A global GIS- and literature-based analysis of the altitude and relief of more than 14,000 ice-free glacial cirques in 56 study areas ranging from 69°N to 77°S shows that average cirque floor altitudes vary with latitude in a zone centered on the Quaternary average equilibrium line altitude (QA-ELA). In addition, relief above cirque floors averages 346 ± 107 m, and is rarely >600 m. In regions where ice-free cirques are abundant, cirque relief is independent of modern precipitation rates, tectonic regime, and relative peak height, limiting peaks to within a fixed distance above the QA-ELA and thus causing them to also parallel the ELA. We propose a physical process model where, under suitable geologic and tectonic conditions, climate exerts a first-order control on mountain range height. In this model, glacial erosion excavates a zone of cirques with floors centered on the QA-ELA, while climate-driven weathering processes on adjacent slopes limit the peaks to within a few hundred meters above this zone.


Glacial cirques are erosional landforms characteristic of alpine glaciation. While the shape and growth of cirques have been studied extensively for decades (e.g., Olyphant, 1981; Evans, 2006; Sanders et al., 2012), there is no global analysis of glacial cirque altitude or relief to date. With evidence mounting that climate exerts a first-order control on the height of mountain ranges worldwide (Egholm et al., 2009; Pedersen et al., 2010; Champagnac et al., 2012), we explore links between cirque altitude and relief, and the relationships between climate, glacial erosion, and mountain height.

Studies have long demonstrated a relationship between cirque floor altitude and glacial equilibrium line altitudes (ELAs) from valley to mountain-range scale (e.g., Flint, 1957; Porter, 1989). Measuring the altitude of cirque floors is even one method of estimating past glacial ELAs (Meierding, 1982; Benn and Lehmkuhl, 2000); Porter (1989) proposed and others have confirmed that cirque floors form in a zone centered around the local Quaternary average ELA (QA-ELA), suggesting a cumulative effect of glacial and non-glacial erosion through multiple glacial cycles (Mitchell and Montgomery, 2006; Foster et al., 2008; Anders et al., 2010). And while some studies on glacial cirque distributions imply that cirques are secondary features that form some distance below peaks (e.g., Hassinen, 1998; Benn and Lehmkuhl, 2000), Anders et al. (2010) argued that if peak height controlled cirque floor altitude from an ice accumulation perspective, areas with high annual precipitation would develop glaciers with a smaller accumulation area and result in cirque headwalls with less relief than in areas with low precipitation.

Without specifically considering cirques, other studies argue that climate, via glaciation, limits the height of mountain ranges (e.g., Brozović et al., 1997; Egholm et al., 2009; Pedersen et al., 2010; Champagnac et al., 2012). This “glacial buzzsaw” hypothesis cites observations that peak altitudes or other topographic measures such as hypsometric maxima have distinct relationships with the ELA, despite variation or gradients in uplift, erosion, or lithology (Brozović et al., 1997). For example, Egholm et al. (2009) confirmed that mountain range heights worldwide tend to correlate with snowline altitudes, and their erosion model predicts the observed hypsometric maxima just below the ELA. Champagnac et al. (2012) used a different approach to show that shortening rate fails to account for at least 25% of mountain range relief, and suggested that climate accounts for the difference.

We use the largest compilation of cirque altitude and relief data from mostly ice-free mountains to date to investigate the relationships between the QA-ELA, cirque floor altitudes, cirque relief, and peak altitudes. We propose a model linking climate to topography that applies to the many mountain ranges around the world currently featuring ice-free cirques. In these locations, cirque floors form in a zone centered on the QA-ELA, and a variety of erosion and weathering processes contribute to limiting the height of the peaks surrounding the cirques, resulting in peak altitudes that also parallel the QA-ELA.


We collected and analyzed cirque floor altitude, headwall relief, and peak height data from 31 study areas; we supplemented this data set with cirque altitude and relief data from 25 published analyses (see the GSA Data Repository1, and Fig. DR1 and Table DR1 therein), totaling 56 study areas (Fig. 1). Three regions spanning >3° latitude were divided into 1° to 2° bins and analyzed separately. For all study areas, we classified the tectonic regime into four categories (active convergent margin, passive margin, continental interior, and volcanic). We also determined the modern average annual precipitation for each range crest based on summed monthly 10-arc-minute normals for ca. A.D. 1950–2000 from WorldClim (www.worldclim.org; Hijmans et al., 2005).

