Prioritization of spatial protection of Šumava National Park, Czechia: comparing reality and the model

ABSTRACT

Natural processes and biodiversity are under pressure, especially in the human-dominated region of Central Europe. Therefore, an effective approach is needed for the conservation of the last fragments of valuable habitats. Šumava National Park is one of the biggest relatively natural forested areas within the cultural landscapes of Czechia and Central Europe. It has been protected since 1991 as a National Park, but management has been changed several times; thus, we would like to create a map with nature protection prioritization based on objective procedures and data to help manage the National Park with regard to its goals. We used occurrence data for 50 selected key fauna and flora species and made habitat suitability models; furthermore, data derived from the habitat mapping layer describing the quality of habitats were prepared. These data were inputs for prioritization ZONATION software resulting in a prioritization map, which could be compared with the current zonation of the National Park.

KEYWORDS:

• Prioritization
• zonation
• management
• protected areas
• Šumava National Park
• Czechia

Key Policy Highlights

• We propose a method for appropriate delimitation of zonation as a key tool of management within protected areas.

• We deploy environmental variables, species occurrence data, and data on habitat quality for delimiting priority rankings.

• We delimit prioritization ranking within Šumava National Park, the largest Czech national park with a turbulent history of management.

1. Introduction

Nowadays, nature, natural landscapes, and biodiversity are under pressure worldwide. In Central Europe, habitat loss, landscape fragmentation, intensifying land use, and anthropogenic pressure are major negative processes influencing and threatened biodiversity (Di Minin & Moilanen, 2012; Gordon et al., 2009; Moilanen, 2013; Montesino Pouzols et al., 2014; Romportl, 2017; Tscharntke et al., 2012). Spatial nature protection should protect biodiversity and nature processes at the landscape scale; it deals with appropriate and effective delimitation, zonation and management of selected areas (Lehtomäki & Moilanen, 2013). This is something which is demanded in the highly human-transformed European landscape, where many interests (e.g. economic, environmental, social) compete with each other and where a balance resulting in sustainability is desired (Kušová et al., 2008). Thus, trade-offs between nature protection and other land uses are an inherent component of landscape management (Di Minin et al., 2013).

National Parks (NPs) are here for extensive nature protection and to face the above-mentioned threats (Soliku & Schraml, 2018); but there are several problems and conflicts. NPs are often designated and managed not based just on knowledge and scientific evidence, but on political or other interests and bases (Müller & Opgenoorth, 2014). Therefore, bridging this gap between science and practice in nature protection is crucial for NP functioning. Proper information, data, discussion, and an agreement on the visions, objectives, and management of an NP are a necessity (Kemkes et al., 2008) for preventing conflicts, which are related to other human activities within an NP (Soliku & Schraml, 2018). Therefore, not only protection but also the development of an area must be considered, especially in the geographical context of Central Europe with its relatively high population density (Brown et al., 2015).

The landscape of Central Europe has been shaped by humanity for centuries and now is more or less cultural and inhabited; thus, conflicts between nature protection and human activities occur more often (Soliku & Schraml, 2018; Watson et al., 2014). The Bohemian Forest, split between Austria, Czechia, and Germany (Bavaria) is an exceptionally large forested area with low human impact within the region (Křenová & Kiener, 2012). This area provides a home for many endangered species such as lynx (Lynx lynx), otter (Lutra lutra), and capercaillie (Tetrao urugallus) (Křenová & Hruška, 2012). However, the protection of the area differs. In Germany, Bavarian Forest National Park (BFNP) has existed since 1970 and zonation and management is clear, to reach 75% natural zone by 2027 (BFNP, 2018). Austria has the smallest part of the mountain ridge and it is not protected as a NP. On the Czech side, Šumava National Park (ŠNP) was established in 1991 (Bláha et al., 2013).

