Spatial variation of hydroclimate in north-eastern North America during the last millennium

Climatic expressions of the Medieval Climate Anomaly (MCA) and the Little Ice Age (LIA) vary regionally, with reconstructions often depicting complex spatial patterns of temperature and precipitation change. The characterisation of these spatial patterns helps advance understanding of hydroclimate variability and associated responses of human and natural systems to climate change. Many regions, including north-eastern North America, still lack well-resolved records of past hydrological change. Here, we reconstruct hydroclimatic change over the past millennium using testate amoeba-inferred peatland water table depth reconstructions obtained from fifteen peatlands across Maine, Nova Scotia, Newfoundland and Qu ebec. Spatial comparisons of reconstructed water table depths reveal complex hydroclimatic patterns that varied over the last millennium. The records suggest a spatially divergent pattern across the region during the Medieval Climate Anomaly and the Little Ice Age. Southern peatlands were wetter during the Medieval Climate Anomaly, whilst northern and more continental sites were drier. There is no evidence at the multi-decadal sampling resolution of this study to indicate that Medieval mega-droughts recorded in the west and continental interior of North America extended to these peatlands in the north-east of the continent. Reconstructed Little Ice Age hydroclimate change was spatially variable rather than displaying a clear directional shift or latitudinal trends, which may relate to local temporary permafrost aggradation in northern sites, and reconstructed characteristics of some dry periods during the Little Ice Age are comparable with those reconstructed during the Medieval Climate Anomaly. The spatial hydroclimatic trends identified here suggest that over the last millennium, peatland moisture balance in north-eastern North America has been influenced by changes in the Polar Jet Stream, storm activities and sea surface temperatures in the North Atlantic as well as internal peatland dynamics. © 2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


Introduction
The last millennium is a key focus for palaeoclimate reconstruction and climate modelling because environmental boundary conditions are broadly comparable with the present day and detailed climate reconstructions can be obtained from wellresolved proxy data with robust chronological control. However, despite its critical effect on human and natural systems, understanding of hydroclimate variability over the Common Era is limited, in part due to a lack of spatially resolved palaeoclimatic data . Within the last millennium, the three most prominent periods of widespread Northern Hemisphere climatic change are: the Medieval Climate Anomaly (MCA), associated with increased temperature ca. 950e1250 CE (Jansen et al., 2007;Mann et al., 2008Mann et al., , 2009; the Little Ice Age (LIA), marked with cooler climate conditions ca.1400e1700 CE (Jansen et al., 2007;Mann et al., 2008Mann et al., , 2009) and recent warming (since 1850 CE). The MCA is often considered a partial analogue for current climate change because mean surface air temperatures, if not rates of change, were comparable with early 21st century warming. For example, reconstructed MCA mean surface temperatures in the North Atlantic region were 0.9e1.4 C above the 1961e1990 reference period (Mann et al., 2009). Conversely, the LIA was associated with mean Northern Hemisphere surface temperature decreases of ca. 0.5 C (Mann et al., 2009), which strongly impacted natural ecosystems and disrupted human society (Lamb, 1977;Grove 1988). The MCA and LIA predate the Industrial Revolution; therefore, they act as excellent case studies to further understand natural ecosystem functioning and climate variability. Both the MCA and LIA have been described as global events (e.g. Mann et al., 2009;Graham et al., 2010;Lüning et al., 2017), but the timing and extent of these periods, as well as their hydroclimatic effects, are regionally variable. For example, whilst extensive MCA droughts have been reconstructed across the Western and Midwestern US (e.g. Cook et al., 2010), the eastward extent of these events remains unclear, partly due to the lack of well-constrained late Holocene hydroclimatic records developed from northeastern US and Atlantic Canada (Shuman et al., 2018). The timing of the LIA in north-eastern North America is also debated since tree-ring reconstructions from northern boreal Quebec demonstrate that cooling began later, in 1450 CE, and that the coolest phase occurred during the 19th century in many regions (Naulier et al., 2015). The majority of existing palaeoclimatic records from north-eastern North America have either; 1) a longer temporal focus (e.g. Newby et al., 2014;Neil and Gajewski, 2017;Blundell et al., 2018), with relatively low temporal resolution for the last millennium, or 2) have been restricted to the latitudinal extremities of the region in Newfoundland Daley et al., 2016) and Maine (Nichols and Huang., 2012;Booth, 2013, 2015). In order to improve our understanding of hydroclimate variability related to past climate change, the spatial coverage of palaeoclimate reconstructions along the eastern seaboard of the North Atlantic must be increased, particularly for the Medieval period (Neukom et al., 2018).
Ombrotrophic (rain-fed) peatlands are sensitive to climate change, with records of past variability in peatland water table depth (WTD) representing the prevailing balance between precipitation and evapotranspiration across the growing season (Charman et al., 2009). The responsiveness of these ecosystems to Holocene climate variability throughout north-eastern US and Atlantic Canada has been clearly demonstrated (e.g. Hughes et al., 2006;Nichols and Huang, 2012;Clifford and Booth, 2013;Clifford and Booth, 2015;Blundell et al., 2018). In addition, comparisons of regional records facilitate the distinction of hydrological changes driven by allogenic forcing (i.e. climate) from autogenic changes driven by internal peatland dynamics and ecohydrological feedback mechanisms (Belyea, 2009;Swindles et al., 2012;Morris et al., 2015). Whilst the characteristics of climatically-driven hydrological responses may vary between peatland sites, broadly synchronous changes in multiple sites can be attributed to climatic control (e.g. Magnan et al., 2018;Piilo et al., 2019).
The current climate of the north-eastern US and Atlantic Canada region is influenced by cold dry Arctic air masses, mild dry Pacific Westerlies, warm moist tropical maritime air masses and oceanic interactions between the Labrador Current and the Gulf Stream. The combined influence of these climatic controls generates longitudinal oceanic to continental, as well as latitudinal terrestrial temperature and precipitation gradients (Fig. 1). Whilst broad-scale climate variability is overlain by smaller scale local variability, east coast temperatures generally decrease with increasing latitude and, whilst latitudinal precipitation gradients have a more complex spatial pattern, generally precipitation decreases from southern boreal Qu ebec northward into the high boreal, subarctic and Arctic. Projections of future annual temperature changes in north-eastern North America suggest a 2e3 C increase by 2050 (Bruce, 2011;Wuebbles et al., 2017), with greater seasonal temperature increases projected for summer months compared with winter months (Wuebbles et al., 2017). Summer (June, July, August; JJA) precipitation-change projections over the same time period contain more complex spatial patterns, with median decreases of up to 10% in northern Nova Scotia and southern Newfoundland and increases of up to 10% in Maine and Qu ebec (Wuebbles et al., 2017).
The aim of this study is to improve understanding of hydroclimate variability during the last millennium in north-eastern US and Atlantic Canada by addressing the following questions: (1) Do the multi-decadal Medieval dry periods detected in the Western and Midwestern US extend to north-eastern North America? (2) Quaternary Science Reviews 256 (2021) 106813 What regional patterns in MCA and LIA climate change can be detected? (3) Do these regional patterns of climate change provide any insight into dominant climate forcing mechanisms? To address these questions, this study reconstructs summer hydroclimate variability over the past 1250 years from fifteen ombrotrophic peatlands across Qu ebec, Newfoundland and Nova Scotia, Canada and Maine, USA.

