1 When and how did deglaciation of Penobscot Bay occur? These are the questions we address in this paper. One question, in particular, motivated this study: was there a calving bay?

2 The reasonable suggestion that calving bays developed in Penobscot Bay (Fig. 1) and further north in the Penobscot Lowland as the Laurentide Ice Sheet retreated across coastal Maine probably originated with Terry Hughes. It was first mentioned in print by Borns and Hughes (1977, p. 204), who refer to “The lowlands of central Maine…,” which we infer to mean Penobscot Bay and the Penobscot Lowland, and who stated, without presenting evidence, that a calving bay developed in these lowlands. Lowell (1994), on the other hand, argued that the water was not deep enough to yield high calving rates, and also that access of the sea to the Lowland was through “…two narrow (<5 km wide) channels…” that would have “…prevented the invasion of an effective calving bay into Maine.”

3 Detailed mapping of striations near Bangor (see Fig. 1 for location) has demonstrated convincingly, however, that ice flow, while initially to the south, changed dramatically during the last phases of retreat. To the east of the Penobscot River, crag-and-tail features indicate that ice flow became westerly, while to the west, flow became easterly (Syverson and Thompson 2008; Thompson and Syverson 2008). The width of the zone of converging flow is <10 km, so the actual “bay” was confined to the Penobscot River valley, which is slightly less than a kilometre wide here, and <30 m deep. There remains the question of whether a calving bay developed further south, where the bay is wider and deeper (Fig. 1). One of our goals in undertaking this study was to determine whether De Geer moraines, visible on recent Light Detection and Ranging (LiDAR) imagery, provided support for the presence of a calving bay there. The LiDAR imagery we used extends from ~20 km west of the bay ~160 km eastward, and from the present coast northward to ~7 km south of Bangor, a mean distance of ~50 km (Fig. 1).

4 Another goal was to date the northward retreat of the ice sheet in Penobscot Bay and across the landscape east of the bay. To do this, we needed dates on some of the moraines. Two moraines have been dated: the Pond Ridge and Pineo Ridge moraines. The western ends of both of these lie within the study area (Fig. 1).

5 In the course of our study of the morainal landscape just south of Pineo Ridge, additional questions arose regarding the timing and character of the Laurentide Ice Sheet retreat and likely subsequent slight readvance that deposited Pineo Ridge.

6 Pineo Ridge was first described by Stone (1899, p. 111–112). Stone called Pineo a gravel plain, not a moraine, although he noticed till in the vicinity. Borns (1966) was apparently the first to recognize that Pond Ridge was a moraine and that Pineo Ridge was a delta-moraine complex. Their ages have been a topic of study ever since.

7 The earliest age estimates are radiocarbon dates, all of which we have calibrated (or recalibrated) using the most recent version of CALIB 7.1 (Stuiver et al. 2019). A ¹⁴C date on seaweed associated with the Pond Ridge moraine, 15.1 ± 0.4 ka was obtained by Stuiver and Borns (1975). More recently, Dorion et al. (2001) report three dates related to the Pond Ridge moraine. The one most closely associated with formation of the moraine, 15.9 ± 0.1 ka (Fig. 2), was on a Nucula sp. shell from glaciomarine sediments interbedded with coarse ice-proximal units on the moraine’s distal side. A maximum age of 16.1 ± 0.1 ka was obtained from a Macoma calcarea shell closely associated with ice-proximal grounding-line deposits 3 km outside the moraine. A min-imum age of 15.2 ± 0.1 ka was obtained from shells in gla-ciomarine sediments on the proximal side of the moraine. Two minimum ages for Pineo Ridge moraine, 15.2 ± 0.1 ka (Kaplan 1999) and 15.1 ± 0.1 ka (Dorion et al. 2001; Fig. 2), were obtained from shells in lake sediments a few kilometres inside the moraine.

8 As these dates were from marine organisms, they required a marine reservoir correction. Reservoir corrections vary both spatially and temporally. Along the perimeter of the Gulf of Maine, McNeely et al. (2006) report 22 values ranging from 430 to 730 years. By comparing dates on shells and on associated logs in the Presumpscot Formation near Portland, Maine, Thompson et al. (2011) inferred a reservoir age of 1000 years. The mean of these 23 estimates is 587 years. We rounded this up to 600 years when calibrating the above dates. This value was also used by Schnitker et al. (2001) and by Borns et al. (2004).