According to IUCN (The International Union for Conservation of Nature), category II, a National Park’s primary objectives are to protect natural biodiversity and support environmental processes, as well as to promote education and recreation (Dudley, 2008). Nevertheless, ŠNP has gone through turbulent development of nature protection. Management and zonation have been a topic for public debate since the beginning of the ŠNP. Not just environmental but also social, economic, and political interests and factors have played a role in the decision-making process. In 1991, the most valuable and the most protected ‘first zone’ covered 22% of ŠNP and was fragmented into 54 patches. In 1995, the first zone shrunk to 13% and became more fragmented, with 135 patches; the second zone had 82% and the third zone 5%. In addition, more intervention management was applied during bark beetle outbreaks (Křenová & Hruška, 2012; Šantrůčková et al., 2010). This does not fully correspond with the objectives of NPs and can have a negative influence on the migration, distribution, and abundance of endangered species (Turner, 1989). In 2020 a new zonation came into force, based on an amended law. The natural zone has 27.7%, the close-to-nature zone 24.6%, the concentrated management zone 46.5%, and the cultural landscape 1.2%. ŠNP is still threatened by human activities: the timber industry, recreation, and road infrastructure and their increasing intensity are among them (Křenová & Kindlmann, 2015; Zýval et al., 2016).

After roughly thirty years of turbulent history ŠNP, there is a chance for widely accepted future nature protection of the area. Therefore, as a material for evidence-based management and verification of recent decisions regarding management and zonation of ŠNP, we prepared prioritization of spatial protection within ŠNP, which is based on ecological modelling and mathematical optimization (Gordon et al., 2009) using objective and detailed data and possibly repeated performances by employing potentially freely available software (QGIS, MaxEnt, ZONATION). We would like to pay attention to zonation as a key tool for spatial nature protection and also to the verification of whether valuable parts of the area are treated well according to the objectives of the reserves design tool. A prioritization tool was used here not for delimiting new reserves, but for the evaluation of a thirty-year-old NP, which was treated in many different ways in the past, and therefore, this evaluation is desired.

2. Material and methods

2.1. Study area

ŠNP is situated in southwest Czechia, along the borderline with Bavaria and Austria. It covers 683.4 km² and is surrounded on the Czech side by the Šumava Protected Landscape Area (ŠPLA) as a buffer zone of the lower level of protection. On the German side, it is partly neighboured by Bavarian Forest National Park (BFNP) and Bavarian Forest Nature Park.

2.2. Data and methods

In the first step, we made habitat suitability models (HSM) as inputs for ZONATION. We gathered all relevant data describing the biotic and abiotic characteristics of ŠNP and influencing habitat suitability for key species. Data included characteristics on geology (prevailing rock formation (Czech Geological Survey, 2022)); geomorphology (altitude, slope, vertical heterogeneity (State Administration of Land Surveying and Cadastre, 2021)); pedology (soil moisture (Brůna et al., 2021)); hydrology (water courses and bodies, mires and bogs, and distances from them (Fojtík et al., 2022)); climatology (annual mean precipitation, temperature, solar radiation (Brůna et al., 2021; Czech Hydrometeorological Institute, 2022)); landscape (land cover type (AOPK ČR, 2013)); forest type (Forest Management Institute, 2022) structure of vegetation derived from LiDAR data (cover of canopy, height of vegetation, percentiles of points from LiDAR by Naesset, number of tree species within pixel 100 × 100 m derived from object-based image classification (ŠNP, BFNP, 2021)); and anthropogenic activities (distances from the roads, settlements, touristic paths, ski slopes (State Administration of Land Surveying and Cadastre, 2021)) (supplemental material 1).

Then, fifty species were chosen for analysis. These are subjects of protection for NATURA 2000 sites overlaying ŠNP and species recognized as important for ŠNP. They were divided into functional groups according to their taxonomic membership: amphibians, reptiles (3 species); birds (14); fungi (1); insects (10); lycopodium, fens, mosses (4); mammals (5); plants (13) (supplemental material 2). Point occurrence data was provided by the ŠNP administration and by the Nature Conservation Agency of the Czech Republic (AOPK ČR) via Czech the National Database of Fauna-Flora Records and Nature Protection [NDOP] (AOPK ČR, 2022a). We selected species with at least ten records from 2000 to 2021 and excluded species exclusively related to water ecosystems due to a lack of data describing the quality of water ecosystems and their relatively small portion within the area of ŠNP. Pre-processing analyses were done in ArcGIS 10.8 (ESRI, 2020) and QGIS (QGIS Development Team, 2021).

