Olympic Peninsula lakes for game worlds

Question: a game set on Washington's Olympic Peninsula needs lakes, and the usual first pass is a circle with a clean, even edge. It does not read as a mountain lake. How did the real lakes form, what shapes do they have, why do some look like a raised rim holding water, and what numbers can a lake generator aim for?

Short answer: the Olympics are not volcanic, so no lake is a crater. Ice, landslides and rivers made them, in two populations with a near-empty band between: valley-floor and lowland lakes below about 500 m, and cirque and bench tarns above about 900 m. The high lakes are small, none over 17 ha, and every natural lake over 25 ha lies below about 220 m.

The typical lake is a cirque tarn shaped like a rounded oval, egg or teardrop: about 1.5 times as long as wide, with a shoreline development index of 1.2 to 1.3. Only one high lake in four is rounder than 1.3 to 1. Lowland lakes run 2 to 3 to 1, with arms and bays. Nearly every lake above 900 m is perched on a step: the ground rises about 125 m within 200 m behind it and falls 40 to 50 m below the water within 200 m past its outlet. That step, not a crater, is what reads as a rim holding water.

Most of these figures were measured for this page from 936 OpenStreetMap lake polygons and SRTM elevations. Section 7 turns them into twelve rules for a generator.

The main published source is Water-Supply Bulletin 43, volume 1, a 1976 USGS and Washington Department of Ecology survey by Bortleson, Dion, McConnell and Nelson [1]. For each lake it sampled it gives measured shoreline development, depth and bottom slope, with aerial photographs and bathymetric maps.

The new measurement covers all 1,317 OpenStreetMap water polygons over 500 m² between 47.30 and 48.20° N and 124.75 and 122.95° W [2], 936 of them natural lakes and ponds of at least 0.1 ha. Each was measured for area, perimeter, length, width, shoreline development, convexity and the Fourier spectrum of its outline, and its surface and surroundings were sampled from SRTM 30 m elevations [3]. The prose calls these figures measured; the tables mark them †. Section 8 gives the method and its limits.

The range is uplifted sea-floor basalt and sedimentary rock, scraped off the Juan de Fuca plate and folded up against North America [4]. Two ice systems shaped its lakes. The Cordilleran ice sheet's Juan de Fuca lobe flowed west along the Strait and the range's north edge roughly 16,900 to 14,000 years ago, scouring the low valleys and leaving drift, moraines and drumlinoid ridges on the northern lowlands [4], [5], [6]. Alpine glaciers flowed from high cirques down the valleys, advancing 25 to 40 miles down the Hoh, Queets and Quinault [5].

Where the lakes are by elevation

Measured natural lakes, reservoirs excluded:

Of the 360 lakes above 900 m, only 98 exceed 1 ha, 22 exceed 5 ha and 5 exceed 10 ha.

The typical outline

A tarn's measured medians are a length to width ratio (Feret) of about 1.5 to 1.6, a shoreline development index (SDI) of 1.2 to 1.3 and a convexity (area over convex-hull area) of about 0.92. The outline usually has one broad bulge or lobe and one flatter or slightly indented side. Only about one high lake in four is below 1.3 to 1, and two thirds of high lakes over 1 ha are not even star-shaped from their centroid: a bend or lobe hides part of the shore from the middle. Lake Angeles is teardrop-shaped [8]; Heart Lake is named for its shape [11]; Moose and Royal lakes are long and narrow, about 2.5 to 1; Gladys Lake is a lobed blob, SDI 2.2 and convexity 0.60.

Lowland and valley lakes are a different family: about 2 to 3 to 1, SDI 1.5 to 2.6, with arms and bays that follow their trough.

By formation type

Cirques are about as wide as long. Medians over 1,593 cirques: width over length 1.05, length 625 m, width 656 m, height range 310 m, wall height 210 m, maximum (headwall) gradient 57°, plan closure 128° [20]. A tarn fills only the floor, so a tarn of 1 to 10 ha is 100 to 400 m across. Its long axis has no preferred direction: of 108 measured high lakes at least 1.4 to 1, a third each run along, diagonal to and across the fall line.

Named lakes

Length is the longest straight line across the polygon (Feret); mean width is area over length. † marks a measured area or SDI from the polygon, or an SRTM elevation at the shore; SDIs in bold are the 1976 survey's.

Lake Crescent's published 12 miles [14] is about 50 % longer than its straight-line length.

