Damir Ugarkovic′ · Nikica ?prem · Nikolina Kelava Ugarkovic′ ·Milan Or?anic′

Abstract Floodplain forests with regular flooding regimes are the largest natural retentions areas in Croatia and are important as natural habitats for ungulates.The aim of this study was to determine the scale of mortality caused by flooding within these forests. Over a 10-year period, data on ungulate mortality (red deer, roe deer and wild boar),flood duration and flooded surface area were recorded.The study was conducted in primary (Posavske ?ume-RET I)and secondary (Opeke II-RET II) retention areas within Lonjsko Polje Nature Park (Sava River region, Croatia).The longest flood period and the largest flooded surface area were recorded in RET I.Total ungulate mortality was 749 individuals, with 482 individuals in RET I and 267 individuals in RET II, predominantly wild boar. Flood mortality did not differ by gender.The highest mortality of wild boar was recorded for the juvenile and yearling age classes.Low mortality of red and roe deer can be attributed to their body size and ecological niches. Differences in mortality between the primary and secondary retention areas corresponded to differences in flood regimes, flood column heights and micro relief structures. In both retention areas, wild boar mortality and flood duration, i.e.flooded surface area, were positively correlated. Because the growth rate of the analysed ungulate populations was higher than the recorded mortality, no long-term effect of floods is expected on species abundance in these areas.

Keywords Floods · Floodplain forests · Mortality · Red deer · Roe deer · Wild boar

Introduction

Climate change has contributed to greater frequency of extreme weather events,such as massive flooding,which has recently intensified(Milly et al.2002;Mudelsee et al.2003;Hirabayashi et al. 2013). Extreme floods affect natural wildlife populations, either by causing direct mortality or through habitat destruction(Wuczyn′ski and Jakubiec 2013).The literature on wildlife responses toflooding events has focused on small mammals(Andersen et al.2000;Williams et al.2001;Jacob 2003),though some studies have addressed the responses of large mammals toflooding (MacDonald-Beyers and Labisky 2005; Wuczyn′ski and Jakubiec 2013;O’Connell-Goode et al.2014).In Florida(USA),the annual death rate of adult white-tailed deer during a flood year was approximately double that of pre-flood years.Water depths of 50 cm have a significant negative impact on deer populations(MacDonald-Beyers and Labisky 2005).The extreme flood of the Odra River in July 1997 caused massive mortality of 3613 game mammals,with deaths of roe deer and wild boar increasing in proportion to the area flooded and duration of flooding (Wuczyn′ski and Jakubiec 2013). The 2011 flooding of the Mississippi River,Louisiana(USA)had minimal effects on survival and site fidelity of black bears,while the same event caused numerous mortalities of deer and feral hog(O’Connell-Goode et al.2014).The Sava River is a major drainage basin of southeast Europe and the greatest tributary of the Danube River.It runs 945 km in length,and the catchment area covers 97,713 km2.Flooding is seasonal,but typically occurs in spring following snow melt in the Alps and in autumn after heavy rainfall.Such a hydrological regime is the norm,but the frequency of storms,droughts and floods are predicted to increase from intervals of 50 years to intervals 5-10 years, while intensity is also predicted to increase (Anonymous 2009). Therefore, along the river course in Croatia, flood control systems have been constructed,relying on three relief canals,two basic water distribution facilities, and five large lowland retention areas(Lonjsko Polje,Mokro Polje,Kup?ina,Zelenik,and Jantak)(Karad?ic′et al.2015).The largest retention area and wetland along the Sava River is Lonjsko Polje,which was declared a nature park in 1990.The floodplain forests here flood regularly,with flood waters persisting for long periods of time.Therefore,interaction between floods and animals is a normal occurrence, though in some years this can cause high rates of mortality,such as in 1965 when more than 2000 deer drowned(Vrataric′ et al.2005).Flood frequency and intensity are predicted to increase with climate change, and the aim of this study was to determine whether this would affect mortality rates of ungulates in the floodplain forests of Lonjsko Polje.

Materials and methods

Study area

The study was conducted in the floodplain forests surrounding the Sava River in Croatia i.e. in the lowland retention area of Lonjsko Polje Nature Park(N 45°20′47′′E 16°47′11′′). The study was conducted at two sites: the first in the Posavske ?ume forest as the primary retention(RET I), and second Opeke II as the secondary retention area(RET II)(Fig. 1).In waterway engineering,retention areas are built to block high water during heavy rains or a sudden upstream melting of snow, in order to prevent of floods in urban, industrial, rural, and agricultural areas (?imunic′2016). The study area is 60% forested, with about 30%covered by wet meadows at elevations between 90 and 110 m. Forests are composed mainly of pedunculate oak(Quercus robur), narrow leaf ash (Fraxinus angustifolia)and common alder (Alnus glutinosa). The climate is continental,moderately warm,and moist with mean annual air temperature of 11 °C and mean annual precipitation of 850 mm (Seletkovic′ and Tikvic′ 2005). The main ungulate species in the study area are red deer(Cervus elaphus),roe deer(Capreolus capreolus)and wild boar(Sus scrofa).The study area is also inhabited by mesocarnivore species:golden jackal (Canis aureus), red fox (Vulpes vulpes) and wild cat (Felix silvestris).

