Study region

In northwestern Québec, the Upper Harricana River is representative of the Abitibi Plains’ hydrological dynamics over the last 250 years.

Study focus

Planning for future spring flood risks involves uncertainties. This research presents a multicentury evaluation of changes in spring mean discharge and flood drivers using streamflow reconstruction (1771–2016), observations (1915–2020) and projections (2021–2100).

New hydrological insights for the region

Using a downscaled CMIP5 ensemble of 10 global climate models (GCMs), generalized additive mixed modeling of mean spring discharge projections matched those of an independent mechanistic model and eight GCMs projected variability in spring discharge by 2100 to be similar to the historical variability reconstructed for the last 250 years across the Abitibi Plains. Results indicate that the projected decline in snow cover (–20 to –30% annual snowfall) and rise in winter and spring temperature may be offset by a greater contribution of rainfall to spring high discharge (+100 to +125 mm). However, two GCMs projected an increase in the magnitude and frequency of high mean spring discharge for the Abitibi Plains. By investigating future mean spring discharge for the Upper Harricana River in reference to past reconstructed variability, this study provides insights to inform the future management of regional water resources. The importance of estimating future regional flood risks from the behavior of multi-model ensembles is highlighted.

Few records of spring paleoclimate are available for boreal Canada, as biological proxies recording the beginning of the warm season are uncommon. Given the spring warming observed during the last decades, and its impact on snowmelt and hydrological processes, searching for spring climate proxies is receiving increasing attention. Tree-ring anatomical features and intra-annual widths were used to reconstruct the regional March to May mean air temperature from 1770 to 2016 in eastern boreal Canada. Nested principal component regressions calibrated on 116 years of gridded temperature data were developed from one Fraxinus nigra and 10 Pinus banksiana sites. The reconstruction indicated three distinct phases in spring temperature variability since 1770. Ample phases of multi-decadal warm and cold springs persisted until the end of the Little Ice Age (1850–1870 CE) and were gradually replaced since the 1940s by decadal to interannual variability associated with an increase in the frequency and magnitude of warm springs. Significant correlations with other paleotemperature records, gridded snow cover extent and runoff support that historical high flooding were associated with late, cold springs with heavy snow cover. Most of the high magnitude spring floods reconstructed for the nearby Harricana River also coincided with the lowest reconstructed spring temperature per decade. However, the last 40 years of observed and reconstructed mean spring temperature showed a reduction in the number of extreme cold springs contrasting with the last few decades of extreme flooding in the eastern Canadian boreal region. This result indicates that warmer late spring mean temperatures on average may contribute, among other factors, to advance the spring break-up and to likely shift the contribution of snow to rain in spring flooding processes.

In northeastern boreal Canada, the long-term perspective on spring flooding is hampered by the absence of long gage records. Changes in the tree-ring anatomy of periodically flooded trees have allowed the reconstruction of historical floods in unregulated hydrological systems. In regulated rivers, the study of flood rings could recover past flood history, assuming that the effects of hydrological regulation on their production can be understood. This study analyzes the effect of regulation on the flood-ring occurrence (visual intensity and relative frequency) and on ring widths in Fraxinus nigra trees growing at five sites distributed along the Driftwood River floodplain. Driftwood River was regulated by a dam in 1917 that was replaced at the same location in 1953. Ring width revealed little, to no evidence, of the impact of river regulation, in contrast to the flood rings. Prior to 1917, high relative frequencies of well-defined flood rings were recorded during known flood years, as indicated by significant correlations with reconstructed spring discharge of the nearby Harricana River. After the construction and the replacement of the dam, relative frequencies of flood rings and their intensities gradually decreased. Flood-ring relative frequencies after 1917, and particularly after 1953, were mostly composed of weakly defined (less distinct) flood rings with some corresponding to known flood years and others likely reflecting dam management. The strength of the correlations with the instrumental Harricana River discharge also gradually decrease starting after 1917. Compared with upper floodplain trees, shoreline trees at each site recorded flood rings less frequently following the construction of the first but especially of the second dam, indicating that water level regulation limited flooding in the floodplains. Compared with the downstream site to the dam, the upstream ones recorded significantly more flood rings in the postdam period, reemphasizing the importance of considering the position of the site along with the river continuum and site conditions in relation to flood exposure. The results demonstrated that sampling trees in multiple riparian stands and along with various hydrological contexts at a far distance of the dams could help disentangle the flooding signal from the dam management signal.

In recent years, the utility of earlywood vessels anatomical characteristics in identifying and reconstructing hydrological conditions has been fully recognized. In riparian ring-porous species, flood rings have been used to identify discrete flood events, and chronologies developed from cross-sectional lumen areas of earlywood vessels have been used to successfully reconstruct seasonal discharge. In contrast, the utility of the earlywood vessel chronologies in non-riparian habitats has been less compelling. No studies have contrasted within species their earlywood vessel anatomical characteristics, specifically from trees that are inversely exposed to flooding. In this study, earlywood vessel and ring-width chronologies were compared between flooded and non-flooded control Fraxinus nigra trees. The association between chronologies and hydroclimate variables was also assessed. Fraxinus nigra trees from both settings shared similar mean tree-ring width but floodplain trees did produce, on average, thicker earlywood. Vessel chronologies from the floodplain trees generally recorded higher mean sensitivity (standard deviation) and lower autocorrelation than corresponding control chronologies indicating higher year-to-year variations. Principal components analysis (PCA) revealed that control and floodplain chronologies shared little variance indicating habitat-specific signals. At the habitat level, the PCA indicated that vessel characteristics were strongly associated with tree-ring width descriptors in control trees whereas, in floodplain trees, they were decoupled from the width. The most striking difference found between flood exposures related to the chronologies' associations with hydroclimatic variables. Floodplain vessel chronologies were strongly associated with climate variables modulating spring-flood conditions as well as with spring discharge whereas control ones showed weaker and few consistent correlations. Our results illustrated how spring flood conditions modulate earlywood vessel plasticity. In floodplain F. nigra trees, the use of earlywood vessel characteristics could potentially be extended to assess and/or mitigate anthropogenic modifications of hydrological regimes. In absence of major recurring environmental stressors like spring flooding, our results support the idea that the production of continuous earlywood vessel chronologies may be of limited utility in dendroclimatology.