Alpine analogues with measured depths

No published depth was found for an Olympic high lake. The only measured high-lake depths found are North Cascades lakes in the same bulletin [1]:

Zr, the bottom slope, is maximum depth as a percentage of mean diameter, Zm·50√π/√A (pages 193 and 234 to 239). The bulletin's Olympic lowland lakes have a Zr of 0.86 to 2.4 %, so cirque tarns are about 4 to 10 times deeper for their size. Lake Crescent works out at about 3.7 % and Quinault about 1.7 %.

These are the landforms that read as a lake on a mound.

a. A cirque tarn behind a rock lip or moraine. A cirque has cliffs on three sides and a fourth side of till forming the lip or sill; erosion deepens the basin, and a moraine may dam a tarn below it [26]:

headwall (cliff, 45-65°)            open side
  \                                  ____ lip / moraine crest (a few m to ~20 m above water)
   \  talus apron                   /    \
    \______  ~~~~~~ water ~~~~~~~ /       \  steep drop to the valley
           \_____ deep basin ____/          \   (100-200 m within 500 m)
           (deepest point toward the headwall / centre)

b. A perched (bench) lake. A cirque floor or shelf hangs above a main valley that the trunk glacier cut deeper; the ground falls away past the outlet. At Hoh Lake, Sol Duc Lake and Lake Constance it falls 150 to 250 m within about half a kilometre of the shore.

c. A kettle in hummocky moraine, a round pond among mounds of till [17], rimmed by only metres, and only on the drift lowlands.

d. Moraine- and landslide-dammed valley lakes. Quinault (moraine) and Crescent, Sutherland and Lena (landslides) have a raised dam at one end only; the rest is valley wall.

e. True crater lakes, like Oregon's Crater Lake or Mount St Helens' crater, are volcanic and absent here.

Measured cross-sections

For each lake of at least 0.5 ha, SRTM elevations were sampled on 16 bearings at 50, 200 and 500 m beyond the radius of a circle of equal area. Rise is ground minus lake surface:

• Nearly every lake above 900 m is perched. Behind it the ground rises about 125 m within 200 m and 220 m within 500 m, a headwall averaging roughly 30° and steeper near the top. In front it drops 40 to 50 m below the water within 200 m and 140 to 170 m within 500 m.
• The lake sits on a step about half-way up a slope, nearer its base. Walls stand on about two thirds of the sides: at +200 m the median share of bearings more than 10 m above the lake is 0.69. The high side and the drop are typically 112 to 135° apart.
• Lowland lakes are the opposite: low banks (+30 m at 200 m) and no drop past the outlet.
• The open side faces N, NE or E for 66 % of 141 lakes above 1,000 m (N 28, NE 38, E 27, SE 10, S 12, SW 4, W 8, NW 14), matching the cirques' aspect: the Cameron Glaciers lie in four north to northeast-facing cirques [27].

Where the deep water is

From the bulletin's bathymetric maps [1]:

• Twin Lakes, lower (cirque): one bowl, deepest (over 90 ft) about 40 % of the way across, offset toward the steep side, where contours are packed. A shelf under 20 ft covers about a third of the lake on the gentle outlet and meadow side.
• Twin Lakes, upper (cirque): one oval bowl, deepest (over 90 ft) slightly toward the far wall, steep all round.
• Tomyhoi (trough): two sub-basins, 40 and 35 ft, at the down-valley end, shoaling toward the inlet, where a braided delta fills the last quarter or so with water under 10 ft.
• Lake Crescent: much of the shore drops off steeply, often as a sheer underwater cliff [14].

From the bulletin's 1971 to 1974 aerial photographs and remarks [1]:

In terms of the mean radius R: bends and lobes are 0.5 to 2 R, deltas and fans 0.1 to 0.4 R, talus, knob and notch roughness 0.05 to 0.2 R on a tarn. The metre scale is below what a map captures, but it is what a player on the shore sees.

Shoreline development index

The SDI, D_L = L / (2√(πA)), is shoreline length over the circumference of a circle of equal area, so a circle scores 1. Elongate lakes score highest, and high values are common in lakes along old drainages or behind dammed streams [1]; crater and kettle lakes score near 1, glacial troughs and oxbows high [21]. A 2022 paper gives medians of 1.67 for Scandinavian lakes (range 1.11 to 4.54) and 2.17 worldwide (1.14 to 10.24), and warns that the SDI rises with lake area and map resolution because shorelines are fractal [30]. Compare lakes only at the same size and resolution.

