Mapping Connectivity in the Swedish Agricultural Landscape

Master’s thesis

Two years

Ecotechnology and Sustainable Development Environmental Science

Mapping Connectivity in the Swedish Agricultural Landscape

William Franzén

MID SWEDEN UNIVERSITY

Ecotechnology and Sustainable Building Engineering

Examiner: Anders Jonsson, anders.jonsson@miun.se Supervisor: Oskar Englund, oskar.englund@miun.se Author: William Franzén, wifr1400@student.miun.se

Degree programme: International Master’s Programme in Ecotechnology and Sustainable Development, 120 credits

Main field of study: Environmental Science Semester, year: Spring, 2020

A BSTRACT

The changes that Swedish agriculture has undergone during the 20^th century has resulted in strongly increased productivity, but at the cost of more intensive

environmental impacts. One of these is loss of biodiversity, which is driven by, e.g., usage of pesticides and loss and fragmentation of habitats. A vital process for

resilient ecosystems is the possibility for species to move between habitats, known as connectivity. One approach to increase connectivity is through strategic

perennialization in the agricultural landscape. The aim of this thesis is to map

structural connectivity in agricultural landscapes in two major agricultural regions in Sweden and explore options for enhancing connectivity by strategic perennialization.

Objectives include the development of a model to map structural connectivity in the Swedish agricultural landscape, identify landscapes where conditions for

biodiversity can be improved by strengthening the structural connectivity, and investigate the potential to improve the conditions for biodiversity by introducing perennial crops in the agricultural landscape. The resulting model is based on circuit theory using the software Circuitscape, in which land cover is treated as electric circuits, which are assigned resistance based on the permeability of different types of land cover. The resistance in the developed model is based partly on human impact and partly on structural differences from areas of high biological values, or value cores, between which connectivity is modelled, in terms of object height- and cover.

Two agricultural production areas were investigated, Skåne plains and Västra Götaland plains, as well as a testing area in Skåne county. Connectivity maps were created and analysed, and potential areas for strategic perennialization were

identified. A strategic perennialization scenario was also modelled in the testing area.

Since the application of the model is structural connectivity, uncertainties regarding how well it relates to functional connectivity varies between species. Structural connectivity has nonetheless been shown to facilitate functional connectivity in

several aspects. No significant difference in connectivity could be found in the testing area following the introduction of strategic perennialization, but this is most likely due to assumptions behind area selection. Therefore, other approaches for

identifying promising locations for strategic perennialization, based on connectivity maps, need to be explored.

Keywords: GIS; Circuitscape; Structural Connectivity; Agriculture; Landscape;

Perennialization; Land use

INDEX

Abstract ... 1

1 Introduction ... 1

1.1 Aim and objectives ... 4

2 Background ... 4

2.1 Connectivity ... 4

2.1.1 Structural & multi-species connectivity models ... 5

2.2 Circuit theory in connectivity studies ... 6

2.2.1 Circuitscape ... 6

3 Method ... 8

3.1 Studied areas ... 9

3.2 Circuitscape input data ... 10

3.2.1 Resistance Surfaces ... 10

3.2.2 Focal points ... 13

3.2.3 Adjustments of input data ... 14

3.2.4 Circuitscape settings ... 14

3.3 Creating and analysing connectivity maps ... 14

3.3.1 Creating total connectivity maps ... 15

3.3.2 Identification of potential areas for improved connectivity ... 15

4 Result ... 16

4.1 Skåne plains ... 16

4.1.1 Agricultural connectivity and potential areas for perennialization... 18

4.2 Västra Götaland Plains... 19

4.2.1 Agricultural connectivity and potential areas for perennialization... 23

4.3 Testing area ... 24

4.3.1 General differences between different resolutions ... 26

4.3.2 Deciduous value cores ... 26

4.3.3 Coniferous value cores ... 27

4.3.4 Agricultural connectivity and potential areas for perennialization... 28

4.3.5 Modification based on strategic perennialization ... 29

5 Discussion ... 31

5.1 Discussion of methodology ... 31

5.1.1 Uncertainties in the model ... 31

5.1.2 Impacts from different resolutions ... 33

5.1.3 Modelling strategic perennialization ... 34

5.2 Result discussion... 35

5.2.1 Prioritization of value cores ... 35

5.2.2 Differences in connectivity results ... 36

5.2.3 Different strengths & weaknesses ... 36

5.3 Conclusions ... 37

5.3.1 Further studies ... 38

6 References ... 39

1

1 I NTRODUCTION

Sweden has a total cropland area of ca 2.5 million ha, of which 44 % is ley, followed by 38 % cereals. A further 17 % is used as grazing lands. The south and southwest parts of the country have the greatest share of croplands as well as grazing land, as can be seen in Fig. 1, with some variations between the different crops and types of agriculture (SCB, 2019).

The Swedish agricultural landscape has gone through significant changes during the 20^th century. During this period, it changed from small-scale agriculture in which croplands, forestry and animal husbandry was integrated with each other, to the rationalised agriculture of today with bigger fields, less natural areas, less variations of crops and increased usage of pesticides and fertilizers (Sandström, et al., 2015).