MaxEnt 3.4.1 software enabling habitat modelling with presence-only data were used (Elith et al., 2011). The principle of the MaxEnt approach is based on maximum entropy estimating a target probability of species´ distribution, which is derived from the above-mentioned occurrence data and environmental variables (Phillips et al., 2006). We run a model with different resolutions of environmental variables to set it appropriately (Sillero & Barbosa, 2021). By comparing models and their AUC (area under the curve showing the quality of the model) of various resolutions (500, 250, 100, 50 and 10 m) we finally selected outputs in the most detailed 10 × 10 m resolution. The models had the best AUC and the finest resolution, which can help management and zonation the most. Therefore, environmental predictors describing biotic and abiotic characteristics of ŠNP were converted into a raster with 10 × 10 m resolution (see supplemental material 1).

We run first the MaxEnt model with all species and environmental variables to analyse the response of variables within the model and set significant variables influencing certain species. In the second round, based on expert knowledge and statistical contribution, we only selected best-fitting variables for certain species (Zurell et al., 2020). Environmental variables were tested for autocorrelation; some correlated (>0.65) variables were excluded from the model run (Dormann et al., 2013). We used 10,000 background points, 500 iterations, and 10% of the data for testing. Then, a habitat suitability model (HSM) for each species with suitability from 0 to 1 (highest) was made. The output format was cloglog because it gives an estimation of the probability of presence between 0 and 1, which is easy to interpret within the prioritization task (Phillips, 2017). In four cases of presence–absence plant data, GLM models in R (R Core Team, 2018) were used and subsequently normalized to a 0–1 suitability scale. See variables selected for the final model run in supplemental material 2.

We acknowledged that despite the selection of several umbrella species (e.g. Boitani, 2000 – wolf (Canis lupus); Carroll et al., 2010 – selected owls; Sirkiä et al., 2012 – capercaillie (Tetrao urogallus)), our pool of species cannot contain a wide diversity of all organisms. Therefore, we also deployed a habitat mapping layer – a unique and detailed description of natural habitats also with the updated version of the layer; the experts mapped the whole country in 2001–2004 to determine Natura 2000 sites (AOPK, 2022b; Chytrý et al., 2010; Härtel et al., 2009). This original layer was updated and mapping was repeated in 2006–2016 (Lustyk & Oušková, 2011). Many attributes for habitats were recorded; therefore, we could compute the quality of habitat for both mappings according to Lustyk and Oušková (2011). For this study, we prepared three derived layers: (1) change of habitat quality between the original and actual layer; (2) quality of habitats in the actual layer; and (3) quality of habitats in the original layer. By using these layers, the quality of habitats and their (dis)continuity and (in)stability is present in our study. Here, we obtained layers with rankings emphasizing habitat quality and its improvement. By using actual and detailed species and habitat data as well, we solve the distinction between species- and ecosystem-based approaches of prioritization and cover both relevant biodiversity features (Ceauşu et al., 2015; Sharafi et al., 2012).

To analyse prioritization within ŠNP, we utilized freely available ZONATION software, a spatial conservation planning tool for hierarchical prioritization (Moilanen et al., 2005; Moilanen & Kujala, 2006). ZONATION works with biodiversity features as inputs (Thomson et al., 2009). The algorithm for conservation priorities starts from the whole landscape and then iteratively removes cells (pixels) of the raster of the lowest conservation value; thus, the most important cells for environmental variables remain. Furthermore, ZONATION can employ further considerations, such as works with compactness and structural connectivity of the landscape, additionally, edge removal provides computation efficiency. As a result, hierarchical prioritization shows a relative proportion of the most valuable ( = priority) area (Moilanen, 2007). ZONATION can weigh input variables (Moilanen et al., 2011) as well.

We computed a ZONATION algorithm for each functional group separately by weighting each species with the same value ( = 1). We also did the same with the above-mentioned habitat mapping layers. Then, these prioritization layers (amphibians, reptiles; birds; fungi; insects; lycopodium, fens, mosses; mammals; plants; habitat mapping layer) were used as inputs for further prioritization weighted 1 for particular functional species groups and 7 for habitat mapping layer prioritization to balance species- and ecosystem-based approaches (Ceauşu et al., 2015).