Elongation and convexity

All measured:

The outline spectrum

For each lake over 1 ha that is star-shaped from its centroid, the radius r(θ) was sampled at 256 angles, Fourier-transformed and divided by the mean radius R. Only 20 of 68 high lakes and 18 of 92 lowland lakes qualify. Median amplitudes as fractions of R:

The RMS radial deviation by band, high lakes (lowland): k = 2, 0.16 (0.28); k = 3 to 6, 0.10 (0.12); k = 7 to 16, 0.04 (0.05); k = 17 to 32, 0.013 (0.017). From k = 3 to 16 the amplitude falls as k^−1.4 for high lakes (IQR −1.6 to −1.1) and k^−1.46 for lowland ones. OpenStreetMap draws a high lake with roughly 20 to 50 vertices, so amplitudes above k ≈ 16 are underestimates.

Fractal dimension and depth

Percolation theory gives d = 4/3 for unscreened perimeters. Lakes worldwide of at least 0.46 km² give d ≈ 1.4 (the corrected value) and Swedish lakes of at least 4.7 km² d ≈ 1.34 [31], [32]. The slope of US lake perimeter against map resolution is 0.21 (D ≈ 1.21), less than earlier estimates [33]. Aim for D ≈ 1.2 to 1.3 on tarns and 1.3 to 1.4 on big lowland lakes.

In the bulletin's terms [1], volume development (mean over maximum depth) is 0.43 to 0.53 alpine, 0.30 to 0.66 lowland, about 0.48 at Crescent and 0.55 at Quinault: low means a conical basin, high a steep-sided, flat-bottomed one. Zr is about 9 to 10 % for cirque tarns, 2 to 4 % for troughs and 1 to 2.4 % for lowland lakes.

Numbers are measured medians unless marked; ranges are what to sample from.

