Water Regulation Ecosystem Services Following Gap Formation in Fir-Beech Forests in the Dinaric Karst

: This paper investigates how variation in forest structural characteristics affects the water retention capacity of gaps and forests in ﬁr-beech forests in the Dinaric Karst. Forests are identiﬁed as a key element of the landscape for provision of pristine water resources, particularly in highly vulnerable karst aquifers characterized by rapid inﬁltration of recharge water, high subsurface permeability, and heterogeneous underground ﬂow. Indicators of hydrologic ﬂuxes (drainage ﬂux, canopy interception, transpiration, and soil evaporation) in a large experimental gap (approximately 0.2 ha in size) and those in a nearby old-growth gap were compared over a 13-year period using the Brook90 hydrological model and their structural characteristics were analyzed. In addition, the hydrologic ﬂuxes were also simulated for a managed forest and an old-growth forest for reference. Water regulation capacity was lowest in the experimental gap, where drainage ﬂux accounted for 81% of precipitation and the sum of canopy interception, transpiration, and soil evaporation (evapotranspiration) accounted for 18%. This was followed by the old-growth gap, where drainage ﬂux accounted for 78% of precipitation and evapotranspiration for 23%. Water retention capacity was highest and generally similar for both forests, where 71–72% of annual precipitation drained to the subsurface. The results of this study suggest that the creation of large canopy gaps in ﬁr-beech forests in the Dinaric Karst results in signiﬁcant and long-lasting reduction in soil and vegetation water retention capacity due to unfavorable conditions for successful natural tree regeneration. For optimal provision of water regulation ecosystem services of forests in the Dinaric Karst, small, irregularly shaped canopy gaps no larger than tree height should be created, mimicking the structural characteristics of naturally occurring gaps in old-growth forests.


Introduction
Through their ecosystem services (ES), forests are identified as a key element of the landscape for provision of pristine water resources, especially in karst ecosystems [1]. ES are broadly defined as the benefits that humans receive from natural ecological processes [2,3], which are generally divided into provisioning, regulating, supporting, and cultural services [4]. Water regulation ES refer to the regulation of hydrologic flows [5,6] and are derived from the relationships between input (precipitation) and output (drainage fluxes). Water regulation ES of forests depend on development stage and structure [7,8], management interventions (managed vs. unmanaged) [9], and edaphic factors [10].
Scientists have repeatedly highlighted the link between the recent rapid loss of forest cover and the reduced water regulation capacity of forest ecosystems [11,12], increased erosion [13], and more destructive flooding, causing hazards and ecological problems as well as economic losses, especially in mountainous regions [14]. However, the effects of management measures on the water retention capacity of forests in gap-based silviculture have been poorly studied [15,16]. Moreover, there is little information on how water regulation ES of forests are affected after canopy release in karst aquifers, with their hydrogeological between input (precipitation; P) and output (drainage fluxes; DF). Sites with the lowest DF estimates and the highest sum of I, TRAN, and SE relative to P, often referred to as ecosystem evapotranspiration (ETP) [65], are assumed to have higher water regulation capacity.
Results of this study could have beneficial implication in hydrologically oriented forest management measures [8] for optimized provision of water regulation ES in karst aquifers [66,67].