RET I covers an area of 11,346 ha,while RET II covers 8054 ha. The dominant microrelief at both analysed retentions are micro depressions (RET I-60%, RET II-64% of the total retention area). In RET I, microplateaus represent 12% of the total area and wet microdepressions 28% (Abramovic′ 2016). In RET II, wet microdepressions represent 14% of the total area, while microplateaus 22%(Tomljanovic′ 2016).

Data collection

Data on ungulate mortality after flooding of the Sava River were obtained over a 10-year period (2005-2014). Based on the Croatian hunting legislation, all hunting organizations are required to record wildlife mortality occurring within their respective hunting grounds (Anonymous 2005). Mortality data were accessed from game management plans and from gamekeeper observations.Surveys for dead animals were conducted during the first 2-5 days after each flood, by foot or from boats. Mortality events were recorded and described by species, gender, age, and location. Animal ages were estimated using replacement and tooth eruption patterns(Boitani and Mattei 1992).Age classes were represented as juvenile (＜1 year), yearling(1-2 years) and adult (＞3 years). Local abundance of ungulate species was estimated during the game census.Applied methods for the population census in the study areas were: (1) year-round observation by the professional gamekeeper, and(2)drive counts twice per year (in spring and autumn) where about 10% of the total area was covered. Based on the same census data, annual growth rates were calculated for the analysed species.

Fig. 1 Map of study area indicating retention area I (Posavske ?ume) and retention area II (Opeke II)

Flood data were obtained for the study period from the Croatian Meteorological and Hydrological Service, Croatian Waters Ltd., and gamekeeper records. Flooded area(km2) was obtained during flood events by survey from boats and from reference to maps and GPS.The number of flood days was recorded as the cumulative sum of all floods(flood days) in 1 year. The impact of flooding on ungulate mortality was observed through three types of floods(flash floods, river floods and storm surge), with river floods and flash floods as most frequent. Floods came directly from the Sava River and indirectly from its tributaries Lonja,Crnava and Strug during the high(maximum)water levels.We studied the influence of natural floods and extreme flooding events, as ‘‘a(chǎn)rtificial’’ floods when a surplus of water was diverted into a retention area or floodplain forest. The combined influence of natural and artificial floods in retentions was examined, as they appeared together in the same area.Spearman correlation coefficients were used to analyse the relationship between mortality, flood duration and flooded areas in the retentions. Differences in mortality and density were tested with the Mann-Whitney U test. Jacobs selectivity index (Jacobs 1974) was calculated according to the following formula: Jacobs index =(r - p)/(r - 2rp + p), where ‘‘r’’ represents the proportion of mortality of particular species and ‘‘p’’ represents the proportion of abundance of particular species compared to the total abundance of the three studied species. The index ranges from - 1 (lack of flood-related mortality) to+ 1 (presence of flood-related mortality). Indices near or equal to zero indicate that species mortality is in proportion to its abundance (Wuczynski and Jakubiec 2013). Data were analysed using SAS/STAT software, V9.4. (SAS Institute Inc., 2002-2014).

Results

Total mortality during the 10-year period of all analysed ungulate species in both retentions was 749 individuals.In RET I, observed mortality was 482 individuals (417 wild boar, 53 roe deer and 12 red deer), while in RET II, mortality was 267 individuals (263 wild boar, three roe deer and one red deer). Mortality in RET I was highest during 2013 and lowest during 2005, while in RET II, it was highest during 2014 and lowest during 2006-2008. Flood duration in RET I was longer than in RET II (97 vs.51 days; p ＜0.001). In RET I, extreme flooding occurred in 2010(three floods during spring and one in autumn)and in RET II in 2014(twofloods in spring and one in autumn).The surface area under flood waters was larger in RET I than in RET II (59 km2vs. 34 km2; p ＜0.001). The most abundant species (p ＜0.01) in the combined retentions was wild boar. Roe deer was more abundant in RET I(p ＜0.01), while red deer was more abundant in RET II(p ＜0.01) (Table 1). Higher mortality (p ＜0.01) of the analysed ungulate species was recorded in RET I than in RET II (Table 1). Wild boar were the most vulnerable ungulate species toflooding in both retention areas(21 and 13; p ＜0.01). During the study period, the lowest mortality was found in RET II for red deer and roe deer.Growth rate during the study period in both retentions was higher than the recorded mortality (RET I 231 ± 31; RET II 289 ± 59; p ＜0.0001).