We used the additive benefit functions (ABF) algorithm in ZONATION (Moilanen, 2007). It emphasizes locations with many occurrences of different species, richness of the inputs (species groups and habitats) rather than rarity, and maximum of each input like the core-area zonation (CAZ) algorithm does (Di Minin & Moilanen, 2012; Lehtomäki & Moilanen, 2013; Moilanen, 2013; Moilanen et al., 2011). In our case of a relatively small, well-mapped area to show a detailed view, and with the importance of habitat resolution, we did not use removal from edge and connectivity measures to produce a detailed site-specific prioritization map.

3. Results

The final prioritization of nature protection was visualized in the map. We can see a pattern of protection with the areas of high protection rankings, e.g.: Mrtvý luh; Plechý – Třístoličník area; Chalupské slatě mires; Jezerní slať mire within Kvilda´s and Horská Kvilda´s surroundings; Modravská slať mires; and Křemelná surroundings around Nová Hůrka. Basically, locations at the highest altitude, along watercourses and mires also commonly correspond with the highest level of the current zonation protection; however, national park zonation creates larger patches than our prioritization with detailed resolution.

An intersection of our prioritization model (divided by the relative distribution of zones from old and current zonation) with current management zonation shows a match on 51.9% of the area. Older zonation only has 13% of the ŠNP area as the first zone (with the highest protection) and that corresponds with the highest 13% of our prioritization only on 5.7% of ŠNP. In contrast, in the case of the current zonation of ŠNP, there is a 16.1% match between the ZONATION model and the old zonation. Therefore, the current zonation protection pattern as a whole and the natural zone delimitation are closer to the ZONATION model than the old zonation (Table 1).

Table 1. Intersection or our prioritization and current zonation of Šumava National Park.

ZONATION prioritization model	Zonation of ŠNP (old/current zone)	Area of our model ZONATION and current zonation intersect (% of ŠNP)*	Area of our model ZONATION and old zonation intersect (% of ŠNP)
1 (72–100%)	I/natural	16.09	9.32
1 (72–100%)	II/close to nature	4.33	18.16
1 (72–100%)	III/concentrated management zone	8.62	1.57
2 (48–71%)	I/natural	6.82	1.70
2 (48–71%)	II/close to nature	6.99	20.20
2 (48–71%)	III/concentrated management zone	10.22	1.13
3 (1–47%)	I/natural	4.82	2.06
3 (1–47%)	II/close to nature	13.28	43.77
3 (1–47%)	III/concentrated management zone	28.81	2.08

* A cultural landscape zone was added to the concentrated management zone in this analysis.

The current zonation of ŠNP is more similar to our model even in the most valuable locations, but the match regarding the comparison of classification of ZONATION and current management zonation is still only slightly more than half of the ŠNP area (Table 1). Based on this, we suggest that the zonation of ŠNP is developed to better protect really desirable areas; on the other hand, there is a big gap between reality and the model. This stems from different roots of current management zonation and how prioritization of the ZONATION algorithm works. The current zonation of ŠNP has been driven by people, based rather on vegetation maps, forestry inventories and political and municipal interests; on the other hand, the inputs we used in our ZONATION model were habitat suitability models of important species, which are the subject of protection (Křenová & Hruška, 2012).

Therefore, we are aware that a full comparison of our prioritization and management zonation is not possible. However, it helps to reveal a new perspective on a way to delimit zonation and valuable locations by using objective data and modelling and compare it with current zonation.

4. Discussion and conclusion

We tried to make a model of protection prioritization based on all dimensions of biodiversity – species and ecosystem-driven; based on habitat quality and habitat suitability modelling using occurrence records and environmental predictors; however, we admit that it is limited by selected species, predictors, or algorithm set-up (Ceauşu et al., 2015). Moreover, land use some activities might not be included sufficiently because it is hard to pronounce their spatial impact precisely; examples include the intensity of tourism and other human activities with rather negative impact on nature conservation (Křenová & Kindlmann, 2015; Verburg et al., 2009). Furthermore, disturbance can change completely ŠNP and appropriate management is desired (Turner, 2005).