Study Sites
The old-growth and managed fir-beech forests studied are located in the northern part of the Dinaric Alps in southeastern Slovenia (45 • 20 N, 14 • 30 E, 860-890 m a.s.l.). The diverse topography of this high karst plateau is characterized by numerous sinkholes, ridges, and slopes. The bedrock is Cretaceous limestone and the soils are shallow, welldrained Eutric Cambisols and Rendzic Leptosols [40]. The climate of the region is montane with annual precipitation up to 1600 mm. The long-term ) mean annual air temperature measured at the nearest meteorological station (Kočevje, 45 • 39 N, 14 • 51 E, 467 m a.s.l.) is 8.3 • C, which corresponds to a temperature of 5.9 • C at the study site (using an environmental lapse rate of 6 • C per km).
The specific old-growth Dinaric fir-beech forest Rajhenavski Rog was defined as a secondary virgin forest and officially declared as an old-growth forest in the first forest management plan of the area in 1892 [68]. It has a complex structure, being a mosaic of different forest development phases (optimal, terminal, declining, juvenile development phase [69], including standing dead trees, fallen logs, stumps, and roots [70] in different stages of decay of coarse woody debris [71]. The forest where the study was conducted was classified as Omphalodo-Fagetum association [72] and had a stem volume of 746 m 3 ha −1 and total basal area of 49 m 2 ha −1 in 2006 [73]. About 68% of the stem volume includes European beech (F. sylvatica L.) and 32% silver fir (Abies alba Mill.). Norway spruce (Picea abies (L.) Karst.), sycamore maple (Acer pseudoplatanus L.), elm (Ulmus glabra Huds.), and lime (Tilia cordata Mill.) comprise less than 1% of the total stem volume. The site faces south, has a slope of 10% and the predominant soil units are Eutric Cambisols and Rendzic Leptosols [74,75]. The average thickness of the O horizon is 3.4 cm (with standard deviation of 1.1 cm), the A horizon is 10.3 cm (with standard deviation of 4.7 cm), and the B horizon is 35.2 cm (with standard deviation of 6.4 cm). In the old-growth forest, a part of the stand and an irregularly shaped canopy gap (approximately 0.07 ha in size), which occurred naturally in the winter of 2002-2003 after windthrow, were used as study sites (Table 1). The canopy gap in the old-growth forest contained patches of different development phases, including dense natural beech regeneration.  [50]; 4 Average value for soil profile [42,49].
The specific managed Dinaric fir-beech forest is located nearby, at a distance of about 1300 m, and has similar elevation, aspect and slope. The predominant soils are also Eutric Cambisols and Rendzic Leptosols. The average thickness of the O horizon is 3.3 cm (with standard deviation of 1.3 cm), that of the A horizon is 16.2 cm (with standard deviation of 4.0 cm), and that of the B horizon is 23.6 cm (with standard deviation of 7.1 cm). In managed and old-growth forests about 30% of the area is bare limestone rock. Up to 40 cm of organic matter may be accumulated in holes and cracks. Similar to the old-growth forest, the managed forest was classified as Omphalodo-Fagetum association [72] and is dominated by European beech with 56% of the stem volume and silver fir with 34%. The rest of the stem volume consists of sycamore maple. A long tradition of sustainable forest management, systematic silvicultural planning, and more recently the development of close-to-nature management [29], with the idea of mimicking natural disturbance regimes and forest dynamics as closely as possible [31], have led to the relatively preserved vegetation composition of these forests [36]. They are traditionally managed with an irregular shelterwood system sensu Mathews [30], in which small canopy gaps are gradually increased and successful natural regeneration of European beech, sycamore maple, and silver fir in mixed forests is encouraged to ensure continuous forest cover and minimize soil degradation [47].
In the managed forest, a closed uneven-aged stand was selected and an irregular experimental clearcut gap (approximately 0.2 ha in size) was established in the winter of 2000-2001, as part of the EU FP5 project Nat-Man. All the trees in the large experimental gap were harvested and carefully removed by horse skidding. The ground vegetation was sparse and no higher than 0.1 m after gap formation. The large experimental gap was selected for this study because it was the only canopy gap where natural regeneration was not successfully established even 12 years after gap formation. Further details on the selected sites are presented in Table 1.

Indicators for Water Regulation ES
The water regulation capacity of gaps and forests was assessed by comparing indicators of hydrologic fluxes relative to precipitation in the open (P) using the water balance equation: where P is precipitation in the open; DF refers to all forms of drainage or runoff, e.g., unsaturated flow and saturated flow through the soil matrix and macropores; I is evaporation from wet canopy surfaces, such as intercepted water, TRAN is transpiration from the canopy and understory vegetation, and SE is evaporation from the ground or soil evaporation (SE). The sum of I, TRAN, and SE is often referred to as ecosystem evapotranspiration (ETP) [65]. SWC refers to soil water content and ∆SWC is the variation in soil water content, assumed to yield 0 in the long run. These hydrological indicators were simulated for each study site using the Brook90 water balance model [77,78] over a 13-year period from 2001 to 2013, except for the old-growth gap, for which the simulation ran from 2003 to 2013.

