Corresponding author: S. St. George, Department of Geography, Environment and Society, University of Minnesota, Minneapolis, MN 55455, USA. (email@example.com)
 Estimates of future flood risks are based on the observations of past floods, but instrumental records of basin hydrology are often too short to assess potential changes in the frequency or magnitude of extreme floods over time. In this study, we show that bur oak (Quercus macrocarpa Michx.) growing along the Red River of the North in North Dakota and Minnesota preserve evidence of past floods within their annual growth rings. Rings formed during major floods often displayed (i) marked reductions in the size of their earlywood vessels or (ii) a more diffuse distribution of vessels throughout the increment. Because of the correspondence between major floods and widespread anatomical anomalies within riparian oaks, we suggest that these features can be described as “flood rings.” The frequency of flood-ring formation varied substantially along the river, which implies that this evidence can only provide an accurate estimate of flood history when it is obtained from many trees sampled across a dense network of sites. The rate of flood-ring formation is primarily influenced by flood magnitude but is also controlled by the timing and duration of inundation relative to the period of cambial growth. Although flood-affected oaks are imperfect recorders of past floods, this approach offers significant potential as a means to estimate the preinstrumental flood history of the Red River within the United States.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Flooding caused by the Red River of the North (Figure 1) is arguably the most damaging recurring natural hazard that affects the northern Great Plains of North America, and major Red River floods are enormously disruptive to communities located along the river in both Canada and the United States. The 1997 flood forced more than 50,000 people to evacuate their homes and caused damage in excess of U.S. $4.8 billion (2012 dollars) [International Joint Commission, 2000; Shelby, 2003], making it the most expensive disaster in the basin's modern history. Because the American portion of the basin was also affected by severe floods in 1979, 2009, and 2011, some observers have argued that flooding in the river's upper course has become more severe and more frequent in recent decades [Flood Diversion Board of Authority, 2012].
 One of the main obstacles to understanding long-term trends in the frequency and magnitude of extreme floods is the brevity of instrumental records of basin hydrology [Klemeš, 1989]. The first stream gauge on the Red River was installed at Grand Forks, North Dakota, in 1882, but most hydrological records for the river began as recently as the 1930s. Working downstream in Manitoba, St. George and Nielsen  reported that tree-ring specimens from bur oak (Quercus macrocarpa Michx.) growing along the Red River contained unusual anatomical features within their annual growth rings that coincided with major nineteenth-century floods. In this paper, we report that riparian oaks growing along the Red River in North Dakota and Minnesota also contain rings with abnormal anatomy and that these anomalous rings formed most frequently during years with exceptional Red River floods. Our results demonstrate that paleoflood hydrology based on anatomical tree-ring signatures is not restricted to the Canadian portion of the Red River basin but can also provide useful information about the occurrence of floods upstream in North Dakota and Minnesota.
2. Study Region
 The Red River of the North arises at the confluence of the Bois de Sioux and Ottertail rivers within the twin cities of Wahpeton, North Dakota, and Breckenridge, Minnesota. The Red River is an 880 km long, single-channel, meandering river that flows northward and serves as the border between North Dakota and Minnesota before entering the Canadian province of Manitoba (Figure 1). The river receives runoff from more than 290,000 km2, making its watershed one of the largest on the northern Great Plains [Brooks and Nielsen, 2000]. Forest cover makes up only 5.5% of the basin [Brooks, 2011] and trees are usually restricted to a narrow (100–200 m wide) corridor surrounding rivers and streams. Species that grow commonly along the river include bur oak, box elder (Acer negundo L.), green ash (Fraxinus pennsylvanica Marshall), and plains cottonwood (Populus deltoides W. Bartram ex Humphry Marshall).
 In the United States, the major population centers of Grand Forks-East Grand Forks and Fargo-Moorhead have received the largest investments for flood protection. After the 1997 flood, Grand Forks and East Grand Forks constructed a 13 km long, 3 m wide flood wall and storm-water pumping system protecting the two cities at a cost of U.S. $409 million [United States Army Corps of Engineers, 2012]. The U.S. Army Corps of Engineers have recommended the construction of a 58 km long diversion channel and other control structures to safeguard the cities of Fargo and Moorhead at an estimated cost of U.S. $1.7 billion [Temple, 2011]. Because the level of protection offered by these structures is estimated from flood-frequency analysis [Foley, 2010], an improved understanding of the history of Red River floods would be very helpful to immediate and near-term decisions about major infrastructure projects in the basin.