Study sites
The fifteen study sites are ombrotrophic peatlands located in Qu ebec, Newfoundland, Nova Scotia and Maine (Fig. 2, Table 1). These study sites are located across two K€ oppen-Geiger climate classification zones based on differences in summer temperatures (Kottek et al., 2006): all study sites in Qu ebec and the most northerly site in Newfoundland (Burnt Village Bog) are located in zone Dfc (snow, fully humid, cool summers and cold winters), whilst the other more southerly sites in Newfoundland (Nordan's and Jeffrey's) and those in Nova Scotia and Maine are located in zone Dfb (snow, fully humid, warm summers with threshold temperature ! þ10 C for at least four months) ( Table 1).
The seven new records developed for this study were obtained from locations on ombrotrophic peatlands with minimal anthropogenic disturbances and were distributed to address spatial gaps in the availability of late Holocene peatland hydroclimatic reconstructions. Whilst there is also a need for more peatland palaeohydrological reconstructions from other regions of the wider study area, such as southern Quebec, New Brunswick and Maine, this study primarily focused on addressing the data gap from more eastern and northern maritime climate areas of Nova Scotia and Newfoundland. One core from each study site was extracted from a lawn microform near the centre of each bog following the protocol of Barber et al. (1994) using an 11-cm-diameter Russian corer, following a stratigraphic investigation using the Tr€ oels-Smith (1955) system. Cores were stored in PVC tubes, wrapped in plastic film and stored at 4 C in preparation for analysis and long-term storage.

Chronologies
Published age-depth models for Burnt Village Bog, Nordan's Pond Bog, Petite Bog, Sidney Bog , Lac Le Fig. 2. Location of study sites that are used within this regional data comparison. Burnt Village Bog (BVB), Jeffrey's Bog (JRB), Framboise Bog (FBB), Petite Bog (PTB), Villagedale Bog (VDB) (black; unpublished sites, this study), Sidney Bog (SDY) and Saco Heath (SCH) (grey; unpublished sites, this study, and published testate-amoebae based peatland hydrological reconstructions (Clifford and Booth, 2013;Clifford and Booth, 2015)) and Lac Le Caron (  , were applied in this study (Fig. A1, Supplementary Information). Updated age-depth models were constructed for Saco Heath, Villagedale Bog, Framboise Bog and Jeffrey's Bog to combine 14 C measurements of Sphagnum stems with published tephrochronologies for the sites (Mackay et al., 2016, Fig. A2, Supplementary Information). Bayesian age-depth models were constructed using the R package "BACON" (Blaauw and Christen, 2011). Ages are reported as calibrated years before present (cal yr BP), with 0 cal yr BP corresponding to 1950 CE. Four time periods of interest have been selected to focus on climate changes during the MCA and LIA (as defined by Mann et al., 2009) and climatic conditions before and after these time periods, defined here as: pre-MCA (1250-1000 cal yr BP), MCA (1000-750 cal yr BP), LIA (550-250 cal yr BP) and post-LIA (150e0 cal yr BP).