Meteorological Data and Soil Hydrological Measurements
Meteorological data were collected above the tree canopy using an automated weather station (Vantage Pro ® wireless, Davis Instruments, Hayward, CA, USA). Hourly mean air temperature and humidity, wind direction and speed, and hourly totals of precipitation were recorded [48]. Missing air temperature and humidity data were substituted with data from the Kočevje meteorological station and missing global radiation data were replaced with data from the Iskrba EMEP station (Environmental Agency of the Republic of Slovenia archives) using site-specific regression functions [39]. Using representative soil samples from the predominant soil units, the available water capacity of the mineral soil horizons was calculated for each study site from pressure plate measurements of field capacity (soil water content at 0.033 MPa). The permanent wilting point (soil water content at 1.5 MPa) and hydraulic conductivity of the saturated soil were also determined. The average soil thickness at all sites was 40 cm, and this depth was assumed to be the rooting depth for all sites. Soil water content (SWC) of the 0-to 40-cm layer was measured monthly at three locations in each site in 2003 and twice monthly in 2004 using time-domain reflectometry (TDR, Prenart Equipment, Frederiksberg, Denmark). Double probes were installed vertically and extended through the 40-cm layer, including the organic layer. Soil-specific calibration curves for vertical TDR probes were used, obtained by the calibration procedures described in Dirksen [79]. SWC was determined once a month in 2001 and 2002 by collecting three replicate volumetric soil samples at 10, 20, and 40 cm depths at each site and measuring weight loss after oven-drying the samples at 105 • C for 24 h. Volumetric water contents were converted to water depths (mm) by multiplying the values by the thickness of the soil layer and correcting for stone content.

The Brook90 Hydrological Model
The Brook90 model was chosen since it was found to be suitable for simulating hydrologic fluxes of stands and gaps in strongly structured forest ecosystems, complex karst morphology, and heterogeneous forest soils with small soil water storage capacity in previous studies [48,49]. However, the model is capable to deal with a dual permeability system [80], whereas a threefold permeability system cannot be represented by this model. The model calculates daily water fluxes (tree transpiration, canopy interception, throughfall, soil evaporation, snow evaporation, drainage (or runoff)) and soil water content at different depths or for the whole rooting depth. Tree transpiration and soil evaporation are calculated separately using the Shuttleworth-Wallace method [81], modified to separate evaporation during the day from evaporation during the night [77]. An important feature of the Brook90 model is that it only considers homogeneous plant cover at a selected site or catchment. Therefore, differentiation between forests, gaps, and regeneration stage was achieved through measured or preset input parameters corresponding to different cover type, plant height and density, leaf area index, interception, and other ecophysiological characteristics. More information on model parametrization can be found in [48,49].

Model Fitting and Testing
Site-specific parameters for running the Brook90 model (see Table S1) were derived by fitting model output with measured monthly throughfall data for managed and old-growth The goodness of fit was assessed by examining the linear correlation coefficient (r), which describes the degree of correspondence between measured and simulated values, the index of agreement (D) [82], which is a descriptive measure of relative error, and the root mean square error (RMSE), which expresses the error between the measured and simulated values (ibid.).

Statistical Analyses
Spearman's rank correlation coefficients (R) were used to evaluate the relationships between annual indicators of hydrologic fluxes (DF, SE, TRAN, and I) for gaps and forests. The nonparametric Kruskal-Wallis test was used to test for differences in annual DF, SE, TRAN, and I estimates among the four study sites, with estimates matched by simulation year. The Wilcoxon signed-rank test with Holm's correction was used for post hoc multiple comparison tests between the study sites. GraphpadPrism ® version 6.04, GraphPad Software MacKiev: La Jolla California USA, 2014. was used to perform the statistical analysis.

