Keywords

1 Introduction

Floods are one of the most common natural processes worldwide. In mountain streams, the catchment predisposition and steep channel characteristics contribute to a quick hydrological response of catchment and very powerful channel responses. Floods in mountain streams, therefore, are characterized by recurrent, highly-turbulent and sediment-laden flows (Wohl 2000, 2006). These characteristics make mountain streams highly hazardous and flood events therein difficult to forecast (Borga et al. 2011, 2014). Floods in mountain environments have caused large amounts of losses and fatalities in the past, with dramatic and recurrent episodes in all mountain ranges worldwide (Weingartner et al. 2003; Kundzewicz et al. 2014a).

The scientific-technical analyses of floods in mountains require long-term records on both occurrence and flow magnitude (Brázdil et al. 2006). However, the widespread and characteristic scarcity of instrumental data and basic hydrological information in mountain terrains clearly hampers these analyses. Augmenting information about past flood events by using historical archives and indirect evidence in the paleorecord is therefore critical to assess flood hazard, evaluate discharge trends or decipher a link to climate change. Paleoflood records based on tree-ring analyses are, therefore, highly convenient in forested mountain terrains (Ballesteros-Cánovas et al. 2015a, b). The interaction between floods and vegetation growing along a mountain stream can leave datable evidence of past floods in tree rings, which allows tracking of past process activity with high spatial and temporal accuracy. Before the formal definition of paleohydrology (Baker 2008), Sigafoos (1964) already stated that the information contained in tree-ring records of riverine trees is unique and constitutes a valuable source of data for hydrological analysis (Ballesteros-Cánovas et al. 2014). Since then, several efforts have been done to improve the tree-ring analyses to determine past flood records (Ballesteros-Cánovas et al. 2015b for a recent review). All these experiences support the initial hypothesis launched by Sigafoos (1964) about the economic value of tree-rings records for paleoflood analyses, specifically in ungauged or poorly gauged sites. However, a pedagogic explanation about the utility of this kind of information is still needed within the hydrological community (Baker 2008).

Particularly in the Tatra Mountains, flood risk is an important issue and has been a known problem for long time. However, despite the fact that first gauging stations were installed in several streams and rivers in the Tatra Mountains foreland in the late 19th century, a lack of flood processes interpretation is still recognized because of the highly fragmented and incomplete gauging network in the region (Fig. 1). This problem is even larger as some crucial gauging stations have been relocated after the occurrence of intense flood events (e.g., in 1934), which destroyed several stations and prevented direct data comparison. Moreover, the political and administrative instability of the region—related to both World Wars—also resulted in several gauging network changes and significant losses of records (Szczepański et al. 1996).

Fig. 1
figure 1

Spatial and temporal distribution of meteorological and flow gauge dataset on the northern slope and foothills of the Polish Tatra Mountains. After Ballesteros-Cánovas et al. (2015a), published by permission of Elsevier

Full size image

Here we synthesize the major research outcomes obtained during the Florist project and report on the use of tree-ring series to decipher the flood history of the Tatra Mountains. We first present the foundation of paleoflood research based on tree rings, provide some fundamental bibliography and then summarize some of the key findings of Ballesteros-Cánovas et al. (2015a, 2016) dealing with the flood dating and peak discharge reconstruction in the region.

2 Fundamentals of Tree-Ring-Based Paleoflood Reconstructions

The methodological basis of paleoflood reconstructions based on tree rings is based on the “process–event–response” concept defined by Shroder (1978). This concept defines the process as any geomorphic agent, in this specific case floods, the event as a specific impact which the process can induce to a tree; and the response the tree’s reaction to this specific disturbance. The main disturbances caused by floods include impact scars, tilted stems and abnormal stem morphologies, eroded roots, re-sprouting and anatomical abnormalities caused by prolonged inundation without needed to invoke mechanical damages (Stoffel and Bollschweiler 2009; Stoffel et al. 2010; Stoffel and Corona 2014). Hydrological processes also control riparian vegetation pattern; these have been used to understand the frequency and duration of flooding events (Kozlowski 2002; Kames et al. 2016).