1. Two populations with a gap. In the box about 39 % of lakes of at least 0.1 ha lie above 900 m, 2 % between 500 and 900 m and 58 % below 500 m; artificial ponds inflate the low share, and inside the park lowland lakes are rare [7]. On a mountain map make about 2 lakes in 3 cirque or bench lakes at 900 to 2,000 m, 1 in 3 valley or lowland water below 500 m, and almost none between, those on benches or behind landslide dams. High water is tarns, basin lakes, chains and many tiny moraine ponds, over half of it under 0.5 ha. Low water mixes rare large troughs and landslide lakes, round kettle-like lakes, common small floodplain and beaver ponds, and marshy lakes (a judgement from the types, not a count).
2. Size by elevation. High lakes are log-normal: median about 0.4 ha, p90 about 3.5 ha, capped at 15 to 20 ha (radius about 35 m, 105 m and 230 m). Only valley-floor lakes may be large, 20 to 3,000 ha, and rarely.
3. No circles. Draw length to width from a log-normal: tarns median 1.5, p10 to p90 about 1.15 to 2.5; valley and trough lakes median 2.2, range 1.5 to 3, length over mean width 4 to 8 for big troughs and 4 to 6 for narrow alpine ones; kettles 1.0 to 1.4. Align trough, landslide and moraine lakes with the valley; orient tarns uniformly to the fall line.
4. Outline harmonics, as fractions of R. Tarns: k = 2 of 0.16 to 0.34, random-phase k = 3 to 16 of about 0.07·(k/3)^−1.4, and finer noise bringing k = 17 to 32 to 0.015 to 0.03 R, erring high because the maps undersample: about 10 % RMS at 2 to 6 lobes, 4 % at 7 to 16 bumps, 1.5 to 3 % finer. Lowland lakes: k = 2 of 0.3 to 0.45, k = 3 to 6 of 12 to 18 %, k = 7 to 16 of 5 to 9 %. At a fixed sampling resolution the SDI should land at 1.2 to 1.35 for tarns, 1.5 to 1.8 for mid-sized lowland lakes and 2.3 to 2.6 for big troughs.
5. Shapes that are not star-shaped. Two thirds of high lakes over 1 ha and four fifths of lowland ones have a bend or lobe radial noise cannot make. Build about 1 tarn in 3, and most lowland lakes, from two or three overlapping ellipses or a bent spine (a curved centreline of varying half-width), then add noise. Make about 1 tarn in 5 clearly two-lobed, like Heart or Gladys Lake (a judgement consistent with a convexity p10 to p25 of 0.77 to 0.85), and big troughs curved ribbons (Crescent) or bodies with 2 to 4 arms (Ozette). Convexity: tarns 0.90 to 0.93 (p10 about 0.77), lowland 0.83 (p10 about 0.5).
6. Perched bowls on a step. A headwall on about two thirds of the perimeter rises about 125 m within 200 m and 220 m within 500 m, steepest near the top (a cirque's maximum is 55 to 65°), over a talus apron of 25 to 35°. The open side faces N, NE or E two times in three. A lip a few metres to about 20 m above the water carries one outlet notch, and beyond it the ground drops 40 to 50 m within 200 m and 140 to 170 m within 500 m. This holds for about 95 % of lakes above 900 m; a bench lake swaps the headwall for a gentler valley wall.
7. Depth. Tarns: maximum 0.09 to 0.10 of the mean diameter (23 to 29 m at 4 to 8 ha), mean 0.45 to 0.5 of the maximum, deepest 30 to 45 % across toward the headwall, contours packed on the headwall and talus side, and a shelf under 6 m over 25 to 35 % of the area on the outlet and meadow side. Troughs: 0.02 to 0.04, steep walls, a shoaling inlet delta. Lowland lakes: 0.01 to 0.02, with wide shallow mucky margins.
8. Inlet deltas as shoreline. Each creek pushes out a rounded fan 0.1 to 0.4 R wide on a tarn, up to about 1 km on a big lake (Barnes Point, 55 ha), under 3 m deep for about its own width, with gravel and sand bars. A tarn has 0 or 1 permanent inlets, often only snowmelt seeps; bigger lakes gain about 67 % more per doubling of area. One trough lake in three gets a braided delta over 15 to 25 % of its length (Tomyhoi).
9. Talus and bedrock bulges. Where the shore meets a wall, add convex talus or avalanche fans every 100 to 300 m (2 to 5 a tarn), each standing out 0.05 to 0.15 R, and sharper bedrock knobs every few tens of metres: the main source of k = 7 to 16 roughness.
10. One outlet, at the low shore. Put it at the rim's lowest point on the open side; lowland and moraine lakes put it down-valley, landslide lakes at the slide, perhaps hidden (Lena drains underground). Shape it as a 5 to 20 m narrowing or spit, not a smooth arc.
11. Edges by elevation. Small and mid-sized lowland lakes: sedge or marsh on 75 to 100 % of the shore (little or none on deep, clear or alpine lakes), logs and snags on almost every forested shore, drowned trees in beaver ponds and landslide lakes, islands on about 5 to 10 % of lakes over 1 ha (Angeles, Ozette, Flapjack). Alpine lakes: boulders, scree and late snow down to the water on the shaded south or headwall side, meadow on the outlet side.
12. Chains and groups. 74 % of water bodies above 900 m have another within 1 km (46 % within 0.5 km), and 51 % of high lakes of at least 0.5 ha have one that size within 1 km. Put about half the high lakes in groups of 2 to 8 in one basin (Seven Lakes Basin, Grand Valley, Royal Basin), stepping down one drainage about 100 m at a time (Gladys 1,641 m, Moose 1,539 m, Grand 1,442 m), each outlet feeding the next lake.

Polygons. OpenStreetMap natural=water ways and multipolygons in the box were downloaded through the Overpass API on 2026-10-08, under the ODbL [2]. Rivers, basins, wastewater ponds and canals were dropped; reservoirs were kept but left out of the statistics. Areas match published values within a few per cent (Crescent 2,038 against 2,075 ha, Ozette 3,035 against 2,954 to 3,151 ha, Constance 4.35 against 4.29 ha), though Quinault's 1,435 ha is under the published 1,509 ha.

Elevations. SRTM 30 m through OpenTopoData, at a shoreline vertex [3].

The ned10m trap. OpenTopoData's ned10m dataset returns values about 3.3 times too small (feet treated as metres) over much of the southern and eastern Olympics: Lena Lake came back at 165 m instead of 556 m. A first pass on it misplaced 195 of the 936 lakes, so every number here uses SRTM.

Rings. Rise and drop use 16 bearings at the equivalent radius plus 50, 200 and 500 m from the centroid. Some +50 m points on long lakes fall on water, so only +200 and +500 m are quoted. The lowest ground at +500 m stands in for the outlet direction.

Resolution. Lakes under 2 ha are drawn with 14 to 29 vertices, which understates the SDI and fine roughness: read those numbers as lower bounds.

Region. The box includes lowland around Port Angeles, Sequim and Hood Canal, where many small ponds are artificial. That inflates the lowland count, not the mountain numbers.