Testate amoebae and reconstructed water table depths
Testate amoeba analysis is a well-established proxy in peatlands that is well-suited to reconstructing past hydroclimatic variability because species assemblages predominantly respond to changes in peatland surface moisture (e.g. Woodland et al., 1998;Mitchell et al., 2008). Within this study, testate amoeba analysis was completed across all cores at 2e4 cm intervals to equate to approximately 40-year resolution. Testate amoebae were extracted from 1 cm 3 subsamples following standard procedures (Hendon and Charman, 1997;Booth et al., 2010). At least 100 individual tests were identified following the recommendations of Payne and Mitchell (2009). Taxonomy followed Charman et al. (2000) and Booth (2008). Testate amoebae water table depth (WTD) reconstructions were obtained using the tolerance-downweighted weighted averaging model with inverse deshrinking (WA-Tol inv) from the North American transfer function of Amesbury et al. (2018). Reconstructed WTD values for all sites were then normalised for comparative purposes Amesbury et al., 2016). The variation in reconstructed WTD was assessed using the standard deviation of normalised WTD values (z-scores) within each site.

Historical climate data
Modelled historical climate data (1901e2010) were obtained for each site using the elevation-dependent ANUSPLIN modelling software (Hutchinson, 2004;McKenney et al., 2011) for comparison with recent reconstructed trends in testate-amoebae inferred peatland WTD. The modelled climate moisture index (CMI; sensu Hogg (1997)) represents effective precipitation, which is the principal climatic control of peatland WTDs (Charman et al., 2009). Spatial patterns of average CMI values have been modelled using inverse distance weighting interpolation (IDW power coefficient ¼ 2, resolution ¼ 2 decimal degrees).
The correlation between testate amoeba-inferred WTD reconstructions since 1901 and modelled CMI values for Northern Hemisphere summer months (JJA) over the same time period confirms that moisture availability is the primary driver of WTD values obtained from testate amoeba assemblages at the study sites and that the transfer function is capturing spatial variability in modern moisture content effectively (R 2 ¼ 0.80, P ¼ <0.001; Fig. 4). These correlations within the modern record justify the application of this transfer function further back in time. Further justification is provided by detrended correspondence analysis (DCA) of the full assemblage data for all sites, which shows that the species ordination along Axis 1 reflects an underlying wetness gradient based on known hydrological preferences of taxa (DCA Axis 1 gradient length ¼ 2.95 and eigenvalue ¼ 0.46; DCA Axis 2 gradient

Testate amoebae-based water table depth reconstructions
Testate amoebae-based WTD reconstructions indicate that on average the hydroclimate of the MCA was wetter relative to pre-MCA conditions at the southern end of the study region in Maine and southern Nova Scotia (Saco Heath, Sidney Bog and Villagedale Bog) and in sites on the north shore of the Gulf of St. Lawrence and the St. Lawrence Estuary (Plaine and Baie) (Fig. 5, Fig. 6A). Conversely, the MCA is associated with a change to a drier hydroclimate in sites in mid-northern Nova Scotia (Framboise Bog, Petite Bog), Newfoundland (Jeffrey's Bog, Nordan's Pond Bog) and James Bay Lowlands (Sterne) (Figs. 5 and 6A). Hydroclimatic conditions are more spatially variable during the LIA compared with the MCA (Fig. 6A, B). Generally, sites in northern Nova Scotia and Newfoundland were relatively wet or becoming wetter during the LIA whilst sites in Maine and many in Qu ebec were relatively dry or becoming drier (Fig. 5). Comparisons of spatial patterns of average LIA and MCA WTDs indicate that wetter MCA conditions were restricted to southern and northern areas of Nova Scotia (Villagedale Bog, Framboise Bog) and James Bay Lowlands (Mosaik, Sterne), whilst sites in Maine (Saco Heath, Sidney Bog) and Newfoundland   The Northern Hemisphere time periods of the MCA (1000-700 cal yr BP) and the LIA (550-250 cal yr BP), as defined by Mann et al. (2009), generally align with coherent WTD changes presented here (Fig. 5). Whilst WTD reconstructions from Framboise Bog suggest that MCA WTD changes begin earlier (ca. 1060 cal yr BP) and Jeffrey's Bog and Nordan's Pond Bog WTD trends extend to ca. 630 cal yr BP (Fig. 5), both of these dates are within the chronological errors of the ascribed Northern Hemisphere time periods for the MCA (±90 cal yr BP; Appendix A, Supplementary Item). General reconstructed WTD trends during the LIA persist beyond 250 cal yr BP in many sites (Saco Heath, Villagedale Bog, Framboise Bog, Morts, Plaine, Burnt Village Bog and Lac Le Caron). However, WTD in these sites deviate from LIA trends before ca. 190 cal yr BP, which is within chronological uncertainty for the Northern Hemisphere time period defined by Mann et al. (2009) and the later regional date of 100 cal yr BP for the end of the LIA in north-eastern North America (Naulier et al., 2015).

