafrost (64-67N) –Isolated/Sporadic permafrost (61-64N) –Non-permafrost (57-61N) Seasonal cycle is larger than interannual variability 0.15 0.10 0.05 0.0 0.15 0.10 0.05 0.0 0.15 0.10 0.05 0.0

16. Shifts in Seasonal Cycle: Lake Area Monthly average lake area fraction, by decade, 1967-2006 All regions show increase in peak lake area, 1970s- 2000s In north, peak occurs in mid-June; peaks occur progressively earlier as we move south Between 1970s and 1990s, JJA lake areas increase 10% in north; decrease 2- 10% elsewhere Could it be that what really happened was a simple shift forward in the snow melt pulse (plus larger snow packs)? Cont Perm Discont Perm Isolated Perm Non Perm Month Fract. Area Approx. 10% increase, 1970s-1990s Approx. 10% decrease, 1970s-1990s

17. Monthly average SWE Melt begins/ends earlier as we move N to S Peak lake area falls generally in 2 nd month of melt All regions show increase in peak SWE, 1970s-2000s –Corresponds to increases in peak lake area Shifts in melt timing are difficult to see here at monthly resolution Shifts in Seasonal Cycle: SWE SWE (mm) Month Cont Perm Discont Perm Isolated Perm Non Perm

18. Shifts in Snow Melt Dates Both observed (NOAA) and simulated (VIC) snow melt has shifted 3-10 days earlier in the 3 latitude bands 61-64N, 64- 67N, and 67-72N. VIC gives a consistently later snow melt date (1-7 days) than NOAA S of 67N –Likely due to NOAA’s snow absence threshold being 50% coverage, compared to VIC’s 0 mm SWE threshold. This shift is a plausible cause of the shift towards earlier dates of peak inundation and lower midsummer lake extent 1967-761977-861987-96 1997-06 1967-761977-861987-96 1997-06 Sporadic and isolated perm. Discont. perm. Cont. perm. No perm.

19. Conclusions Permafrost degradation is not the only hypothesis that explains the changes in lake area observed by Smith et al (2005) The bulk of the “changes” in lake area could also be caused by: –shift towards earlier snowmelt –larger spring snow packs The observed geographic pattern in lake area changes could be the result of: –Summer sampling in the north falls around the peak in lake area –Summer sampling in the south falls mainly in the recession following the peak

20. Conclusions Assessment of more remote sensing imagery, spanning more years and a longer portion of the year, would help solve this puzzle Model limitations: –We should double-check model performance in tundra region –We should try to simulate lakes actively thawing through permafrost – any differences? How many lakes are actively thawing through permafrost has big implications for feedbacks to global warming

21. Thank You