 We collected tree-ring samples from 176 living bur oaks at nine sites along the course of the Red River, ranging from its upstream limit to the Canada-U.S. border (Figure 1, Table 1). Site selection was guided by criteria that (1) ensured all sampled trees were inundated during recent major floods and (2) allowed us to compare botanical evidence to nearby hydrological data. All sampled stands were located within the flood zone for both the 1997 and 2009 Red River floods (flood maps obtained from the International Water Institute, May 2011). We also noted field evidence of recent high water at several sites, including debris lodged in branches several meters above the ground surface, basal scouring, alluvial clay plastered around the lower trunks, and abrasion or impact scars. We chose sites close to active U.S. Geological Survey (USGS) streamflow gauges (0–4 km up or down river), so hydrological data could be used to estimate the depth and duration of past flooding at each stand of oaks. Most of the trees sampled were less than 30 m from the edge of the river, and several were flooded by the river at the time of collection. Because the anatomical response to flooding exhibited by bur oak is most prominent near the base of the tree (St. George et al., 2002), we used increment borers to collect one sample from each tree as close to the ground as possible (median sampling height was 40 cm). After all samples had been examined for anatomical abnormalities, each one was dated by visually matching its sequence of relative tree-ring widths against the pattern described by other trees from the same site [Stokes and Smiley, 1968].
Table 1. Metadata for the Quercus macrocarpa Tree-Ring Sites Along the Red River of the North Developed by This Study
Number of Trees
Span of Record
St. Vincent, Minnesota
Pembina Golf Course
Pembina, North Dakota
Grand Forks, North Dakota
Grand Forks, North Dakota
Oak Grove Park
Fargo, North Dakota
Red River Trail
Fargo, North Dakota
Fargo, North Dakota
Bois de Sioux Golf Course
 Normal growth rings in bur oak (Figure 2a) feature one or two ranks of large conductive vessels within the earlywood followed by latewood composed primarily of fiber and flame parenchyma. We noted three main types of anatomical abnormalities within our specimens. First, some rings had earlywood vessels that were markedly smaller (“shrunken vessels”) than those present in prior or subsequent growth rings (Figure 2b). Second, although most rings have few or no vessels within their latewood, selected rings had relatively large vessels scattered through both the earlywood and the latewood (“extended vessels”; Figure 2c). Third, some rings had unusually small vessels present through the entire growth layer (“shrunken and extended vessels”; Figure 2d). Bur oak growing within the riparian forest along the Red River of the North exhibited unusual anatomical features in roughly 4% of their growth rings (771 out of the total 18,286 rings). For most years, trees at a given site did not have any rings that contained anomalous vessels (Figure 3; see Table S1 in the Supporting Information). Instead, the majority of sites had 2 or 3 years (or more) where a large fraction of the trees at that location exhibited one or more types of anatomical abnormality.
 Those cases where a high proportion of trees within a given site contained anomalous rings nearly always coincided with a flood recorded at the closest stream gauge. When three or more trees at a site featured unusual vessel characteristics within the same ring, peak stage at the nearest gauge exceeded flood stage (determined by the U.S. National Weather Service) in 47 of 53 cases. Several floods were associated with frequent anatomical anomalies at multiple sites. Growth rings formed in 1950 showed unusual vessel characteristics at the four northernmost sites (Pembina Golf Course, Seventh Street in St. Vincent, Lincoln Park, and Riverside Park). This geographic pattern matches the hydrological profile of the 1950 Red River flood, which was most severe along the downstream reach of the river. Both the 1996 and 1997 floods were associated with frequent vessel anomalies at Pembina Golf Course, Lincoln Park, Oak Grove Park, and Welles Park. One third of oaks sampled at Fargo (18 of 60) had abnormalities within their 2009 ring, which was coincident with the flood-of-record for that community (recorded at USGS gauge 05054000). Several oaks at Pembina Golf Course (15 of 20) and Lincoln Park (5 of 20) have vessel anomalies in their 1979 ring, which was concurrent with a major flood at Pembina and the Canadian portion of the basin.