Spatial patterns of variability in reconstructed water table depth
Within site variability in reconstructed WTD was generally higher during the LIA (SD: 0.64) than any other time period (pre-MCA: SD 0.56; MCA: SD 0.57; post-LIA: 0.48) ( Fig. 7C and E). Compared with the MCA, these LIA elevated levels of WTD variation appear to be higher in sites in Maine, central Nova Scotia, southern Newfoundland and the James Bay region of Qu ebec ( Fig. 7B and C). However, there are no clear trends based on latitude, longitude or climate zone and a student t-test of WTD variability during the MCA and LIA indicates that there is no statistical difference in variability between time periods when all peat records are considered together (t ¼ 0.9, df ¼ 14, p-value ¼ À0.4). The post-LIA period is generally characterised by the lowest within-site variability in reconstructed WTD compared with the pre-MCA, MCA and LIA ( Fig. 7D and E). There is a latitudinal pattern of WTD variability in the post-LIA period (Fig. 7D), with highest values concentrated in the southern sites: regionally calculated average standard deviation values demonstrate the pattern, with sites in Nova Scotia and Maine registering two to three times the variability of sites in Newfoundland and Qu ebec. However, once again there are no statistical differences in WTD variability between post-LIA and LIA (t ¼ 1.5, df ¼ 14, p-value ¼ 0.2), post-LIA and MCA (t ¼ À0.9, df ¼ 14, p-value ¼ 0.4) and post-LIA and pre-MCA (t ¼ À0.3, df ¼ 14, p-value ¼ 0.8) when all peat records are considered together.

Modelled climate in north-eastern North America (1901e2010)
Historical climate data obtained from the ANUSPLIN modelling software (Hutchinson, 2004;McKenney et al., 2011) reveal a southnorth decreasing annual temperature trend across the study region (7 to À2.3 C; R 2 ¼ 0.93, P ¼ <0.001), which is also reflected in the average growing season length (226 -146 days; R 2 ¼ 0.95, P ¼ <0.001) and to a lesser extent in average annual precipitation (743e1264 mm; R 2 ¼ 0.68, P ¼ <0.001), although there is no significant relationship between latitude and precipitation (Fig. C1, Supplementary Information). The modelled climate moisture index (CMI; sensu Hogg (1997)) represents effective precipitation, which is the main climatic control of peatland WTD (Charman et al., 2009). There is no significant relationship between latitude and average modelled CMI during the Northern Hemisphere summer months (JJA) (Fig. 8; Appendix C, Supplementary Information). The CMI indicates that three of the four most southern sites (Saco Heath, Sidney Bog and Petite Bog) have experienced summer moisture deficits (from À1.6 to À0.2 cm) between 1901 and 2010, whilst summer precipitation exceeds evapotranspiration (from 1.1 to 4 cm) at all other sites within this study ( Fig. 8; Appendix C, Supplementary Information).