At the level of forest stands, flood activity modulates the riparian forest. Extreme flood events may eliminate trees along the channel and flood plain. Survivor trees may benefit from less competition, more light, nutrients and water availability (Osterkamp et al. 2012; Stoffel and Wilford 2012). At the same time, new plants may germinate on bare surfaces. These vegetation changes condition the growth rate and the age structure of the forest, which has been used to date the minimum age of new landforms and its linkage with fluvial dynamic (Sigafoos 1964).

At the scale of individual trees, and in the case of Tatra Mountain streams, we most commonly observe mechanical damage in the form of stem wounds (Fig. 2a). Previous studies have reported that in this region, the transport of boulders and woody debris during intense flood events can impact trees, removing the bark and destroying the meristematic cambium tissue (Zielonka et al. 2008). Scars on trees can, however, also be generated by falling trees or impacts during logging, which is quite widespread in the region. After wounding, trees react upon this disturbance by compartmentalizing the wounded surface (Stoffel and Klinkmüller 2013; Ballesteros et al. 2014) so as to minimize rot and decay. Therefore, the partial removal of bark and cambium tissues during mechanical wounding will favor the formation of callus pads at the edge of open wounds and dissimilar growth disturbances according to tree species.

Fig. 2
figure 2

Examples of main dendrogeomorphic evidence related with floods in the Tatra Mountains: a scar on trees and b exposed roots

Full size image

Conifers growing in the Tatras Mountains respond to wounding in a way which allows them to minimize the risk of cell embolism and cavitation and by forming smaller cells. In addition, Abies, Larix and Picea have been reported to form tangential rows of traumatic resin ducts (TRD) after mechanical disturbance (Stoffel 2008; Schneuwly et al. 2009a, b; Ballesteros et al. 2010a), as well as a significant reduction of both ring width and the size of the lumen of earlywood tracheid (Arbellay et al. 2012a, b; Ballesteros-Cánovas et al. 2010a). Specifically, TRDs normally form tangentially within the current growth ring and close to the wound, but can also be observed in subsequent rings (Bollschweiler et al. 2008; Stoffel and Hitz 2008). This anatomical feature represents a nonspecific defense response to injury that compartmentalizes the wood, thereby preventing the attack of fungi and insects, as well as the loss of cell moisture (Shigo 1984). The occurrence of TRD in the tree-ring records represents a valuable indicator of the existence of ‘‘hidden’’ scars, and can even provide information on the seasonality of events (Stoffel and Beniston 2006; Schneuwly-Bollschweiler and Stoffel 2012). In broadleaved trees, the main reaction of trees to wound damages is a significant decrease in mean vessel size within rings formed during flood conditions (Arbellay et al. 2012b; Ballesteros-Cánovas et al. 2010b).

Tilted trees are also a common feature in this region. The sudden, unidirectional pressure on the stem generated by geomorphic events (e.g., deposition of material by a debris flow) as well as the destabilization of the root-plate system by bankfull erosion cause that the stems of many trees close to the channel are rotated (Fig. 2b). Trees are responding to tilting with the formation of reaction wood (Stoffel and Bollschweiler 2008, 2009). This anatomical feature is recognized macroscopically through the occurrence of eccentric growth rings in the stem. The recognition of compression wood in the tree-ring records allows determination of the moment of tilting, and consequently dating the flood event. By using dendro-mechanical models, the magnitude of the tilting can also correlated with the flow energy and consequently provides information about the peak discharge at a given cross-section (Ballesteros-Cánovas et al. 2015c).