Results
Weather data showed considerable interannual variation from 2001 to 2013 (Figure 1). The annual P was lower than the long-term P  in 2003 (−22%) and 2011 (−30%), when the difference between P and reference evapotranspiration (Ref ETP) also indicated the highest water deficit (−63% in 2011 and −50% in 2003 compared to the long-term annual mean). This corresponds with the extreme drought conditions in summer 2003 as reported by several authors [83][84][85][86][87], which led to widespread disturbances in European forest ecosystems. P above the long-term annual mean was measured in 2004 (+11%) and 2010 (+17%), when the difference between P and Ref ETP also indicated the highest water surplus (+33% in 2004 and +50% in 2010). Mean annual air temperature was not as variable as annual P. Only in 2002 and 2007 were mean annual air temperatures higher than the long-term annual mean.

Model Fitting and Testing
Table S1 summarizes the input parameters used in the Brook90 model simulations. Monthly throughfall ( Figure S1) and daily soil water contents ( Figure S2) were well sim-

Model Fitting and Testing
Table S1 summarizes the input parameters used in the Brook90 model simulations. Monthly throughfall ( Figure S1) and daily soil water contents ( Figure S2) were well simulated in both time and magnitude for all sites). For throughfall fitting, the average D was 0.772 and average RMSE was 60.2 mm month −1 . For model testing, the average D was 0.772 and average RMSE was 60.2 mm month −1 (Table S2). Throughfall measurements were limited for financial reasons to monthly or biweekly recordings using a series of nine funnel collectors (240 cm 2 each) arranged along a regular grid with 5 m × 5 m spacing. Mean deviation of recorded throughfall quantities was 5% in the managed forest and ranged between −6% and 21%. In the old-growth forest mean deviation was 10% and ranged between −12% and 22%. In the old-growth gap, mean deviation of recorded throughfall quantities was 6% and ranged between −16% and 17%, respectively. In the experimental gap, mean deviation of recorded throughfall quantities was only 2% and ranged between −4% and 19%. To cover throughfall variability and assure spatial representativeness, a higher number of throughfall collectors is recommended [88], e.g., 35 to 40 collectors for heterogeneous uneven-aged single species and mixed forest of two or more tree species, with small canopy gaps or even more than 40 collectors in case of big canopy gaps. For soil water content fitting, the average D was 0.865 and the average RMSE was 13.0 mm day −1 . For model testing, the average D was 0.750 and the average RMSE was 15.6 mm day −1 (Table S2).

Indicators of Water Regulation Ecosystem Services (ES)
Annual DF estimates followed the pattern of annual P, with the lowest values in 2003 and 2011 (Figure 2), which were 75% of the average annual P from 2001 to 2013. DF estimates were higher in relatively wet years (2004 and 2010) for all sites, when annual P was greater than 115% of the 2001 to 2013 mean. heterogeneous uneven-aged single species and mixed forest of two or more tree species, with small canopy gaps or even more than 40 collectors in case of big canopy gaps. For soil water content fitting, the average D was 0.865 and the average RMSE was 13.0 mm day −1 . For model testing, the average D was 0.750 and the average RMSE was 15.6 mm day −1 (Table S2).