MCA hydroclimatic change (1000-700 cal yr BP/950e1250 CE)
North American palaeoenvironmental reconstructions often register drought events during the MCA (e.g. Cook et al., 2010;Ladd et al., 2018;Shuman et al., 2018) that extend from the west coast of the US to the Midwest (Booth et al., 2006(Booth et al., , 2012 and the mid-Atlantic states (Cronin et al., 2010;Oswald and Foster, 2011). Some of these events are defined as 'mega-droughts', which are events that persist beyond a decade . The most prominent dry period during the MCA, reconstructed from peatlands in the western Great Lakes area, occurred ca. 1000 cal yr BP, at the start of the MCA (Booth et al., 2006) and this has also been recorded in Mg/Ca ratios from ostracods and oxygen isotopes from benthic foraminifera preserved in marine records obtained from the US east coast in Chesapeake Bay (Cronin et al., 2010). Dry conditions at this time (within chronological errors of ca. ±100 years) are evident in WTD records from sites in Maine (Saco Heath), Nova Scotia (Petite Bog, Framboise Bog), Newfoundland (Burnt Village Bog, Nordan's Pond Bog) and Qu ebec (Baie, Lebel, Morts, Plaine, Lac Le Caron) within this study (Fig. 5). However, unlike peatland WTD records in the Great Lakes (Booth et al., 2006), the dry conditions reconstructed here (with the exception of Framboise Bog and Plaine; Fig. 5) are not pronounced and are recorded as smaller shifts relative to WTD lowering during the LIA and post-LIA period. There is, therefore, no clear evidence from multiple adjacent sites at the multi-decadal sampling resolution to support a pronounced 'drought event', centred ca. 1000 cal yr BP in terrestrial records from Maine, Nova Scotia, Newfoundland or Qu ebec.
Towards the end of the MCA, two other dry events, centred on ca. 800 and 700 cal yr BP, have been recorded in the western Great Lakes area (Booth et al., 2006) and there are prolonged widespread droughts recorded in the Atlantic Palmer Drought Severity Index at ca. 850 cal yr BP (Cook et al., 2010(Cook et al., , 2011. Some sites within this study do contain dry shifts around these timings: low water tables are reconstructed at 850/800 cal yr BP in some more northerly sites (Petite Bog, Framboise Bog, Jeffrey's Bog, Lebel, Morts, Mosaik, Sterne, Lac Le Caron) and 700 cal yr BP in some more southerly sites (Villagedale Bog, Saco Heath, Jeffrey's Bog, Baie and Lebel) (Fig. 5). However, as was the case with the dry shifts recorded ca. 1000 cal yr BP, these dry excursions are not key prominent events within the last millennium. The subtle dry shifts recorded in sites within this study may be the muted manifestations of the drought events centred on more western regions of North America. This therefore suggests that either the climatic drivers of these Medieval drought events have reduced influence in this region of northeastern North America (owing to North Atlantic oceanic conditions and more abundant moisture sources associated with the majority of study sites) or that other climatic forces were more dominant in this region at this time. Alternatively or additionally, reconstructed water table lowering during the MCA within this study may be related to the warmer MCA temperatures which promoted peat growth and enhanced peat accumulation, thereby creating an autogenically-driven apparent lowering of the water table (e.g. Yu et al., 2003;Swindles et al., 2012).
Whilst drier average hydroclimate conditions during the MCA relative to pre-MCA are evident in WTD reconstructions from sites in central-northern Nova Scotia (Framboise Bog, Petite Bog) and Newfoundland (Jeffrey's Bog, Nordan's Pond Bog), wetter conditions were prevalent at the southern end of the study region in Maine (Saco Heath, Sidney Bog), southern Nova Scotia (Villagedale Bog) and around the Gulf of St. Lawrence (Baie, Plaine) (Figs. 5 and 6A). These results contribute to the growing body of evidence in support of wetter conditions in north-eastern US, particularly in Maine, during times of extreme drought in the western US from palaeoenvironmental records (Shuman et al., 2018). Support for these hydroclimatic spatial patterns is displayed in both modern mid-summer drought records (Cook et al., 2010) and MCA palaeoenvironmental reconstructions. Examples of other reconstructed wetter conditions in the MCA compared with pre-MCA can be found in Maine in pollen records (Gajewski, 1988), lake level reconstructions (Dieffenbacher-Krall and Nurse, 2005) and plant lipid biomarkers (n-alkanes) (Nichol and Huang, 2012). Existing studies from Qu ebec also support wetter MCA conditions relative to the LIA  (Paquette and Gajewski 2013;Holmquist et al., 2016), but reveal complex temperature patterns of warmer temperatures proximal to the James Bay Lowlands and cooler temperature in more oceanic regions around the Gulf of St. Lawrence (Viau et al., 2012).
The spatial extent of wet MCA conditions relative to the pre-MCA period in north-eastern North America is restricted to Qu ebec (St. Lawrence Estuary and the Gulf of St. Lawrence), Maine and southern Nova Scotia (Fig. 6A). This north-eastern extent is in agreement with findings from a recent synthesis study that suggested wetter MCA conditions extended to southern Nova Scotia (Shuman et al., 2018). Reconstructed WTD at sites north of Villagedale Bog in Nova Scotia are lower during the MCA relative to pre-MCA WTD (Fig. 6A), as supported by a water table lowering in a peat bog from Prince Edward Island (Peros et al., 2016), which is at a similar latitude. Therefore, a boundary between contrasting hydroclimatic shifts during the MCA in the Atlantic Provinces may lie between the locations of southern Nova Scotia and Prince Edward Island (43.5e46 N).
A notable exception to reconstructed wet MCA conditions in Maine is evident in a previously published record from Sidney Bog, which reported dry conditions throughout the MCA (900e550 cal yr BP) (Clifford and Booth, 2015). This contrasts with the near-surface WTD reconstructed from the same peatland in the present study, which was obtained from a different coring location. These discrepancies in water table reconstructions from the same peatland could either be a result of site-specific conditions at each coring location (e.g wind exposure vs. sheltered areas), non-climate related influences, such as autogenic feedbacks (Swindles et al., 2012) and/or secondary environmental gradients (Juggins, 2013) that could be influencing one or both of these records. Identifying autogenic processes and non-climatic drivers of WTD would require the analyses of more replicate cores obtained from this peatland. However, confidence in the reconstruction obtained from this study is provided by the aforementioned proximal records, which support regionally wet conditions during the MCA in the south of study area.