9 ¹⁰Be exposure ages have also been obtained on the Pineo Ridge moraine. Hall et al. (2017) dated twelve boulders on the proximal side of the moraine (Fig. 2). Three, ranging in age from 16.2 to 17.0 ka, were deemed outliers; the remaining nine yielded a peak age of 15.3 ± 0.7 ka on a “camel” plot. The uncertainty includes uncertainty in the production rate. Koester et al. (2017) dated seven boulders on a moraine projecting southward from beneath the Pineo delta, yielding an age of 14.5 ± 0.7 ka. As the moraine would have been below sea level when the Pineo delta formed and for some time thereafter, the young age is likely due to shielding by water (Hall et al. 2017).

10 In the Greenland Ice Core Project (GRIP) ice core, there are two dips to more negative δ¹⁸O values during the Older Dryas (Fig. 3), one at 16.1 ± 0.2 ka and a larger one at 15.7 ± 0.2 ka (GICC05 chronology from Andersen et al. 2006). The ages are based on multi-parameter counting of annual layers (Andersen et al. 2006) and the uncertainties are based on the maximum counting error and are approximately ±2σ. These dips reflect brief periods of colder climate.

11 Between 17.8 and 16.7 ka, the mean δ¹⁸O in the GRIP core was -40.8‰, and between 16.1 and 15.5 ka it was -42.4‰, a change of 1.6 ‰ (Fig. 2). Such a change corresponds to a decrease in mean annual temperature of ~5°C at the core site (Buizert et al. 2014; Fig. 3). In ice cores from central and northern Greenland, such temperature changes are believed to largely reflect changes in winter temperatures (Denton et al. 2005; Buizert et al. 2014). The change in summer tempera-ture is only about one third of that in winter (Denton et al. 2005, p. 1169). Thus, the 5°C decrease in mean annual tem-perature likely resulted from a ~2.5°C decrease in summer temperature and a ~7.5°C drop in winter temperature; the annual temperature range would thus have increased ~5°C.

12 As ablation is closely related to the number of positive degree days (e.g., Braithwaite 1995), we next investigate the change in positive degree days implied by this cooling. We approximate the annual temporal temperature cycle by a sinusoidal function, Display large image of Equation 1 where θ is the temperature, θ_m is the mean annual temperature, R is half the annual temperature range, and tis the time in days. The number of positive degree days, D_d, is then: Display large image of Equation 2 where Display large image of Equation 3 is the day when the temperature first rises above 0°C and Display large image of Equation 4 is the maximum summer temperature. Carrying out the integration yields: Display large image of Equation 5

13 Let’s first choose R = 13°C to roughly match current conditions along the coast in both Maine and northwest Green-land and calculate D_d for a series of reasonable values of θ_m. We then decrease θ_m by 5°C and increase R by 2.5°C in accord with the above analysis and calculate D_d for another series of possible values of θ_m. Finally, we calculate the percentage change in D_d for each choice of θ_m (Fig. 4). For example, on the Maine coast between 17 and 16 ka, a 5°C decrease in θ_m from -5°C to -10°C, resulting from a 2.5°C decrease a summer temperature, a 7.5°C decrease in winter temperature, and a 2.5°C increase in R, would result in a 49% reduction in positive degree days (light dashed line in Fig. 4).

14 On the Maine coast at 17 ka, θ_m was likely between -1°C and -8°C. The former is based on an analysis of geomorphic features indicative of permafrost (Hooke and Fastook 2007, p. 649), and the latter is obtained by adding the present mean annual temperature difference between the GRIP site and coastal Maine, 39°C, to the 17 ka GRIP temperature (~-47°C). Thus, a 5°C cooling could have reduced D_d by between 37% and 68%. This would have had an appreciable effect on the ablation rate. Cooling episodes like the Oldest Dryas (Fig. 3) are believed to be accompanied by decreases in the strength of the Atlantic meridional overturning circulation (AMOC), and thus in an increase in sea ice in the North Atlantic (Denton et al. 2005, p. 1175). This, too, would have inhibited melting along the margin of the ice sheet in the Gulf of Maine. Thus, we think it likely that these cold events resulted in pauses in retreat of the ice sheet, or possibly slight readvances, and that they were responsible for the Pond Ridge and Pineo Ridge moraines. Supporting this interpretation is the observation that the duration and depth of the second of these coolings was significantly greater than that of the first, consistent with the appreciably larger size of the Pineo Ridge moraine complex in comparison with the Pond Ridge moraine. As shown in Figure 2, the dates of these cold events are generally slightly older than the relevant ¹⁴C and exposure age dates, although within the limits of uncertainty of some of the latter.