In the tree-ring records, growth suppression can be recognized in trees suffering from crown decapitation or stem burial (Procter et al. 2012; Kogelnig-Mayer et al. 2013). It has been reported that crown or branch losses reduce photosynthetic activity, which in turn affects the growth rate. Similarly, stem burial may prevent water supply to the roots which in turn will lead to a decrease in growth rates, although the opposite reaction has been reported as well in case when a thin layer of nutrient-rich sediments is deposited around the trunk (Kui and Stella 2016; Friedman et al. 2005).

Finally, the last botanical evidence of past floods in the Tatra Mountains is exposed roots, which are caused by bankfull failures (Osterkamp et al. 2012; Stoffel et al. 2012). Exposed roots present clear signal in the tree-ring records, reflecting sudden or continuous erosion process by noticeable changes in wood anatomy (Stoffel et al. 2013).

The large set of potential evidence coming from trees can be used in paleoflood reconstructions, but also points to differences in the intensities of different responses. Stoffel and Corona (2014) provide a detailed overview on criteria used to assign signal intensity based on the anatomical features for different species as well as on the existing indices and minimum sampling sizes. In the case of paleoflood studies, flood scars on trees are the most useful botanical indicators of paleofloods since they do not only provide information on past events with seasonal resolution, but also represent paleostage indicators (i.e. height of the scar; Baker 2008) in flow discharge estimations (Ballesteros-Cánovas et al. 2015a).

Based on literature review (n = 52 studies performed in North America and Europe; Ballesteros-Cánovas et al. 2015b), the number of sampled trees per study was 104 ± 122 trees, of which 60 % were broadleaved trees and 40 % conifer trees.

3 Paleoflood Reconstruction in Tatra Mountain Streams Using Tree Rings

3.1 Study Site

The Tatra Mountains (Tatras) are located in the Carpathian arc (max. elevation 2655 m a.s.l.) between Poland and Slovakia. The Tatras are a crystalline and metamorphic core mountain range covered by nappes of Mesozoic sedimentary rocks. During the Pleistocene, the Tatras underwent at least three glaciations, which strongly reshaped the region, left conspicuous relief forms and moraine deposits. This mountain terrain is highly susceptible to geomorphic processes such as debris flows, snow avalanches and floods, which have caused several disasters during the last century. The study was conducted in four streams, i.e. Rybi Potok (RP), Roztoka (RS), Strążyski (ST) and Chochołowski streams (CH), covering the western-central and eastern parts of the northern slopes of the Tatra Mountains (Fig. 3).

Fig. 3
figure 3

a Location of the study site in the Carpathians. b Location of the investigated catchments together with meteorological stations

Full size image

In this region, climate is influenced by regional air mass oscillations and local topography, with a predominance of polar marine (65 % of annual incidents) and polar continental air masses (25 %; Niedźwiedź et al. 2015). The Tatras also form a considerable barrier to air mass movements resulting in heavy rainfall events with 24-h sums of up to 300 mm (30 June 1973; Niedźwiedź 1992). Annual precipitation varies from 1100 mm at the foothills (Zakopane, 844 m a.s.l.) to 1660 mm at timberline (Hala Gąsienicowa, 1550 m a.s.l.) and 1721 mm on the summits (Lomnicky stit, 2635 m a.s.l.). The most effective precipitation events that result in flash floods are largely concentrated in the summer months as shown in Fig. 4.

Fig. 4
figure 4

Hydrometeorological situation at the northern foothills of the Tatra Mountains: a Average precipitation (cm) from 19 meteorological stations (28 years: 1954–1982); b Number of water flow larger than 1.5 times than averages as measured at Nowy Targ water gauge for the period 1898–1983 (Ballesteros-Cánovas et al. 2015a). After Ballesteros-Cánovas et al. (2015a), published by permission of Elsevier

Full size image

Vegetation at the study sites is mostly formed by subalpine forests of Norway spruce (Picea abies (L.) Karst.) at higher altitudes and a mixed forest with large proportions of P. abies, and Silver fir (Abies alba Mill.) at lower altitudes. In general, stream channels are formed in gravel and loamy moraine deposits, which cover granitic and pegmatitic bedrock (Bac-Moszaszwili et al. 1979).