 Prior work in Canada [St. George and Nielsen, 2000, 2002, 2003] and France [Astrade and Bégin, 1997] demonstrated that prolonged inundation can cause oaks to reduce the mean transverse area of earlywood vessels within their annual rings. Oaks in North Dakota and Minnesota exhibit the same reduction in earlywood vessels during flood years but also form vessels throughout their entire growth ring, a response to inundation that has not been reported previously for this species. Because of the observed match between major floods and widespread anatomical anomalies within flooded oaks, we suggest that it is appropriate to describe these features as “flood rings.”
 The frequency that oaks formed flood rings varied substantially across the network, and sites separated by only a few kilometers often exhibited different responses to the same floods. For example, oaks sampled at the Seventh Avenue site in St. Vincent, Minnesota, contained vessel anomalies in only 2.8% of their growth rings, whereas trees from the Pembina Golf Course site on the opposite side of the river formed anomalous rings more than twice as often (6.9% of all rings). These discrepancies may be the product of very modest topographic differences between sites or among trees within the same stand, although we did not note any relationship between the proximity of trees to the river and the frequency of flood-ring formation. Because the frequency of flood-ring formation varies considerably across our network, we conclude that flood-ring evidence can only provide an accurate assessment of the occurrence of past floods when it is obtained from many trees sampled across a dense network of sites.
 Those cases where flood rings were present in many trees across a stand were nearly always associated with a major flood at that location, which implies the flood history preserved within riparian oaks is a censored record that only documents high-magnitude events. Some major floods also left behind little or no evidence in the regional flood-ring record. The 2010 Red River flood was the fifth-largest event (by peak stage) since 1900 recorded at gauges in Wahpeton, Fargo, and Grand Forks, but flood rings were either absent or extremely rare at sites in all three locations. The Red River flood of 2006 was similarly a top-five event at Fargo, Grand Forks, and Pembina but was recorded by multiple trees at only one site (Pembina Golf Course). The absence of substantive flood-ring evidence for these major events implies that flood timing and duration may also influence the ability of riparian oaks to act as surrogate flood recorders. The 2010 flood began exceptionally early (exceeding flood stage at Fargo on March 13) and was over by the third week in April. The 2006 flood was a relatively brief event (flood stage at Fargo was exceeded for only 21 days) that also ended in mid-April. In contrast, floods in both 2001 and 2009, which produced extensive flood-ring evidence at several sites, persisted until the middle of May (the 2009 flood exceeded flood stage at Fargo for 60 consecutive days). These observations confirm that flood magnitude is not the only factor that influences the rate of flood-ring formation in riparian oaks and suggest that the timing and duration of inundation relative to the period of active cambial growth may also play a role in determining whether or not trees preserve evidence of past floods within their annual growth rings [St. George and Nielsen, 2002].
 Prior research in Manitoba used anatomical tree-ring signatures in riparian oaks to extend the record of extreme floods in the downstream portion of the Red River of the North back several centuries [St. George and Nielsen, 2003]. In this study, we have shown that oaks farther upstream in North Dakota and Minnesota contain similar anatomical signatures within their annual growth rings and that these features are also linked to recent major floods. Our analysis also makes it clear that flood-affected oaks are imperfect recorders of past floods, as they contain abundant evidence of certain events but give little to no indication of the occurrence of other floods of similar magnitude. This behavior reflects the fact that trees are complex biological systems whose ability to record floods is influenced by phenology and the immediate growth environment of individual organisms. If longer tree-ring sequences were developed from older living trees or ancient wood recovered from archeological settings, it could be possible to use such an extended record to estimate the frequency of floods in the upper Red River basin during the period prior to direct hydrological monitoring.
 We are grateful to D. Holehouse, M. Jacobson, and F. Fernando for providing assistance during field collections in North Dakota and Minnesota. We also thank M. Fugazzi (City of Grand Forks), F. Haberman (City of Fargo), N. Brandt (Pembina Golf Course), W. Niesche (City of Breckenridge), and W. Beyer (Wahpeton Parks & Recreation Department) for granting permission to sample. The final manuscript was improved by comments from E. Bigio and two anonymous referees. This study received partial support from the University of Minnesota's Undergraduate Research Program.