Tree rings from ring-porous species have often been used as flood proxy. Many ring-porous species produce characteristic flood rings in response to stem submersion during vessel formation. Flood rings have earlywood vessels that are more numerous and/or of smaller cross-sectional area than “normal” rings. This study aimed at determining if diffuse-porous balsam poplar and trembling aspen, like ring-porous black ash, produce anatomically distinct annual tree rings in response to flooding. More specifically, we asked (i) if periodic tangential bands of vessels (hereafter PTBV) could be as easily identified/quantified as flood rings and (ii) if PTBV could be associated with spring flooding. Sampling of black ash, balsam poplar and trembling aspen trees took place along a flood exposure gradient in the floodplain of Lake Duparquet in northwestern Québec. Two observers recorded flood rings and PTBV. Consistency between observers was greatest when identifying flood rings. In both diffuse-porous species, PTBV occurred less abundantly than flood rings in black ash. They also occurred less often in balsam poplar than in trembling aspen. Years in which PTBV were initiated early in the growing season were associated with years in which flood rings occurred. Like flood rings, early occurring PTBV were more abundant in springs characterized by high mean river discharge, extensive snow cover, cold temperatures and heavy precipitation. Early-occurring PTBV dominated in flooded sites and late-occurring ones dominated in the control site. However, PTBV of the late-types were also observed in both flood exposures indicating that spring flood may not be the only factor modulating their formation. While flood rings seem to be associated with a change in the transport of growth regulators resulting from stem submergence and excess water, PTBV may be reflective of rhythmic alterations in the transport of growth regulators resulting from either water excess or deficit. Despite promising findings, many questions remain before PTBV in riparian diffuse-porous species can be widely used as a flood proxy. Why do species and individual trees differ in their ability to record them? What is the full range of environmental conditions triggering PTBV’s formation especially in unflooded sites and in the late growing season?

In eastern boreal Canada, long-term perspective in water resources and hydroelectric dam management is currently limited by the lack of long-term hydrological records. The research for new paleohydrological proxies would help fill this hydrological data gap and provide regional hydroclimatic predictive trajectories in the context of climate change. The development of long annually resolved series of earlywood vessel cross-sectional area has recently demonstrated a high potential for reconstructing high and low discharges. This study analyzes a network of 10 sites scattered around Lake Duparquet. The region covers an area of about 20 000 km2 including four river basins characterized by natural and regulated rivers, and unflooded control sites. The objectives were to assess 1) the spatial coherency in flood-rings chronologies among sites and Lake Duparquet, and among hydrological regimes (natural, regulated and unflooded control) and 2) their degree of association with i) annually resolved chronologies of earlywood vessel cross-sectional area and number, ii) a reconstruction of the Harricana River spring discharge and iii) discharge data from eleven hydrometric stations distributed in the study area. It was hypothesized that flood rings would be consistent among natural rivers and absent from regulated rivers. Results showed high spatial coherency among natural rivers with flood rings recording the major floods of the last 250 years. Flood ring and earlywood vessel chronologies were strongly correlated to both reconstructed and instrumental discharge data. On regulated rivers, trees were younger than at the other sites and mainly spring floods that occurred prior to dam creation and the few extreme floods after dam creation were recorded by flood rings. One hypothesis is that older trees (before dam) most likely recorded the natural dynamic of the river, while younger trees (after dam) most likely recorded dam management maneuvers and spring flood exceeding dam capacity. Flood rings and earlywood vessel chronologies provided comparable and complementary hydrological evidence. Flood rings were easily identified visually allowing fast determination of major flood years whereas developing earlywood vessel chronologies, while being more tedious and time consuming, allowed capturing a larger spectrum of hydrological conditions

In eastern boreal Canada, variability in river discharge is poorly understood at the multi-century scale due to short instrumental records. In recent decades, increased magnitude and frequency of spring floods have raised concerns about the potential effects of climate change on flood risk. Unlike tree-ring width, flood rings have a demonstrated dendrochronological utility for reconstructing high discharge in boreal environments. In this study, twelve chronologies of earlywood vessel cross-sectional area (a new hydrological proxy) and ring width were developed from riparian Fraxinus nigra trees periodically flooded in spring. These chronologies were used as predictors of Harricana River spring discharge, which was reconstructed for the period 1771–2016. The reconstruction captured 69% of the variance over a 102-year calibration period. The reconstruction indicates that the magnitude and frequency of spring high discharge has increased since the end of the Little Ice Age (1850–1870 CE) and since 1950. The change from a multi-decadal frequency in the late 19th century to a decadal and then interannual frequency in the late 20th century is associated with an increase in snow cover over much of central-eastern Canada. The association between the reconstructed spring discharge and spring atmospheric circulation indices NINO3.4, AMO, NAO may also have changed in these periods and further work is needed to assess the stability of these associations. The correlation between reconstructed and instrumental spring discharge at the regional scale, as well as the shared features in reconstructed discharge and other paleorecords from subarctic Québec suggest a common hydrological signal across the study area and for the early 20th to 21st centuries. The unprecedently low and high spring discharge in recent decades compared to the historical natural variability of the last 250 years also suggests that the increase in flood frequency and magnitude originates from climate change.