ŠNP has experienced turbulent changes in management; therefore, there were attempts to improve zonation. Bláha et al. (2013) proposed delimitation of the first zone based on valuable habitats and core areas of capercaillie (Tetrao urogallus) occurrence. The priority locations mentioned in the results vastly correspond with those in the article (Bláha et al., 2013). However, the methods and character of results are different; our approach is based on the modelling of the habitat suitability of more key species and use of the more precise data for modelling and also an updated habitat mapping layer. Furthermore, ZONATION software was used, which is different from simple overlaying data as in the mentioned article (Bláha et al., 2013). ZONATION generated a map for priority on the whole area of ŠNP, whereas Bláha et al. (2013) gave a proposal only for Zone I.

The new zonation of ŠNP is closer to our prioritization than the older one. However, a map shows the intersection of our model and current zonation. It is helpful for delimitation locations, where prioritization and zonation are similar and also to find ‘gaps’, locations, where zonation is significantly different from our results. It can be helpful for future management of the area. Again, the mentioned valuable locations are more or less the same from the point of view of zonation and prioritization as well. On the other hand, some locations, e.g. non-forested areas are valuable from a priority point of view, but they are in concentrated management zones, which deserves conservation concern.

Forested habitats in particular are protected, which cover a vast area of ŠNP (Bláha et al., 2013; Křenová & Hruška, 2012). However, our results often show priority locations on dynamically changing or open habitats of the ŠNP, such as disturbed forests or mires (Janík & Romportl, 2018) where habitat heterogeneity supports species diversity (Bengstsson et al., 2003). Despite the prioritization of certain locations within ŠNP, to preserve the ecological functions and resilience of the protected areas, it is crucial to protect as large an area as possible and ŠNP as a whole (Bengstsson et al., 2003). Our prioritization results are based only on the ŠNP area; therefore, lower priority within ŠNP does not mean valueless location in a broader geographical context (Lehtomäki & Moilanen, 2013). For example, it is a case of large carnivores and generalists, such as wolf (Canis lupus) and lynx (Lynx lynx), for whom landscape integrity and a low level of habitat fragmentation are crucial. Prioritization on this scale is negligible for them because they need a larger suitable area and, in this case, they used practically the whole area of ŠNP (Turner, 2005).

Spatial context matters in the case of ŠNP at least three-fold. (1) We did not use connectivity measures because our aim was oriented more on finding really suitable and priority locations within a relatively unfragmented area; however, in the case of connectivity between protected areas, we would use them (see Moilanen et al., 2005; Moilanen & Wintle, 2006). (2) Moreover, ŠNP creates a transboundary park with BFNP, it raises its connectivity and the relevance of its protection not only for national but also for continental or global priorities of nature conservation. We elaborated priorities in ŠNP first, because BFNP has solved the question of zonation – it is different from ŠNP (Janík, 2020; Moilanen & Arponen, 2011; Moilanen et al., 2013; Montesino Pouzols et al., 2014). (3) There is a dynamic development of consumer-producer and predator-prey relationships; their overlapping and spatial interactions could be continuously monitored to capture the current distribution of key species, such as the above-mentioned generalists (Rayfield et al., 2009).

From a methodological point of view, we used the ABF algorithm because it favours richness to retain high-quality occurrences of all features and creates a rather continuous area, not fragmented hot spots for each species or group (Di Minin & Moilanen, 2012; Moilanen, 2007).

To sum up, we used prioritization by using objective data and software for evaluating protection within ŠNP with its turbulent development management. Newly applied management zonation corresponds better to our prioritization than the former one. Core locations deserving protection are rather similar; however, there is still a huge difference between current zonation and our prioritization. The results are a contribution to strategic planning and sustainable development of protected areas (Carroll et al., 2010; Gordon et al., 2009) and can serve as a supporting material for bridging the gaps in management of ŠNP. Furthermore, prioritization within a broader spatial context of a national or regional scale could be the next step for the investigation of spatial nature protection prioritization.

Software

We used ArcMap 10.8 and QGIS for pre-processing and post-processing data during the analysis (changes in resolution, formats, and tools for normalization). MaxEnt 3.4.1 was used for the preparation of habitat suitability models and RStudio for GLM. ZONATION was used for prioritization. The map was created in ArcMap 10.8 and ArcGIS Pro software.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data which are depicted as a result of this study are openly available online: https://doi.org/10.5061/dryad.pk0p2ngqz.

Additional information

Funding

This work was supported by the Faculty of Science, Charles University [grant number SVV 244-2605731].