15 Hall et al. (2017) objected to correlating the Pineo and Pond moraines with these cold episodes. They argue that although deglaciation in Maine was characterized by “rapid ice retreat followed by a (sic) short-lived stillstand/readvance” during the Oldest Dryas, the isotope record “shows extremely cold conditions throughout” this time period. Thus, they contend that the retreat and the two stillstands or readvances could not be due to changes in air temperature. The reference Hall et al. (2017) cite for their temperature record (Alley 2000), shows data only going back to 16 ka. Based on the GRIP δ¹⁸O record (Fig. 3), the warming event that likely initiated retreat of the Laurentide Ice Sheet from its LGM position at the edge of the continental shelf, began at ~30 ka, although the actual retreat of the ice margin did not start until ~22 or 23 ka (e.g., Schnitker et al. 2001). The period from 22 ka to ~17 ka was one of gradual warming under the influence of increasing insolation (Fig. 3; Buizert et al. 2014; Koester et al. 2017); at 17 ka, when cooling started despite a continued increase in insolation (Fig. 3), the ice sheet was likely not fully adjusted to the prevailing climate. Thus, it would have continued to retreat, but would have been ripe for pauses (or slight readvances) at the times of the two cold periods centered at 16.1 and 15.7 ka.

16 In short, the 16.1 ± 0.2 and 15.7 ± 0.2 ka ages for the Pond Ridge and Pineo Ridge moraines based on the Greenland δ¹⁸O record are consistent with, but better supported than, either the ¹⁴C ages, which are prone to various uncertainties, not the least of which is the marine reservoir correction, or the ¹⁰Be exposure ages with their uncertainties. Construction of these major moraines, however, took some time. Thus, deposition of the Pond and Pineo moraines undoubtedly began prior to 16.1 and 15.7 ka, respectively, and retreat from them later than these dates.

17 Borns (1967) noted numerous small NE-trending moraines south of Pineo Ridge. Recent LiDAR imagery has revealed even more of these moraines, trending more easterly, between Pineo Ridge and Penobscot Bay. These are now recognized as De Geer moraines (Fig. 5), deposited along the grounding line of a rapidly retreating ice sheet terminating in the sea (De Geer 1940; Larsen et al. 1991; Sinclairet al. 2018). It is commonly argued that De Geer moraines form approximately annually, and we adopt this view: an ice sheet advances slightly during the winter, bulldozing till into a low ridge, and then retreats rapidly during the summer, resulting in a space of a few tens to a couple of hundred metres between successive moraine ridges (e.g., Eusden 2014; Sinclair et al. 2018). The mean distance between these small moraines is generally consistent with the best estimates of the rate of retreat of the ice margin across coastal Maine (Hooke and Hanson 2017, p. 294).

19 De Geer moraines are readily visible on high-resolution LiDAR imagery (e.g., Fig. 5). The LiDAR map layers were uploaded to ArcMap 10.4.1 software using the ArcGIS online feature. To illuminate the topography, all hillshade data used a 315° false sun, shining approximately perpendicular to the moraines. The crests of nearly all moderately continuous moraines were mapped.

20 To correlate moraines on opposite sides of areas devoid of moraines, such as bodies of water, topographic highs, and some topographic lows, we located a well-defined, continuous moraine above or below the area lacking moraines and counted the number of moraines north or south of this moraine on both the east and west sides of the barren area. This yielded high-confidence correlations between prominent moraines across the barren areas, especially across the Penobscot Bay.

21 Our westward continuation of the Pond Ridge moraine (Fig. 6) is based on counting the approximate number of moraines between the Pond Ridge and Pineo Ridge moraines. This involved some uncertainty because the morainal sequence south of Pineo is lobate and erratic. The moraine we believe to be the westward continuation of the Pond Ridge moraine was roughly the same distance from Pineo as the Pond Ridge moraine is in its type locality, and is of approximately the same size and character as the type Pond Ridge moraine.

22 In some cases, apparently continuous sequences of De Geer moraines occur between two prominent moraines. In these cases, the time elapsed between deposition of the prominent moraines can be determined reasonably accurately simply by counting the intervening De Geer moraines. More commonly, however, we did not find continuous sequences of mappable De Geer moraines that covered the entire distance between prominent moraines. In these cases, the time elapsed between deposition of the prominent moraines was estimated by calculating several retreat rates based on short sequences of De Geer moraines between the prominent moraines. Using these retreat rates, we estimated the time elapsed between deposition of the prominent moraines. By accumulating these estimates, we calculated the time elapsed since deposition of our extrapolation of the Pond Ridge moraine. Most prominent moraines likely represent less than, say, five years of accumulation (e.g., Fig. 5). As our estimates of moraine ages (Fig. 6) are to the nearest century, we did not make any adjustment for the additional time required to accumulate a prominent moraine.

23 Flow directions west and east of Penobscot Bay were determined by averaging the orientation of numerous normals to moraines within 5 × 5 km areas. Two areas were studied east of the bay, and three west of the bay (Fig. 7 and Table 1). Areas were chosen based on their proximity to the bay and the abundance of mappable moraines.