Indicators of Water Regulation Ecosystem Services (ES)
Annual DF estimates followed the pattern of annual P, with the lowest values in 2003 and 2011 (Figure 2), which were 75% of the average annual P from 2001 to 2013. DF estimates were higher in relatively wet years (2004 and 2010) for all sites, when annual P was greater than 115% of the 2001 to 2013 mean. Annual DF estimates were highest in the experimental gap (81% of P), followed by the old-growth gap (78% of P), old-growth forest (72% of P), and managed forest (71% of P) (Figure 3, Table 2). In the experimental gap, DF accounted for 1784 mm or 86% of P in wet years and 1028 mm or 77% of P in dry years. In contrast, in the old-growth gap, DF accounted for 1706 mm or 82% of P in wet years and 942 mm or 70% in dry years. Spearman correlation analysis showed that annual DF estimates for each of the four sites were significantly correlated (R > 0.918, p < 0.001), although the correlation was highest in the old-growth and managed forest cases (R = 0.999, p < 0.001). In wet years, DF accounted for Annual DF estimates were highest in the experimental gap (81% of P), followed by the old-growth gap (78% of P), old-growth forest (72% of P), and managed forest (71% of P) (Figure 3, Table 2). In the experimental gap, DF accounted for 1784 mm or 86% of P in wet years and 1028 mm or 77% of P in dry years. In contrast, in the old-growth gap, DF accounted for 1706 mm or 82% of P in wet years and 942 mm or 70% in dry years. Spearman correlation analysis showed that annual DF estimates for each of the four sites were significantly correlated (R > 0.918, p < 0.001), although the correlation was highest in the old-growth and managed forest cases (R = 0.999, p < 0.001). In wet years, DF accounted for 1581 mm or 76% of P in the managed forest and 1607 mm or 77% in the old-growth forest (Table 2), while in dry years it accounted for 878 mm (66%) and 896 mm (67% of P), respectively.    77  75  79  70  70  71  66  63  69  67  65  69  SE  13  11  15  3  2  4  2  2  2  2  2  2  TRAN  10  10  10  22  21  22  21  18  25  19  16  21  I  0  0  0  4  6  3  10  10  11  12  12  13  ETP = SE + TRAN + I  25  21  21  29  29  30  34  31  37  33  31  35 As expected, the experimental gap had significantly larger annual SE estimates (11% of P) compared to the forests (2% of P in the managed and 1% in the old-growth forest) and the old-growth gap (2% of P) (p < 0.001). For the old-growth gap, a statistically significant decrease in SE was shown during the simulation period (SE = −6.33 × year + 12,743, N = 11, R 2 = 0.645) due to shading of progressive natural tree regeneration. Statistical tests showed that annual SE estimates were significantly correlated only in the case of managed and old-growth forests (R = 0.975, p < 0.001). In dry years, the experimental gap and both forests had greater SE (170 mm or 13% of P in the experimental gap and 27 mm and 29 mm or 2% in both forests) than in wet years (193 mm or 9% of P in the experimental gap and 21 mm and 254 mm or 1% in both forests). SE in the old-growth gap was 60 mm in wet and 39 mm in dry years, accounting for 3% of P.
Annual estimates of TRAN were significantly lower in the experimental gap (7% of P) compared to the other sites (p < 0.001). TRAN accounted for 18% of P in the managed and 16% in the old-growth forest and gap. Annual TRAN estimates for forests and gaps showed no significant temporal trend and were significantly correlated only in the case of managed and old-growth forests (R = 0.929, p < 0.001). In wet years, TRAN was much lower than in dry years at all sites, accounting for 107 mm or 5% of P in the experimental gap and 266 mm or 13% of P in the old-growth gap. In dry years, TRAN accounted for 138 mm or 10% of P in the experimental gap and 292 mm or 22% in the old-growth gap. In contrast, TRAN accounted for 327 mm or 16% of P in managed forest and 287 mm or 14% in old-growth forest in wet years, while in dry years TRAN accounted for 287 mm or 21% of P in the managed forest and 248 mm or 19% in the old-growth forest.
Annual I estimates differed significantly among the four sites (p < 0.001) and were higher in the forests than in the gaps, as expected due to the presence of tall trees in the forests. I was highest in the old-growth forest (11% of P), followed by the managed forest (10%), and was significantly correlated in the case of both forest sites (R = 0.972, p < 0.001). In the old-growth gap annual I estimates accounted for 4% of P and statistically significantly increased during the simulation period (I = 6.77 × year − 13510, N = 11, R 2 = 0.802). Annual I estimates in the experimental gap were close to 0 mm in dry and wet years, as expected from the sparce ground vegetation that reached only 0.5 m in height 12 years after gap formation. In contrast, in forests, I estimates showed only a small increase in wet years compared to dry years, from 139 mm (10% of P) to 192 mm (9% of P) in the managed forest and from 166 mm (12% of P) to 224 mm (11% of P) in the old-growth forest.
The lowest water regulation capacity was present in the experimental gap, where annual DF was highest (81% of P) and the sum of annual I, TRAN, and SE referred to as ecosystem evapotranspiration (ETP), accounted for 18% of P. In wet years, DF was 756 mm or relative to P 9% higher and ETP was 7 mm or relative to P 9% lower in the experimental gap than in dry years. This was followed by the old-growth gap, where annual DF was 78% of P and annual ETP was 23% of P. In wet years, DF was 764 mm or 12% of P higher and ETP was 17 mm or 10% of P lower in the old-growth gap than in dry years. Water regulation capacity was highest and generally similar for both forests. Annual DF estimates were only 1% of P higher in old-growth compared to managed forest and were 180 mm or 10% of P higher in wet years compared to dry years in both forests. In addition, annual ETP estimates were lower by 1% of P in the old-growth forest compared to the managed forest. ETP was lower in wet years compared to dry years in both forests, with a difference of 89 mm or 8% of P in the managed forest and 92 mm or 7% of P in the old-growth forest.