LIA hydroclimatic change (550e250 cal yr BP/1400e1700 CE)
The WTD spatial trends reconstructed for the beginning of the LIA in this study indicate a north east/south west split in hydrological conditions of opposing conditions compared with MCA WTDs: sites in northern Nova Scotia and Newfoundland were wet or becoming wetter whilst sites in Maine and many in Qu ebec were dry or becoming drier (Fig. 5). Such dry conditions at the start of the LIA have also been reported from a synthesis of North American drought data (Cook et al., 2009).
Average LIA hydroclimate conditions show a complex spatial pattern with wet conditions restricted to sites of James Bay Lowlands and southern and northern Nova Scotia (Figs. 5 and 6B). This spatial pattern of hydroclimate change is supported by some other published proximal records: WTD reconstructions obtained from a peatland in Prince Edward Island reported wet conditions during the LIA (Peros et al., 2016) and drought events are reported between 500 and 600 cal yr BP from other peatland WTD reconstructions in Maine (Clifford and Booth, 2013;Clifford and Booth, 2015). Pollen-based temperature reconstructions also reveal similar spatial trends with cooler LIA temperatures in the region surrounding the James Bay Lowlands and Nova Scotia (Viau et al., 2012). However, not all existing paleoenvironmental records support these average LIA climatic spatial patterns: higher lake levels were reconstructed during the LIA elsewhere in Maine (Dieffenbacher-Krall and Nurse, 2005), as reflected in a review of north-eastern US Late Holocene climate, which identified elevated moisture levels and/or reduced drought events throughout the LIA from 550 cal yr BP (Marlon et al., 2017). Furthermore, dry conditions are recorded at ca. 600 cal yr BP in the south-west of Qu ebec based on pollen records obtained from lake sediments (Paquette and Gajewski, 2013;LafontaineeBoyer and Gajewski, 2014) and more frequent fires during the LIA until 1850 CE in western Qu ebec indicate the dominance of dry Arctic air mass during this time (Bergeron et al., 2004).
The complex and varying spatial patterns in LIA WTDs that have emerged across the study region may be attributed to hydroclimatic variability and/or site-specific sensitivities to hydroclimate change. More variable climatic conditions during the LIA compared with the MCA have been reported from elsewhere in this region (Lafontaine-Boyer and Gajewski, 2014;Holmquist et al., 2016). The peatland WTD variability between the LIA and MCA was not statistically different when all sites in this study were considered together. However, elevated within-peatland multi-decadal hydroclimate variation during the LIA was identified in reconstructions from James Bay Lowlands and sites in Maine and Nova Scotia (Fig. 7). Site-specific sensitivities to hydroclimate change during the LIA may relate to local permafrost aggradation and limited snow accumulation, both of which would create relatively dry, highly localised conditions. For example, the hydrological regimes of the exposed northern peatlands in Qu ebec and Newfoundland may have been affected by persistent frozen layers at times during the LIA, as proposed by Van Bellen et al. (2011) to explain reconstructed peatland palaeohydrological trends in James Bay Lowlands. In addition, treeless coastal peatlands, such as those located along the Gulf of St. Lawrence, are highly exposed to winds which can limit snow accumulation during winter, allowing for the build-up of frozen peat layers and removing the additional snow water supply at the start of the growing season .
Many of the sites analysed in this study contain pronounced wet and dry shifts during the LIA (Fig. 5) but there is little consistency in the pattern of these events between sites since they vary in timing, duration and extent. The exception to this is the change towards drier conditions in all peatlands centred ca. 400 cal yr BP (1500 CE) (Fig. 5), a time of cold temperatures and permafrost development in Qu ebec (Arlen-Pouliot and Bhiry, 2005;Fillion et al., 2014). Whilst this widespread dry-shift may reflect an apparent drying in northern peatlands affected by frost, the timing of this dry-shift may also correspond to a mega drought reported in the 16th century across North America (Ladd et al., 2018), thus extending its known eastern extent. Tree-ring based temperature reconstructions from the Qu ebec-Labrador Peninsula show that regional terrestrial LIA cooling began in 1450 CE (Naulier et al., 2015), with the coolest phase of the LIA detected following the Tambora eruption in 1815 CE until the middle of the 19th century (Gennaretti et al., 2014). This LIA temperature minimum in the Qu ebec-Labrador Peninsula was therefore a century later than the coolest LIA phase detected when the Northern Hemisphere is considered as a whole (1700 CE, Mann et al., 2009). Similarly, the 19th century was highlighted as a particularly cold period at the St. Lawrence River at Qu ebec City based on historical records of ice bridge formation (Houle et al., 2007). The site-specific LIA trends of reconstructed WTD from this study do extend beyond 1700 CE (Fig. 5) and are comparable with 19th century regional dates (Naulier et al., 2015). However, chronological uncertainties for the majority of study sites overlap both time periods and therefore preclude refinement of LIA timings within this study.