Because local ELA data for our study areas are in many cases unavailable, we compared our cirque floor altitudes with an “average” ELA derived from the pole-to-pole Last Glacial Maximum (LGM) and modern “snowline” transects along the East Pacific published by Broecker and Denton (1989). While snowline data may have been derived using topography-based methods in valleys containing cirques—and therefore the cirque floor altitudes may indirectly influence the results—it is unlikely that these snowlines were measured directly from cirque floor altitudes. We define the Quaternary average East Pacific ELA (which we refer to as the QAEP-ELA) simply as the midpoint between the modern and LGM East Pacific snowline at each degree of latitude (Porter, 1989). While the local QA-ELA for many of our study locations likely differs from the QAEP-ELA, and the true QA position of the East Pacific ELA might not be precisely at the midpoint between the modern and LGM snowlines, we believe this QAEP-ELA serves as an acceptable proxy for showing first-order relationships between a global QA-ELA and our cirque data (Porter, 1989).

We located >5000 cirques from 31 study areas primarily using digital topographic data available from the U.S. Geological Survey (Gesch et al., 2002). We defined cirques as landforms with a curved, steep headwall and a flat or overdeepened bottom, and that occupy a valley head (Evans and Cox, 1995; Mitchell and Montgomery, 2006). Many cirques in this analysis contain a tarn. We only included cirques with a clearly defined, distinct headwall rising to a drainage divide; in most cases these were the highest overdeepenings in the valley (Fig. DR1). Because we needed to measure the altitude of cirque floors, we excluded basins containing substantial glacial ice. Our analysis therefore generally excludes exceptionally high mountain ranges whose peaks extend far above the modern ELA, the implications of which we explore in the Discussion section. We determined the altitude, latitude, and longitude of each cirque outlet or tarn surface with a 30 m or higher-resolution digital elevation model (U.S. Geological Survey National Elevation Dataset; ned.usgs.gov) except where noted. We measured individual cirque relief for >6200 cirques in 34 study areas by subtracting the cirque floor (tarn or outlet) altitude from the highest point of each cirque’s contributing drainage area. We then calculated the means and standard deviations of latitude, cirque floor altitude, adjacent peak altitude, and cirque relief.

We also compiled published cirque floor altitude, relief, and adjacent peak height data from 25 additional study areas (17 publications, n > 9000 cirques). We limited these studies to ones where all cirques within a geographic area were included, the definitions of “cirque” and “cirque floor” were similar to ours, and the number of cirques and the average cirque floor altitude were presented. We have complete data sets of individual cirque altitude, relief, and peak height from two of the published studies (Mitchell and Montgomery, 2006; Anders et al., 2010), but for most studies, only averages (means) were reported. In some cases, we calculated mean relief or adjacent peak altitude using the data presented. We determined the centroid latitude and longitude of published cirque studies using location maps or other information from the publication.


Mountain ranges with abundant, ice-free cirques are found on all continents except Australia and at latitudes ranging from 69°N to 77°S, and are located in all four types of tectonic regimes (Fig. 1). The number of cirques in each study area ranges from 5 to >3500. Modern annual precipitation values range from ∼100 mm/yr to >3000 mm/yr, with a mean annual precipitation of ∼1100 mm/yr.

Mean cirque floor altitudes range from ∼440 to 4380 m.a.s.l. (m above sea level), and are normally distributed around the East Pacific ELA (Figs. 2A and 2B). The cirque altitude for the study location in Antarctica (77°S) is the most striking outlier; however, the actual QA-ELA in that location is likely hundreds of meters higher than the QAEP-ELA (Fountain et al., 2006). Correlations between the residual of mean cirque altitude and QAEP-ELA with precipitation, peak altitude, and tectonic setting are weak (see Figs. DR2–DR6).

Mean cirque relief for all study areas is 346 ± 107 m (mean ± 1σ), and only one study area has a mean cirque relief of >600 m (Fig. 3A). The relief of individual cirque basins has a similar distribution; average cirque relief is 382 ± 150 m (n = 6249), and 92% of cirques have relief of <600 m (Fig. 3B). Correlations between mean cirque floor relief and precipitation, peak altitude relative to the QAEP-ELA, and tectonic regime are weak (Figs. 3B–3D).

The average altitudes of peaks adjacent to ice-free cirque basins range from 600 to 4970 m.a.s.l. depending on latitude, and the majority (78%) of peaks are <600 m above the East Pacific ELA at each latitude (Fig. 2C). The mean cirque basin relief (∼350 m) is very similar to the mean vertical difference between the QAEP-ELA and the modern East Pacific ELA (∼400 m), and the peaks generally parallel but do not exceed the modern East Pacific ELA. Because our data set is limited to mountain ranges with ice-free cirque basins, any cirques adjacent to peaks rising well above the modern ELA would likely be ice covered and therefore excluded from our analysis.


Our major findings are that cirque floors are normally distributed about the QAEP-ELA and that cirque relief is limited to a few hundred meters despite variation in location, tectonic setting, position of peaks relative to the East Pacific ELA, and modern precipitation (Fig. 2). The average altitude of peaks adjacent to ice-free cirques parallels the QAEP-ELA but is 346 ± 107 m higher. We propose a process model to explain these relationships.