What is still unknown

• Olympic high-lake depths: none found; the cirque depths above are North Cascades analogues.
• PJ Lake is not mapped by name, so it has no measured shape.
• Kettles: no source names an Olympic kettle; calling the round lowland lakes kettles is an inference.
• The 2022 SDI medians come from a search summary, because the paper returned an access error.
• Oxbows: none is tagged as one in OpenStreetMap here, so they were not measured.

1. [1]
   G. C. Bortleson, N. P. Dion, J. B. McConnell, and L. M. Nelson, "Reconnaissance data on lakes in Washington, vol. 1: Clallam, Island, Jefferson, San Juan, Skagit, and Whatcom Counties," Washington Dept. of Ecology, Water-Supply Bull. 43, 1976. Accessed: Oct. 8, 2026. [Online]. Available: https://apps.ecology.wa.gov/publications/documents/wsb43a.pdf
2. [2]
   OpenStreetMap contributors, "OpenStreetMap data, ODbL," queried through the Overpass API. Accessed: Oct. 8, 2026. [Online]. Available: https://overpass-api.de
3. [3]
   Open Topo Data, "SRTM 30 m elevation dataset," Open Topo Data. Accessed: Oct. 8, 2026. [Online]. Available: https://www.opentopodata.org
4. [4]
   U.S. National Park Service, "Geology of Olympic," Olympic National Park. Accessed: Oct. 8, 2026. [Online]. Available: https://www.nps.gov/olym/learn/nature/geology.htm
5. [5]
   U.S. National Park Service, "Glaciation," in NPS Natural History Handbook: Olympic, NPS History. Accessed: Oct. 8, 2026. [Online]. Available: https://npshistory.com/handbooks/natural/1b/nh1bc.htm
6. [6]
   HistoryLink.org, "Vashon glacier begins to melt and recede from Puget Sound region and Columbia Basin around 16,900 years ago," HistoryLink.org, essay 5087. Accessed: Oct. 8, 2026. [Online]. Available: https://www.historylink.org/file/5087
7. [7]
   S. J. Brenkman et al., "Unveiling a legacy of fish introductions to mountain lakes using historical records and eDNA surveys in a National Park," Front. Conserv. Sci., vol. 6, Art. no. 1698619, Jan. 2026. Accessed: Oct. 8, 2026. [Online]. Available: https://www.frontiersin.org/journals/conservation-science/articles/10.3389/fcosc.2025.1698619/full
8. [8]
   Washington Trails Association, "Lake Angeles," Washington Trails Association. Accessed: Oct. 8, 2026. [Online]. Available: https://www.wta.org/go-hiking/hikes/lake-angeles
9. [9]
   Washington Trails Association, "PJ Lake," Washington Trails Association. Accessed: Oct. 8, 2026. [Online]. Available: https://www.wta.org/go-hiking/hikes/pj-lake
10. [10]
    Wikipedia contributors, "Paternoster lake," Wikipedia. Accessed: Oct. 8, 2026. [Online]. Available: https://en.wikipedia.org/wiki/Paternoster_lake
11. [11]
    Wikipedia contributors, "Seven Lakes Basin," Wikipedia. Accessed: Oct. 8, 2026. [Online]. Available: https://en.wikipedia.org/wiki/Seven_Lakes_Basin
12. [12]
    U.S. Geological Survey, "Ozette Lake: A natural seismograph along the northern Cascadia Subduction Zone (video)," U.S. Geological Survey. Accessed: Oct. 8, 2026. [Online]. Available: https://www.usgs.gov/programs/cmhrp/news/ozette-lake-a-natural-seismograph-along-northern-cascadia-subduction-zone-video
13. [13]
    Wikipedia contributors, "Lake Cushman," Wikipedia. Accessed: Oct. 8, 2026. [Online]. Available: https://en.wikipedia.org/wiki/Lake_Cushman
14. [14]
    Wikipedia contributors, "Lake Crescent," Wikipedia. Accessed: Oct. 8, 2026. [Online]. Available: https://en.wikipedia.org/wiki/Lake_Crescent
15. [15]
    Hiking Project, "Lena Lake," Hiking Project. Accessed: Oct. 8, 2026. [Online]. Available: https://www.hikingproject.com/trail/7001958/lena-lake
16. [16]
    Wikipedia contributors, "Mount Storm King," Wikipedia. Accessed: Oct. 8, 2026. [Online]. Available: https://en.wikipedia.org/wiki/Mount_Storm_King