Post-LIA (1850 CE-present day) hydroclimatic change
Reconstructed temperatures from the Gulf of St. Lawrence indicate that 20th century warming was greater than any experienced during the last millennium (Thibodeau et al., 2010) and records of precipitation across the study region show increases that are projected to continue because of increases in the concentration of atmospheric water vapour (Kunkel et al., 2013). All sites within this study, with the exception of Jeffrey's Bog in southern Newfoundland, exhibit pronounced WTD changes during the post-LIA period at the multi-decadal sampling resolution (Fig. 5) and a complex spatial pattern of hydroclimate change is evident across the study region: sites in the middle of the study region (midnorthern Nova Scotia and southern and eastern Newfoundland) are reconstructed as wetter than preceding periods, whilst sites at the latitudinal extremities (Maine, southern Nova Scotia and northern Newfoundland) and some sites in Qu ebec are drier ( Fig. 6C and D). These hydroclimatic spatial patterns are consistent across comparisons of both the MCA to post-LIA period and the LIA to post-LIA period ( Fig. 6C and D).
The unstable hydroclimate conditions in Maine and southern Nova Scotia in the post-LIA period may be caused by their proximity to sub-tropical storm tracks, which bring intense rainfall events to this region and could enhance variability in WTD. Whilst several studies have demonstrated that summertime extratropical cyclones have been decreasing in frequency since the 1980's (e.g. Chang et al., 2016;Colle et al., 2013), the intensity of these storms and the amount of precipitation that they deliver have increased because warmer temperatures increase the capacity of the atmosphere to hold water vapour (Pfahl et al., 2015;Colle et al., 2013). Under projected future climate warming, hurricane intensity in the North Atlantic basin is projected to further increase over the late 21st century, increasing precipitation rates and the number and occurrence days of category 4 and 5 storms (Knutson et al., 2015;Ting et al., 2019). Rapid increases in hurricane intensity are facilitated by reduced vertical wind shear along the U.S East Coast, which acts as a protective barrier to hurricanes, and greater sea surface temperature warming in this region (Ting et al., 2019).

Spatial expression of MCA and LIA hydrological change: implications for climate drivers
Reconstructed expressions of climate drivers during the MCA support a combined ocean-atmosphere climatic forcing mechanism: the Atlantic Multidecadal Oscillation (AMO) was in a persistent positive warm phase (Mann et al., 2009;Steiger et al., 2019) and the North Atlantic Oscillation (NAO) was in a positive phase (e.g. D' Arrigo et al., 2006;Trouet et al., 2009;Cronin et al., 2010). Whilst recent analysis of observational data and climate model simulations have demonstrated that the AMO signal is not distinguishable from climatic noise background (Mann et al., 2020), the associated climatic manifestations in eastern North America include a strengthening of the Polar Vortex thus generating a more zonal Polar Jet Stream and a northward storm track shift across the eastern Atlantic (Ineson et al., 2011;Martin-Puertas et al., 2012) (Fig. 9). The resultant increase in winter storm precipitation during the MCA in north-eastern North America would be accompanied by increases in summer-autumn precipitation arising from increases in the frequency and greater northwards extension of tropical cyclones (Mann et al., 2009).
Increases in reconstructed moisture availability during the MCA were restricted to the sites in the south of the study region in Maine, southern Nova Scotia and sites in Qu ebec proximal to the Gulf of St. Lawrence and the St. Lawrence Estuary, whereas sites in northern Nova Scotia, Newfoundland and Quebec's James Bay Lowlands were drier (Fig. 6A). Warm MCA temperatures would promote peatland productivity rates, particularly in northern sites with cooler average temperatures, and therefore contribute an autogenically-driven lowering of the WTD. However, the spatial patterns of reconstructed hydroclimatic trends most likely support a narrowing of the Westerly Wind belt, with a latitudinal limit in southern Nova Scotia, and elevated hurricane activity between southern Nova Scotia and southwestern regions of the Gulf of St. Lawrence and St. Lawrence Estuary (Fig. 9).
The lack of evidence for pronounced Medieval drought events in north-eastern North America (Section 4.1) may be a result of the sampling resolution within this study (ca. 40 years), since such events may not be captured in the WTD records or may only be registered as a hydrological excursion in one sample. However, the results from this study are in support of the growing body of evidence that there was a Pacific-based Medieval mega-drought driver (e.g. Coats et al., 2016), as first postulated by Booth et al. (2006) based on peatland records of hydroclimatic change in the western Great Lakes region. Steiger et al. (2019) have recently refined the theory behind climatic drivers of medieval drought through analysis of modelled hydroclimate conditions from the Palaeo Hydrodynamic Data Assimilation product .
Spatial patterns of WTD changes within north-eastern North America at the start of the LIA were opposite to those expressed during the MCA: sites in Maine and many in Qu ebec were dry or becoming drier whilst sites in northern Nova Scotia and Newfoundland were wet or becoming wetter (Fig. 5). These trends can therefore be explained by the switch to the negative phase of the NAO at the beginning of the LIA (e.g. Cronin et al., 2010), which would have contributed towards a reduction in the strength of the Polar Vortex and thus generated a meridional flow of the Polar Jet Stream (Ineson et al., 2011;Martin-Puertas et al., 2012) and longterm movement of storm tracks (Bakke et al., 2008) (Fig. 9).
A recent review of late Holocene North Atlantic Ocean circulation by Moffa-S anchez et al. (2019) highlights that MCA-LIA changes are difficult to reconcile because of differences in palaeoenvironmental reconstructions. For example, some studies have supported an AMOC slow down during the LIA (Lund et al., 2006), whilst others have inferred an AMOC strengthening (Rahmstorf et al., 2015;Thornalley et al., 2018). Commonalities between several studies do, however, support a cooling and increased drift ice export from north-east Greenland to the subpolar North Atlantic likely within the East Greenland Current (EGC) during the LIA, increasing the amount of polar waters reaching the Labrador-Baffin Bay region relative to Atlantic waters (Moffa-S anchez et al., 2019). This would therefore reduce sea surface temperatures in areas proximal to the northern oceanic sites within this study (Nordan's Pond Bog, Burnt Village Bog).
The spatial pattern of hydroclimate conditions across the study region is more complex when average LIA hydroclimatic conditions are considered, with more spatially variable average patterns of WTD change emerging compared with both the MCA and initial LIA conditions (Figs. 5 and 6B). The spatial patterns of reconstructed LIA WTD resemble patterns of Atlantic storm density tracks associated with a negative Summer North Atlantic Oscillation (SNAO) (Dong et al., 2013, Fig. 9). The SNAO, which is driven by the AMO (Sutton and Dong, 2012;Dong et al., 2013), has been reconstructed to have been in a persistent negative phase during the latter half of the LIA (Linderholm and Folland, 2017) and may therefore have played a role in shaping average LIA hydroclimatic conditions across the study region.

