Sina Muster, Julia Boike, Moritz Langer, Annett Bartsch, Anne Morgenstern, Guido Grosse, Kurt Roth
Zooming out:
From local snapshots to a pan-arctic
inventory of Arctic ponds and lakes
Water bodies in the Arctic are ubiquitious.
Bathurst Island, Canadian High Arctic
Thermal contraction crack polygons
Lena Delta, SiberiaMackenzie Delta, Canada
Mackenzie Delta, Canada
Baffin Island, Canadian High Arctic
Permafrost, Julia Boike, 2010
Examples: Thermokarst
Coastal lowlands, Laptev Sea, Siberia
Ponds emit 40% of landscape–scale CO2 emissions in Siberian polygonal tundra in the Lena Delta.
(Abnizova et al., GBC 2012)
During freezing ponds
produce as much CH4 per square meter as the
average tundra landscape during summer
(Langer et al., RSE, 2014)
Ponds as biogeochemical hotspots
Ponds = water bodies with surface area smaller than 100x100 m
Limits of Global Inventories
• Ponds and small lakes are not mapped on the pan-arctic
scale
• Global lakes and wetland
database (GLWD) maps lakes larger 0.1 km² (100*1000 m)
• MODIS water mask has a resolution of 250 m -> yields confident lake areas larger 0.25 km² (500 x 500 m)
8
Source: Lehner & Döll, 2004
Scientific questions
1. How many ponds and small lakes are there?
2. How can we scale high-resolution but
local water body maps to the global
scale?
Sites
Lena Delta Yukon Delta
Kolyma Lowlands
Yamal Peninsula Polar Bear Pass
Mackenzie Delta
Seward Peninsula Barrow Peninsula
Baku Seoul
Minsk
Tehran Moscow
T'Bilisi Stockholm
P'yongyang
Copenhagen RigaRiga OsloOslo
Astana Astana
Vilnius Vilnius Helsinki Helsinki
Ulaanbaatar Ulaanbaatar
040W
140W
60W
20W 80W
160W
120W
180W
100W
60E
40E 80E
120E 140E
20E
100E 160E
180E
60N
80N
40N 40N
Permafrost
continuous (90-100%) discontinuous (50-90%) isolated patches 0-10%) sporadic (10-50%)
(after Brown et al. 1997)
0 2.5 5 10km
High-resolution mapping
• Kompsat-2, TerraSAR-X, Geoeye, aerial photos
• 0.3 m to 4 m resolution
• 2 km² to 500 km² coverage
5 km
regional
size distributions
Upscaling scheme
12
regional fraction of water surface
unmixing low-resolution
satellite data
representative water body count
High-resolution maps
Representative water body count
13 13
Arga Complex
First Terrace Ice
Complex
(1) Minimum sampling area
(2) Variability within the region
Lena Delta, Siberia
Similar
conditions for lake ice
formation and growth?
Regional probability density functions
14
Lena Delta Barrow Peninsula
Polar Bear Pass Mackenzie Delta
Δp=0.15
Δp=0.02 Δp=0.02
mapped not mapped
Barrow Lena Delta
Inter-regional variability: number and area
15
80 to 99% of the water body number are not mapped
mapped not mapped
5 to 50 % of the
water body area are not mapped
16 16
Upscaling to Pan-Arctic Lake Distribution
error N = xAy
Lena Delta, Siberia
Muster et al., RS, 2013
Conclusions
We need to zoom in before we can zoom out.
• Ensure that probability density functions are representative
• quantify regional variability to give a measure of uncertainty
Representative regional probability density functions can then be used to implement subgrid-scale
information in coarse-scale grids.
Photos courtesy of:
Julia Boike | Konstanze Piel | J.A. Kraulis/Corbis