The land use history of the Tatras starts with Medieval ore mining, through pasturing to intense logging in the 18th and 19th centuries. The local forests have been used intensively for grazing, with peaks in grazing pressure during the 19th and mid-20th centuries, changing the characteristics of soil, vegetation, and forests, and leading to an increase in flood risk. This region has been also traditionally affected by intensive logging associated with the steel industry in the second half of the 18th century. The Tatra National Park was enacted in 1954 but pasturing locally continued until 1978 and logging remains permitted, at least in some areas of the park. The long history of timber exploitation—as fuel for the local industry and as construction material—has substantially changed the character of forests in the study region. The use of the valley floor and stream channels as transportation routes furthermore intensified this process.

3.2 Temporal Reconstruction of Floods in Tatra Mountain Streams

The reconstruction of past flood events was carried out in six representative stream sectors, previously selected due to the lack of signs of human activity or the influence by other geomorphic processes (e.g., landslides, snow avalanches). A detailed geomorphic survey and inspection of trees were performed to select trees presenting suitable past evidence of flood activity. As stated above, work focused on scars on trees due to (i) their suitability to date scars using TRDs; and (ii) the utility of scar height to estimate peak discharge. Trees growing in suitable positions (i.e. exposed to the flow) and/or with visible damage were sampled using increment borers and by following standard dendrogeomorphic sampling procedures (Stoffel and Corona 2014). Additional cores, wedges or even cross-sections were taken from dead trees. Undisturbed specimens of P. abies and A. alba were also sampled to build 5 reference chronologies. We also recorded specific information from each tree such as tree height, remarkable growth characteristics, diameter at breast height (DBH), photographs and tree location (GPS).

After field data acquisition, samples were mounted on woody supports, sanded and polished. We scanned all samples with a resolution of 3200 dpi for imagery analysis using CooRecorder and CDendro (Cybis Elektronik and Data AB; Larsson 2003a, b). The reference cores were cross-dated against site chronologies employing both visual and statistical techniques. Mistakes and growth discrepancies (i.e., missing, wedging, and false rings) were identified and corrected. Growth disturbances (GD) were analysed and dated under a stereomicroscope, with a special focus on injuries and related features such as TRDs, including their location in the tree ring, and callus tissues. Other GD such as abrupt growth suppression or release as well as the occurrence of compression wood were only used to support the interpretation of events.

The definition of flood events was based on the weighted index proposed by Kogelnig-Mayer et al. (2011), which considers the number of GD as well as their intensity for each year. The threshold used to distinguish flash flood signals from noise was set to Wit ≥ 0.5 and GD ≥ 2, as suggested in previous work for hydrogeomorphic processes (e.g., Schneuwly-Bollschweiler et al. 2013).

A total of 1111 increment cores were sampled from 218 P. abies and A. alba trees affected by past flood activity. They allowed identification of 480 GD and definition of 47 flood events between A.D. 1866 and 2012 (Fig. 5). Our results reveal that the catchment presenting the largest floods activity is RP with 23 reconstructed events since A.D. 1866 (0.15 events year−1 with a sample size = 97), whereas the smallest activity was observed at DR with 4 reconstructed floods since A.D. 1926 (0.04 events year−1 with a sample size = 21). At DCH and ST, tree rings allowed the reconstruction of 19 (0.24 events year−1 with sample size = 73) and 9 (0.12 events year−1 with sample size = 22) since AD 1934 and 1938, respectively (Table 1).