The key element of this process is the excavation of cirques over time via scouring and headwall retreat, resulting in floors at or near the QAEP-ELA. Because cirque headwall relief is not significantly affected by precipitation, height of peaks relative to the QAEP-ELA, or tectonic regime, we reason that temperature controls the altitude of cirque floors and thus peak height. For example, if cirque relief was controlled by accumulation area rather than the ELA, high precipitation rates might be correlated to lower cirque relief because less accumulation area would be needed to develop an erosive glacier (Anders et al., 2010). Similarly, cirques on peaks rising higher above the QAEP-ELA might also be expected to have less relief. We see no such relationships and conclude instead that the ELA-controlled zone of cirques forms a base level above which peaks rarely rise more than 600 m. While consistent with other studies indicating that hypsometry decreases rapidly above modern ELAs (Egholm et al., 2009; Pedersen et al., 2010), our results specifically relate cirque formation and peak height to an average ELA rather than modern or LGM ELAs.

Why cirque floors should occur at the QAEP-ELA and why relief is ∼350 m are not directly answered by our data. However, several studies describing enhanced weathering and erosion in alpine regions may explain how climate influences these observed results. Sanders et al. (2012) showed how rock fracturing on cirque headwalls and within the bergschrund of a cirque glacier via ice segregation may drive headwall retreat. Headwall erosion in this manner is most effective when ice is present in the basin. Frost cracking within certain temperature—and thus altitude—windows may also drive erosion rates comparable to uplift rates (e.g., Walder and Hallet, 1985; Hales and Roering, 2009). Temperatures in this range also occur during winter months of interglacials, and thus can drive erosion even when glaciers are not present. Furthermore, the steep and fractured landscape above cirque floors becomes preferentially susceptible to mass-wasting processes such as landsliding (e.g., Moon et al., 2011). We speculate that the fluctuation between glacial and interglacial phases facilitates erosion; weathering and ice-related processes continue to weaken and remove rock on the steep headwalls during interglacial times, while ice effectively attacks the headwalls and removes debris during glaciations (Sanders et al., 2012). High rock-fracture density can also lead to locally high subglacial erosion rates (e.g., Dühnforth et al., 2010), so perhaps rock fracturing at and above the QA-ELA facilitates headwall retreat and mass-wasting processes on the steep cirque headwalls. If fracture formation, subglacial erosion, and mass wasting are all important processes, extremely durable lithologies or slopes covered with permanent or frozen-bed ice may not experience the same extent of cirque excavation. This hypothesis implies that climate is a primary control on cirque floor altitude and thus peak altitude in these locations, and that cirque erosion is not limited to times when the basin is occupied by a cirque glacier.

The development of cirques at the long-term ELA, and thus the primacy of climate in limiting mountain height, is common but not ubiquitous. Several mountain ranges have been shown to not fit the glacial buzzsaw model due to factors such as high uplift rates, durable lithology combined with steep slopes (e.g., “teflon peaks”), or frozen-bed glaciers (e.g., Thomson et al., 2010; Ward et al., 2012; Godard et al., 2014). The high regions of such ranges, not included in our study, extend above the QAEP-ELA, and any cirques present were probably ice covered throughout the Quaternary. We hypothesize that any cirques on these exceptionally high peaks are unlikely to have the same altitude and relief traits of those where ice cover is intermittent, because the role of physical weathering is limited. If cirques do form at the QAEP-ELA on these slopes, by definition they will have more relief than those in our analysis. While the highest mountain ranges worldwide likely prove an exception to our cirque-driven glacial buzzsaw model, the frequency and global distribution of mountain ranges with ice-free cirques contribute to a global glacial buzzsaw effect.


Absent other factors such as rapid rock uplift or unusually durable lithology, ice-free cirques worldwide are located at altitudes centered on the QAEP-ELA. Because cirque relief is limited, adjacent peaks in cirque-rich mountain ranges rise only 346 ± 107 m above the QA-ELA. These observations suggest that climate-driven weathering and erosion processes are primary factors in cirque formation and limit the heights of mountain peaks adjacent to cirques.

This work was supported by the College of the Holy Cross (Massachusetts, USA), an Ardizzone Fellowship (Mitchell), and a Summer Research Scholarship from Herman R. and Mary Charbonneau (Humphries). We thank N. Shea for GIS data collection, R. Bertin for statistical analysis assistance, and S. Brocklehurst and three anonymous reviewers for their helpful comments.

1GSA Data Repository item 2015028, summary cirque altitude, relief, peak altitude, East Pacific equilibrium line altitude (ELA), and shortening rate data, and a description of the statistical analysis of residuals for the correlation between cirque altitude and East Pacific ELA, is available online at www.geosociety.org/pubs/ft2015.htm, or on request from editing@geosociety.org or Documents Secretary, GSA, P.O. Box 9140, Boulder, CO 80301, USA.