Mortality of ungulates was positively correlated(p ＜0.0001) with flood duration in RET I and positively correlated (p ＜0.05) with flooded area in RET II(Table 2). Wild boar mortality was positively correlated with flood duration in RET I (p ＜0.0001) and in RET II(p ＜0.05).Also,surface of the flooded area was positively correlated with wild boar mortality in RET I (p ＜0.001)and in RET II (p ＜0.05). Mortality of red deer and roe deer was unrelated toflood duration or surface of flooded area in both retention areas (Table 2).

Table 1 Abundance and mortality of analysed ungulate species (mean ± SD)

Table 2 Spearman correlation coefficients between mortality of ungulate species and flood duration (days) or surface of flooded area (km2)

Red deer and roe deer had a higher (p ＜0.0001;p ＜0.001) density of mortality in RET I than in RET II(Table 3). Densities of mortality for wild boar and all analysed ungulate species were similar for the two retention areas (Table 3).

The observed mortality of juvenile and yearling wild boar was positively correlated with flooded area in RET I(p ＜0.0001). In RET II, mortality of yearling wild boar and flooded surface area were a positively correlated(p ＜0.05) (Table 4).

Table 3 Density of mortality of analysed ungulate species(mean ± SD)

Table 5 presents the differences in mortality of ungulates caused by flood per age class. Juvenile red deer and roe deer showed higher (p ＜0.0001; p ＜0.001) mortality in RET I than in RET II, and juvenile wild boar had the highest mortality. The yearling category of roe deer and wild boar showed higher (p ＜0.0001; p ＜0.001) mortality in RET I than in RET II, with wild boar having the highest mortality. Adult red deer and wild boar showed higher (p ＜0.001) mortality in RET II than in RET I,while higher (p ＜0.0001) mortality of adult roe deer was recorded in RET I than in RET II. Also, no mortality was recorded for juvenile red deer or roe deer in RET II,yearling red deer in RET I,or yearling roe deer in RET II.Mortality rates did not differ by sex for any of the analysed ungulate species (RET I, U test = 0.0014, p ＜0.9698;RET II U test = 0.1194, p ＜0.7297).

High negative values of the Jacobs index showed low susceptibility toflood mortality for the analysed ungulate species in RET I. Although the highest rate of flood mortality was recorded for wild boar, high species relative abundance resulted in a high negative Jacobs index. Low mortality rate of red deer and roe deer in RET I was confirmed by the Jacobs index (Fig. 2). Low susceptibility toflood mortality was revealed by the Jacobs index in RET II(Fig. 2). This was due to high species abundance and low observed proportion of losses (Fig. 3).

Table 4 Spearman correlation coefficients between mortality of ungulate species per age category and retention flooded surface

Table 5 Mortality of ungulate species per age category in analysed retentions (mean ± SD)

Discussion

The biotope of a floodplain region is usually well adapted to the natural phenomenon of high waters. Anthropogenic interventions(construction of retention areas)in floodplain forest areas alter flood dynamics and ungulate species often cannot readily adapt to these new conditions(Vrataric′et al.2005). Natural floodplain forests typically flooded in Croatia outside the vegetation period (Prpic′ 2005) and ungulate species were threatened when floods occurred.Today, the maximum flood height in the forest is considerably higher than during former flood periods, and under the present conditions, the flooded forest area has expanded. Flood waters also remain in retention areas during vegetation periods (Prpic′ 2008). These changes in flood regime threaten ungulates to a greater extent than in preretention basin times. Differences in mortality of the analysed ungulate species between RET I and RET II corresponded to differences in the flood regime, flood columns, micro relief and landscape structure. Maximum flood columns in RET I reached up to 7.35 m, while in RET II they reached 5.40 m (Prpic′ et al. 2005). Micro relief and landscape structure characteristics were seen in the fact that wet micro-depressions accounted for twice as much area in RET I as in RET II. Also, RET I had significantly larger flooded surface area than RET II,explaining the higher mortality there.