Discussion
The results of this study indicate that the water retention capacity of the vegetation and soil in the large experimental gap was significantly lower compared to a gap in a nearby old-growth forest. The formation of a large experimental gap with a diameter greater than the height of the surrounding forest caused unfavorable environmental conditions that hindered successful natural tree regeneration for a long time [47]. The interception surfaces were unable to recover quickly after gap formation due to colonization by grasses and forest understory that benefit from high irradiance and nutrient leaching from the soils [15,89,90]. This was not the case in the irregularly shaped old-growth gap, where patches with different development phases, including dense natural beech regeneration were present. Studies in small-and medium-sized experimental gaps also reported dense natural tree regeneration as early as four years after gap formation [47], with hydrologic fluxes comparable to those in the old-growth gap [49].
Knowledge of water regulation ES of forests is a highly relevant goal of forest management to ensure multifunctionality and climate resilience [91][92][93]. Furthermore, disturbances associated with human activities and natural processes can have widespread and long-lasting impacts on hydrological, landscape, and ecosystem processes [94]. Reduced evapotranspiration in large canopy gaps will tend to increase drainage flux and contribute to reduced water retention capacity, as demonstrated in our study. The Brook90 model estimated that 71-72% of annual precipitation (1271 to 1281 mm) drained from the forest to the subsurface. In the experimental gap, almost all precipitation falls directly on the ground vegetation cover or soil, which havea smaller interception surface than the multilayered mixed foliage of fir-beech forests [31]. Accordingly, drainage accounted for up to 81% of annual precipitation (1452 mm) in the large experimental gap, indicating a long-lasting reduction in water retention capacity even 12 years after formation. The higher drainage flux estimates for the gaps compared to the two forests in our study are explained by the lower water use by vegetation in the gaps compared to forests [16,95]. In addition, the water holding capacity of the gap root zones is lower compared to forests [37]. Gap formation also increased water drainage in a mixed forest in Germany [16], in a European beech forest in Austrian Kalkalpen [96], and in a pure European beech forest in Denmark [89]. For the mixed forest in Germany, the difference in soil water content between forest and gap was due to lower canopy interception in the gap [16], while for the beech-dominated Danish forest it was found that the main cause of differences in soil water content between forest and gap conditions was the lack of transpiration in the gap [97]. In addition, the progressive growth of natural tree regeneration and edge trees resulted in increased canopy interception and water extraction by roots, which could modify the gap effect as early as the second year after gap formation to reach forest conditions (ibid.). This is consistent with the results of our study, in which patches with different development phases, including remnants of standing decaying trees and the progressive growth of natural tree regeneration in the old-growth gap [98], contributed to lower drainage flux (78% of annual precipitation or 1387 mm) than in the large experimental gap. Moreover, drainage flux in wet years was 82% of annual precipitation (1706 mm) in the old-growth gap, while in dry years drainage flux was only 70% (942 mm), indicating a higher water retention capacity compared to the large experimental gap. Katzensteiner [27] also reported lower drainage values for beech regeneration (85% and 75% of seasonal precipitation in 1996 and 1997, respectively) compared to canopy gaps (90% and 81% of seasonal precipitation in 1996 and 1997, respectively). Nevertheless, small and medium-sized canopy gaps in fir-beech forests with diameters smaller than the height of the surrounding forest ensure suitable conditions for successful establishment of natural tree regeneration as early as four years after canopy gap formation [47], resulting in similar drainage fluxes and evapotranspiration estimates as in the old-growth gap [49].
Differences in water retention capacity between experimental and old-growth gaps were exacerbated in the dry years of 2003 and 2011 when extreme drought led to widespread disturbances in European forest ecosystems [83][84][85][86][87]. The annual P measured at the nearest meteorological station was 30% lower than the long-term P  in 2003 and 22% lower in 2011. Drainage flux in the large experimental gap accounted for 77% of annual precipitation (1028 mm) in dry years, while it accounted for only 70% (942 mm) in the old-growth gap. Soil evaporation was 10% of P higher in the experimental gap than in the old-growth gap, while transpiration and interception were higher in the old-growth gap, as expected. Observed records and climate simulations of some meteorological variables (e.g., temperature, evapotranspiration) [99] and shifts in precipitation patterns [100] show increasing trends in the intensity, duration, severity, and spatial extent of droughts worldwide [101]. The Dinaric Karst region has also experienced multiple severe and prolonged droughts over the last century, which, together with secondary insect damage, have led to episodes of forest decline [102]. This should be considered when developing forest management strategies for climate change adaptation [103,104].
Water regulation capacity of managed forest and old-growth forest was generally similar and higher compared to both gaps. In the managed forest with the currently practiced traditional shelterwood management system, annual drainage flux was 71% of annual precipitation (1271 mm) and 72% (1281 mm) in the old-growth forest, while evapotranspiration was 29% of annual precipitation (510 mm) in the managed forest and 28% (500 mm) in the old-growth forest. Estimated values of the drainage flux for the forests were similar to those reported by Katzensteiner [27] for a European beech forest in Austrian Kalkalpen, with drainage flux values corresponding to 75% and 65% of growing season precipitation in 1996 and 1997, respectively. In wet years, drainage flux was 10% of annual precipitation higher and evapotranspiration 7% (managed forest) and 8% (old-growth forest) of annual precipitation lower in both forests than in dry years.