17. [17]
    Wikipedia contributors, "Kettle (landform)," Wikipedia. Accessed: Oct. 8, 2026. [Online]. Available: https://en.wikipedia.org/wiki/Kettle_(landform)
18. [18]
    T. Abbe, J. Bountry, G. Ward, L. Piety, M. McBride, and P. Kennard, "Forest influence on floodplain development and channel migration zones," presented at Geol. Soc. Amer. Annu. Meeting, Seattle, WA, USA, Nov. 2003. Accessed: Oct. 8, 2026. [Online]. Available: https://gsa.confex.com/gsa/2003AM/webprogram/Paper64592.html
19. [19]
    Washington State Recreation and Conservation Office, "Allen's Marsh / Old Joe's Slough fish passage," Salmon Recovery Portal, project 15-1292. Accessed: Oct. 8, 2026. [Online]. Available: https://srp.rco.wa.gov/project/100/40207
20. [20]
    I. S. Evans and N. J. Cox, "Size and shape of glacial cirques: comparative data in specific geomorphometry," in Proc. Geomorphometry 2015, Poznań, Poland, Jun. 2015. Accessed: Oct. 8, 2026. [Online]. Available: https://geomorphometry.org/uploads/pdf/pdf2015/EvansCox2015geomorphometry.pdf
21. [21]
    Wikipedia contributors, "Shoreline development index," Wikipedia. Accessed: Oct. 8, 2026. [Online]. Available: https://en.wikipedia.org/wiki/Shoreline_development_index
22. [22]
    Wikipedia contributors, "Lake Ozette," Wikipedia. Accessed: Oct. 8, 2026. [Online]. Available: https://en.wikipedia.org/wiki/Lake_Ozette
23. [23]
    LakeLubbers, "Lake Quinault, Washington, USA," LakeLubbers. Accessed: Oct. 8, 2026. [Online]. Available: https://lakelubbers.com/?p=7003
24. [24]
    Washington Trails Association, "Hoh Lake," Washington Trails Association. Accessed: Oct. 8, 2026. [Online]. Available: https://www.wta.org/go-hiking/hikes/hoh-lake
25. [25]
    Washington Department of Fish and Wildlife, "Constance," High lakes. Accessed: Oct. 8, 2026. [Online]. Available: https://wdfw.wa.gov/fishing/locations/high-lakes/constance
26. [26]
    Wikipedia contributors, "Tarn (lake)," Wikipedia. Accessed: Oct. 8, 2026. [Online]. Available: https://en.wikipedia.org/wiki/Tarn_(lake)
27. [27]
    Wikipedia contributors, "Cameron Glaciers," Wikipedia. Accessed: Oct. 8, 2026. [Online]. Available: https://en.wikipedia.org/wiki/Cameron_Glaciers
28. [28]
    Wikipedia contributors, "Barnes Point," Wikipedia. Accessed: Oct. 8, 2026. [Online]. Available: https://en.wikipedia.org/wiki/Barnes_Point
29. [29]
    D. Seekell, B. Cael, E. Lindmark, and P. Byström, "The fractal scaling relationship for river inlets to lakes," Geophys. Res. Lett., vol. 48, no. 9, Art. no. e2021GL093366, May 2021. Accessed: Oct. 8, 2026. [Online]. Available: https://nora.nerc.ac.uk/id/eprint/530379/
30. [30]
    D. Seekell, B. B. Cael, and P. Byström, "Problems with the shoreline development index: A widely used metric of lake shape," Geophys. Res. Lett., vol. 49, no. 10, Art. no. e2022GL098499, May 2022. Accessed: Oct. 8, 2026. [Online]. Available: https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2022GL098499
31. [31]
    B. B. Cael and D. A. Seekell, "The size-distribution of Earth's lakes," Sci. Rep., vol. 6, Art. no. 29633, Jul. 2016. Accessed: Oct. 8, 2026. [Online]. Available: https://pmc.ncbi.nlm.nih.gov/articles/PMC4937396/
32. [32]
    B. B. Cael and D. A. Seekell, "Corrigendum: The size-distribution of Earth's lakes," Sci. Rep., vol. 7, Art. no. 42155, Feb. 2017. Accessed: Oct. 8, 2026. [Online]. Available: https://pmc.ncbi.nlm.nih.gov/articles/PMC5304202/
33. [33]
    L. A. Winslow, J. S. Read, P. C. Hanson, and E. H. Stanley, "Lake shoreline in the contiguous United States: Quantity, distribution and sensitivity to observation resolution," Freshw. Biol., vol. 59, no. 2, pp. 213-223, 2014. Accessed: Oct. 8, 2026. [Online]. Available: https://pubs.usgs.gov/publication/70048525