Wild boar was the most vulnerable toflood-caused mortality of the three ungulate species studied in both retention areas. Wild boar prefers forested habitats and display remarkable plasticity in spatial and activity patterns(Boitani et al. 1994). Among other factors, weather conditions also affect the duration and distribution of wild boar activity (Lemel et al. 2003). Marshes are the most frequently used areas year-round (Dardaillon 1986). These marsh areas, especially micro-depressions and wet microdepressions are regularly flooded. Therefore, flood events cause mortality of wild boar in the retention areas of floodplain forests. The highest water levels and longer flood duration was during spring, coincident with the reproduction period of all three ungulate species.However,the lower mortality rates of red deer and roe deer were likely related to body size(Wuczynski and Jakubiec 2013).The higher mortality(Table 4)of wild boar could be due to the smaller size of younger age classes of this species,and their inability to escape suddenly rising water and high water flow energy in micro and wet micro-depressions(Vrataric′ et al.2005).Wuczynski and Jakubiec(2013)also reported higher mortality rates of young and female animals than older or male animals caused by floods.The high mortality rate among young animals is generally due to small body size, low resistance and lack of experience(Vrataric′ et al. 2005; Welbergen et al. 2008). As reported,animal size appears to be crucial during the initial phases of the flood (Wuczyn′ski and Jakubiec 2013). Small animals, like piglets, are less likely to survive highly rising floodwaters or a sudden strike, or long swimming in floodwaters (Vrataric′ et al. 2005).

Among the ungulates, wild boars have the highest fertility (high number of foetuses), and have early sexual maturity and a short gestation period(Servanty et al.2009).Adult female survival can be considered the main influence of wild boar population dynamics (Fonseca et al. 2011).According to ?prem et al.(2016),the percentage of female pregnancy in Croatian wild boar population is over 86%,with nearly 65% of juvenile females pregnant and a mean of 6.02 foetuses per animal. Species with a high reproductive output have more variable annual survival (Gaillard et al. 2000), suggesting they react more to severe conditions, but are more likely to recover quickly (Mysterud and S?ther 2011). Also, wild boar numbers are increasing consistently throughout Europe and in Croatia,and hunting is the main cause of mortality for this species(Massei et al. 2015).

Fig. 2 Jacobs index as an indicator of susceptibility toflood mortality caused by drowning

Fig. 3 Abundance and observed mortality of the analysed ungulate species in retention I and retention II

The calculated Jacobs index for the analysed ungulate species yielded high negative values, indicating low susceptibility toflood mortality. This can be attributed to the low observed mortality in relation to species abundance and annual growth rates, especially for red deer and roe deer. The primary retention area had significantly higher mortality rates of these species than the secondary retention area, due to the higher water levels and longer flooding periods.However,wild boar is characterised by the highest reproductive rate,with annual populations growth rates that may exceed 2.0(Bieber and Ruf 2005;Keuling et al.2013;Massei et al. 2015). In this study area, the annual growth rate of wild boars was from 2.5 to 3.0 (Abramovic′ 2016;Tomljanovic′ 2016). Due to the annual growth rates that were higher than the observed mortality, long-term negative effects of floods on ungulates are not predicted.

Wildlife species have strategies that allow them to survive negative effects of such extreme events (Finkelstein et al.2010;Gerisch et al.2012).Gamekeepers reported that some individuals survived floods by migration or moving to higher ground,i.e.microplateaus.In this study areas,this includes forests of pedunculate oak and horn beam that are typically not flooded. Due to its ecological niche, the mortality of roe deer caused by floods was low. Since roe deer favours agricultural forest settings, moist habitats covered with lush swampy vegetation are far from ideal in satisfying its biological and ecological requirements(Grube?ic′ and Krapinec 2005), and therefore roe deer tend to avoid such habitats.Red deer finds wet microdepressions extremely favourable since these natural wallowing sites provide the necessary defence against insect invasions in the summer months (Grube?ic′ and Krapinec 2005; Schlaghamersky and Hudec 2008). During summer,floods are very rare and therefore the mortality of red deer was also very low. Changes in site ecology by flooding affect red deer behaviour, as the animals change their rhythm and behavioural patterns, such as frequent migrations during flood conditions.

Conclusions

Floods in retention areas of the floodplain forests in Croatia cause ungulate species mortality. Mortality caused by floods was higher in the primary than in the secondary retention. In both retention areas, flood duration and flooded surface area significantly affected rates of wild boar mortality, while floods seem to have a minor effect on red and roe deer survival. Juvenile categories of all analysed ungulate species were the most threatened.

Natural floods and extreme artificial floods did not threaten long-term survival of ungulates occupying floodplain forests in Croatia. Although wild boar was most threatened by floods, the local population growth rate was significantly greater than the rate of flood-caused mortality.Thus, floods caused no lasting effects on local wild boar abundance. Floodplain forests, despite floods, provided suitable habitats for ungulates.

Monitoring of flood-based mortality, recruitment and survival in all ungulates populations is recommended,with particular emphasis on wild boar populations.

AcknowledgementsThe authors would like to thank to the gamekeepers from Posavske ?ume and Opeke II for their assistance in field work. Also, thanks to Linda Zanella for providing language editing and proofreading.

Conflict of interestThe authors declare that they have no conflict of interest.

Author’s contributionsD. U designed the study. D. U and N. ? wrote introduction and discussion, while N. KU analysed data and wrote results. M. O revised the manuscript. All authors read and approved the final manuscript.