Fig. 5
figure 5

Reconstructed chronology of major floods for the northern slopes of the Tatra Mountains and 5-year moving average of process activity highlighting periods of enhanced and limited flood activity. After Ballesteros-Cánovas et al. (2015a), published by permission of Elsevier

Full size image
Table 1 Overview of sample size (n), growth disturbances (GD; n) and weighted index (Wit) for each year of the time series and for the four streams analysed
Full size table

Our results also suggest that a large majority of dated events (82 %) was limited to individual catchments, and no years can be found with reconstructed floods in all catchments. The flood with the largest spatial replication took place in 1970 and was documented in DCH, ST, and RP, whereas in ten other years with floods signs were found in only two catchments. The temporal occurrence of floods in the Tatras also allows distinction of at least three periods with enhanced flood activity. As shown in Fig. 5, greater activity occurred between 1946 and 1949 (5-year average: 0.8 event year−1), 1955–1963 (5-year average: 0.7 event year−1), and 1979–1987 (5-year average: 0.6 event year−1).

3.3 Scar-Based Peak Discharge Reconstruction in Tatra Mountain Streams

After the analysis of flood occurrence in the Tatra Mountains (in terms of temporal frequency), we performed a reconstruction of flood magnitude based on a selected subset of trees presenting the most suitable evidence in the form of clear scars. The specific criteria to select the trees were (i) the presence of scars previously assigned to a specific flood event (see point 3.2); and (ii) scars facing the direction of flow and (iii) exhibiting a shape typical of flood impacts (see Ballesteros-Cánovas et al. 2010a, b for details). During fieldwork, each tree containing scars was accurately positioned with a GPS (Trimble GeoXT 6000, precision <1 m). We also measured scar height as well as the distance of trees with respect to the thalweg with a laser distance meter.

For each tree position, we computed peak discharge (q)–water stage (W) relationship W (q, i) by means of the two-dimensional hydrodynamic model IBER, which is a numerical model simulating turbulent-free, unsteady surface flows and solves depth-averaged, two-dimensional shallow water (2D Saint-Venant) equations using a finite volume method with a second-order roe scheme. In terms of digital terrain information we used LiDAR data with 1 × 1 m resolution as a basis for the mesh. Bed friction was evaluated using Manning’s n roughness coefficient, which was initially assessed in the field by using homogenous roughness units (Chow 1959). In this study, we used Manning’s n of 0.04 for the main channel and 0.1 for the overbank sections. To compute inlet water discharge (i.e. steady flow) into each study reach, we modelled successive inlet water discharges. At each tree, we used the best match between modelled water table and scar height to define the mean and standard deviation for the estimated average peak discharge of each flood event.

A total of 55 trees showed visible scars inflicted by sediment and/or wood transported during past flood events (namely 25 in DCH, 22 in RP, 6 in ST, and 3 in DR). Our results point out that the largest event took place in RP in 1903 with a reconstructed peak discharge of 115.9 ± 59.2 m3 s−1 (Table 2). By contrast, the smallest discharge of 11.1 ± 4.9 m3 s−1 was reconstructed for the 1934 event in catchment ST. This flood event was also dated in catchment DCH with reconstructed peak discharge of almost 39.7 ± 11.8 m3 s−1. On average, the range of discharges for reconstructed events was generally much higher in DCH and RP (92.8–39.7 and 115.9–28.6 m3 s−1, respectively) than in ST (45.8–11.1 m3 s−1) and DR (24.1 ± 7.6 m3 s−1). In terms of uncertainties related to the reconstructions, our methodology points to an average standard deviation of almost 41 %. The largest deviation (up to 80 %) was found for the 1997 event in ST, whereas the lowest deviation was observed at DCH and for an event in 1983 (~16 %). The detailed analysis of the impact of the hydrogeomorphic position of trees used in the reconstruction suggested that the trees showing lower deviation were located: (i) in sections with straight channel configurations or on the internal side of channel bends (10–20 %), (ii) far away from neighbouring trees (23 %), or (iii) next to the boundary of bankfull channel (~20 %). Trees showing the largest variability in scar heights were, by contrast, located on the external side of channel bends (up to 40 %), in areas with high tree density (41 %), as well as within the central channel or in overbank positions (up to 30 %).