Conclusions
In fir-beech forests in the Dinaric Karst, the loss of leaf area due to large canopy gap formation significantly reduces the water storage capacity of soil and vegetation over the long term. The unfavorable environmental conditions for natural tree regeneration even 12 years after the formation of the experimental gap resulted in lower evapotranspiration and higher drainage flux compared to a gap in a nearby old-growth forest. Patches of different development phases, including remnants of standing decaying trees and progressive growth of natural tree regeneration in the old-growth gap showed higher transpiration and canopy interception, and lower soil evaporation and drainage flux compared to the large experimental gap.
For optimal provision of water regulation ES of forests in the Dinaric Karst, continuous cover forestry should be promoted, emphasizing the size and shape of canopy gaps as a critical factor for successful natural tree regeneration. Small, irregularly shaped canopy gaps no larger than tree height should be created, mimicking the structural characteristics of naturally occurring gaps in old-growth forests.
Supplementary Materials: The following are available online at https://www.mdpi.com/1999-490 7/12/2/224/s1, Table S1: Values for selected parameters used in the Brook90 model simulations; Figure Table S2: Linear regression (y = a + b × x) coefficients, linear correlation coefficient (r), index of agreement (D), root mean square error (RMSE), and sample size (n) describing the goodness-of-fit between Brook90 model simulated (y) and measured (x) daily values for the soil water contents (0-40 cm layer) and monthly throughfall (mm) for two calibration periods.
Funding: This research was funded by the Slovenian Research Agency within the project "Infiltration processes in forested karst aquifers under changing environment" (no. J2-1749) and the core funding for the Program Group "Forest biology, ecology and technology" (P4-0107).