Last millennium environmental change as context for current and projected future conditions
The MCA is often referred to as a partial analogue for current and future climate change since it was associated with warmer conditions. However, reconstructed differences in interannual hydroclimate variability between the MCA and the LIA in the American southwest led Loisel et al. (2017) to conclude that the MCA may be a misleading analogue for future climate since interannual LIA hydroclimate conditions were more variable and more similar to 21st century projections than MCA hydroclimate. The multi-decadal peatland WTD reconstructions from north-eastern North America in this study also reveal more stable site-specific WTD reconstructions during the MCA than the LIA. However, these differences in variability are not statistically different when all study sites are considered together, indicating that the variability recorded across the study region is likely related to peatland-specific conditions rather than regional-scale hydroclimate variability. The post-LIA spatial patterns of reconstructed peatland WTD in northeastern North America from this study are more similar to the LIA than the MCA, as many sites in northern Nova Scotia and southern Newfoundland that became drier during the MCA became wetter during the LIA and post-LIA period (Fig. 6A,C,D). However, the 'warm LIA' scenario, suggested as an analogue for future hydroclimate variability by Loisel et al. (2017), is likely not appropriate for peatlands of north-eastern North America since it cannot account for important regional mesoclimatic LIA influences, such as permafrost aggradation, which impact peat hydrology and accumulation dynamics in this region.

Conclusions
Analysis of the testate amoebae-based water table reconstructions developed from peatlands in north-eastern North America has identified complex hydroclimatic spatial trends that varied over the last millennium. Climatic changes associated with the Medieval Climate Anomaly (MCA) and the Little Ice Age (LIA) have been detected in these peat records; however, such changes are not uniform across the region. Southern peatlands were relatively wet during the MCA, whilst northern and more continental sites were relatively dry. This spatial trend in MCA hydroclimate supports findings from other recent regional palaeoenvironmental reconstructions such as Ladd et al. (2018) and Shuman et al. (2018), and helps to constrain the northern geographical limit of these wet MCA conditions to southern Nova Scotia. West and Mid-West American MCA drought events do register as dry shifts in these Fig. 9. Synoptic climate map highlighting main climatic influences on MCA and LIA hydroclimate in north-eastern North America. SST ¼ sea surface temperatures, AMO ¼ Atlantic Multidecadal Oscillation, SNAO ¼ summer North Atlantic Oscillation (storm densities represented by yellow shading with darker (lighter) yellow representing greater (fewer) storm densities), EGC ¼ East Greenland Current. Blue and red circles respectively represent wet and dry reconstructed WTD at the study sites. Light blue arrows represents cold Labrador Current and red arrow represents warm Gulf Stream. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) north-eastern records; however, they are not prominent hydroclimatic features. There is therefore no evidence that the Medieval droughts extended across the Atlantic seaboard of north-eastern North America, based on the multi-decadal resolution of WTD reconstructions from this study. This finding is consistent with the hypothesis that MCA mega-droughts were primarily forced from Pacific climate drivers (e.g. Steiger et al., 2019). Future research that develops palaeoclimate records from currently under-represented areas in north-eastern North America, including New Brunswick and Maine, would help to further characterise late Holocene regional climate variability and test the absence of Medieval droughts in this region.
LIA hydroclimate change in north-eastern North America was often spatially and temporally variable rather than displaying clear WTD directional shifts or latitudinal trends. These variable regional palaeoenvironmental reconstructions support conclusions from synthesis papers, which demonstrate that some dry periods in the LIA were comparable with those reconstructed during the MCA (Ladd et al., 2018;Mann et al., 2009), despite manifesting from different climate forcings or peat hydrological influences such as LIA drying due to local permafrost aggradation. The spatial patterns of peatland WTD reconstructed in this study indicate that the MCA and LIA expressions of climate were likely influenced by complex interactions of external and internal climate drivers and feedbacks that contributed towards a more zonal Polar Jet Stream and increased hurricane activity during the MCA and a meridional Polar Jet Stream flow, combined with Atlantic storm precipitation patterns and extensions of cold polar waters during the LIA. The spatial and temporal diversity of peatland responses to hydrological change reconstructed in this study highlights the complex balance between regional hydroclimate variability and internal peatland dynamics that occurred in north-eastern North America over the last millennium.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.