Table 2 Peak discharge reconstructions based on scar heights and water levels as modelled for each stream reach and each reconstructed (i.e. tree-ring data) event considered
Full size table

4 Implications for Paleoflood Reconstructions in the Tatra Mountains

4.1 Climate Triggers and Flood Variability

The temporal flood reconstruction as well as the analyses of the location of TRDs within the tree rings have allowed an assessment of the seasonality of events, at least if they occurred during the growing seasons between late spring and early fall. Based on a comparison with records from the closest rain gauge stations, we described the potential hydrometeorological triggers of floods by analysing the maximum 1-, 3- and 5-day April–October rainfall events. We found that for 1-day events, rainfall totals ranged from 28.8 to 168.4 mm (mean 80.2 ± 34.3 mm), whereas the 3- and 5-day precipitation sums ranged between 50.3 and 247.7 mm (mean 130.4 ± 51.3 mm) and 65.2 and 258.1 mm (mean 149.4 ± 53.5 mm), respectively (Table 2). The statistical test used (nonparametric Friedman test) confirmed the existence of non-significant differences of rainfall threshold values between the catchments. The most intense event was recorded in DCH with a daily precipitation total of 333.5 mm (24 July 1980). The most replicated event (ST, RP, DCH) occurred in 1970 and has not only been confirmed by flood records in Zakopane, but also corresponds to the second largest recorded rainfall in the Tatra Mountains on 20 July 1970, when the 1-, 3-, and 5-day accumulated rainfall totals were 150.1, 206.4, 243.2 mm, respectively (Table 3).

Table 3 Precipitation records (1-, 3- and 5-day totals) associated with dated flash flood events
Full size table

These results confirm the observation by Niedźwiedź et al. (2015) who suggest that rainfall above the threshold of 50 mm day−1 can cause floods in the Tatras (in our study we provide an average value of 80.2 mm, but with a lower minimum at 28.8 mm). The seasonality and evidence of intense daily rainfalls during May and October are in agreement with local observations and previous studies (Kotarba 2004; Niedźwiedź et al. 2015). Rain-on-snow and sudden snowmelt due to warmer springs may trigger floods as well, as stated by Zielonka et al. (2008). However, our results suggest that the main weather mechanism leading to floods is related to prolonged rainfall induced by North cyclonic circulation (Nc) (Niedźwiedź 1992; Niedźwiedź et al. 2015). This situation implies generalized and widespread occurrence of heavy rainfalls, explaining the situations with floods such as the well-replicated event of 1970, for which we observe the highest daily rainfall sums in all meteorological stations. At the same time, we also observe a large number of asynchronous flood events between catchments. While differences in catchment characteristics may play a role in flood generation, we also think that local effects, such as the presence of rainfall shadow areas (i.e. orographic effect; Borga et al. 2014) may play an important role. This hypothesis is supported by the coefficient of variation on rainfall measurement observed in rain gauge stations along the Tatra Mountains. For instance, the year 1972 was characterized by prolonged rainfall, affecting mostly the Central and Western parts of the Tatra Mountains with maximum 3-day precipitation recorded at the Hala Gąsienicowa station (218.1 mm) and reconstructed floods in catchments DCH and ST (CV:038). This rainfall shadow effect has been described for other mountain areas as well (Buytaert et al. 2006). Our results, therefore, imply that significant space-time variability exists in catchment response.

Reconstructed peak discharges confirm dissimilar catchment responses as well. The specific discharge analyses of both gauged and reconstructed flow records clearly support the observed variability of catchment responses. On the other side, we also observe that reconstructed floods from the first half of the 20th century were generally larger than those measured by the gauging stations during the second half of the century, especially in catchments ST and DCH. These observations can be explained by the evolution of land uses in the Tatra Mountains. Therefore, the strong human impact from the eighteenth to the first half of the twentieth century resulted in increased soil erosion and surface runoff, thereby favouring the formation of braided channels. Our results may also support the idea that, in general, average peak discharges could have been reduced due to the impact of intense forest recovery on runoff in the study region during the second half of the twentieth century (Wyżga et al. 2012).

4.2 Implications for Flood Hazard Assessments

The paleoflood records provided in this study have implications in terms of flood hazard definition. Despite the fact that existing historical archives describe the occurrence of intense floods triggered by summer precipitation since at least the seventeenth century (Krzemień 1991; Starkel 1996; Kotarba 2004; Gorczyca et al. 2013), the existing gauging network in the Tatra Mountains and their foreland is highly fragmented and incomplete, which therefore renders proper hazard definition and makes long-term flood studies based on systematic records very difficult. Our results provide two main outcomes which are considered very useful for an improved understanding of flood hazards in the region.

First, we provide unique and quantitative rainfall thresholds associated with long-term flood occurrences. These rainfall thresholds should be used as a baseline for the determination of probabilities of threshold exceedances under future climate conditions. By using regional climate model (RCMs) output and downscaling procedures, the expected future frequency of intense precipitation events could be compared with our data so as to provide data on changes in the duration, severity, and frequency of intense precipitation events leading to the triggering of floods in the region over the next few decades. This is specifically relevant as evidence exists on an increase in the activity, at least in the long term, of cyclonic circulation types which are responsible for floods in the region (Niedźwiedź et al. 2015).

Secondly, the inclusion of non-systematic paleohydrological data, derived from tree-ring analysis, has an important impact on the results of the flood frequency analysis. Through the use of Bayesian Markov Monte Carlo Chain algorithms (Gaume et al. 2010; Viglione et al. 2013), we combined the reconstructed flood magnitude with the existing flow gauge series to determine changes in the flood frequency quartiles. Our results highlight that the largest changes in flood frequency (up to 25.5 % for T = 100 years) have occurred at Strążyska (ST) (Fig. 6). This implies that the flood hazard at this site would have been underestimated had the assessment of flood frequency distributions been based solely on systematic data. This finding is of high importance as the Strążyski Stream (ST) drains the catchment upslope of the main population center of the Polish Tatras, Zakopane. Consequently, our results should be taken into account for the design of hydraulic infrastructure and the definition of hazard planning. Our results also help to substantially reduce uncertainties in the estimation of the 100-year flood. Under a stochastic flood risk assessment approach, this uncertainty could be included in future risk assessments in the area, as it has been done in other regions in the past (Apel et al. 2004; Ballesteros et al. 2013).

Fig. 6
figure 6

Flood frequency distribution based on systematic flow gauge series (1) and the series consisting of systematic flow data and the reconstructed paleodischarges (2) After Ballesteros-Cánovas et al. (2016), published by permission of Elsevier

Full size image

Finally, and similarly to what is done on the southern side of the Tatra Mountains (Slovakian sector; Gáal et al. 2010), our paleoflood records could be included to deliver regional flood analysis, given that both datasets pass the homogeneity test based on the Hosking and Wallis (1997) algorithm. If so, a combined analysis on either side of the main divide could lead to a completely new setting for flood hazard assessments at the regional scale, which could then have clear and important implications for risk-based land management.

5 Conclusions

The lack of data on past floods and their impacts on the understanding of flood processes and their impacts in the Tatra Mountains and their foreland has been one of the key drivers of the FLORIST project (Kundzewicz et al. 2014b). Tree-rings studies, as the ones presented here, have been applied to Tatra Mountain streams to overcome some of the past drawbacks related with this lack of data. Despite some difficulties to deal with and to combine non-systematic (paleoflood) and systematic (gauged) records, the application of tree-ring records in Tatra Mountain streams has undoubtedly contributed to an improved understanding of past process activity, hydrometeorological triggers of floods; flood variability during the last century, and have facilitated redefinition of flood hazards in the region.