Figure 1. The geographical distribution of agriculture in Sweden, presented as % of total land area of counties.

This development has come with the advantage of a strongly increased productivity, but at the cost of environmental impacts (Sandström, et al., 2015), e.g.,

eutrophication, erosion, loss of soil carbon and loss of biodiversity. Eutrophication is caused by nutrient leaching from soil to bodies of water. Agriculture is the largest contributor to the problem in southern Sweden (Havs och vatten myndigheten,

2 2019). Erosion can be caused by wind as well as water and is one of the leading

factors for soil loss in Europe, with agricultural land use being one of the primary causes (Panagos, et al., 2015). Soil Organic Carbon (SOC) is a measure for soil quality.

Soil also act as carbon sink and has potential to be used for carbon sequestration in agricultural lands if current practices are changed toward such that increase SOC (Lal, 2004; Lugato, et al., 2014). Under current practices however, agriculture is a major driver for SOC-loss (Power, 2010).

Biodiversity loss in the agricultural landscape is driven by, e.g., usage of pesticides and loss of- and fragmentation of habitats. Agricultural fields can in themselves also make out barriers for species belonging in other landscapes (Berglund, et al., 2018).

Biodiversity is crucial both for the agricultural landscape as well as other

environments since it is a prerequisite to effectively deliver ecosystem services.

Biodiversity is the variability of living organisms, including the diversity between and within species and of ecosystems (United Nations, 1992). A higher biodiversity may contribute to the productivity of ecosystem services. Production of biomass, e.g., has been found to be significantly higher in areas with a higher diversity compared to others (Cadinale, et al., 2012). Biodiversity also contributes to the stability and resilience of ecosystems. In a system where several species serve the same function, known as functional diversity (Chillo, et al., 2011), this diversity is not likely vital for it to function. However, in the event of some disturbance or a change that threatens certain species in the system, a higher diversity increases the stability by making it possible to compensate for the loss of species. Diversity within a certain species, or genetic diversity, also makes it less sensitive to different stresses, e.g., different crops have been found to have a higher resistance to pest and diseases (Science for

Environment Policy, 2015).

A vital process for species is the possibility to spread, which affects the density and distribution of populations in the landscape, promotes colonization of places and rescues small populations from local extinction through migration. A measure for the extent to which a certain species can spread between habitats in a landscape is

connectivity. Connectivity is a function of the quality and quantity of habitats, meaning that without enough habitats of high enough quality, connectivity does not exist (Berglund, et al., 2018). Connectivity can be either functional or structural. For functional connectivity, the structure of the landscape is weighed together with how a certain organism functions in order to determine its spreading possibilities, while

3 for structural connectivity only the physical structure of a landscape in terms of distance or pathways between habitats are considered (Kindlmann & Burel, 2008).

Sweden has 16 national environmental goals, several of whom overlap in the aim of preserving and enhancing biodiversity. One that has a specific focus on biodiversity in the agricultural landscape is “A varied Agricultural Landscape”, which relate to the maintenance, protection and development of ecosystem services and biodiversity in it. The state of agricultural ecosystem services is still considered to be at an

acceptable level, but they are at risk to decline or vanish with a further decrease of biodiversity (Jordbruksverket, 2018). In Sweden, 33 % of red-listed species has the agricultural landscape as an important biotope, with a further 22 % using it

occasionally. Some of the most vulnerable species include beetles, butterflies and vascular plants. The most important biotopes for the red-listed species include different kinds of grasslands, which are heavily on the decline following the rationalisation of agriculture during the 20^th century in Sweden (Sandström, et al., 2015). Grasslands have today stabilized at around 450 000 hectares. For croplands, however, the trend largely continues with a decline at around 10 000 hectares per year, with the loss of small biotopes as a result (Karlsson & Wallander, 2019).

If the environmental goal of a varied and rich agricultural landscape is to be achieved and ecosystem services to be maintained in the long run, action must be taken in order to stop the decline of biodiversity. Some key points that have been pointed out to achieve this, e.g., the maintenance and increase of small biotopes that differ from the surrounding landscape that make out habitats for species who do otherwise not belong in the current landscape (Karlsson & Wallander, 2019).

One possible approach for this is the strategic introduction of perennials in the agricultural landscape. This has the potential to mitigate environmental impacts in the agricultural landscape while providing economic incentives by potentially increasing production of food as well as biomass for e.g. material and bioenergy.

Trees can for example be planted to act as windbreaks, slowing and spreading water runoff, and for taking up nutrients, thus decreasing wind- and water erosion as well as eutrophication (Christen & Dalgaard, 2013). As opposed to annual crops,

perennials increase soil quality by increasing SOC (Asbjornsen, et al., 2013). For biodiversity, perennialization promotes habitats by increasing resource heterogeneity and landscape complexity. This in turn may enhance ecosystem resilience and

services such as biotic regulation trough pollination and pest control (Asbjornsen, et al., 2013). It can also increase connectivity between different biotopes in the

4 landscape. It have been found that rows of trees, e.g., can function as ecological corridors for certain woodland beetles for which agricultural fields may be obstacles (Oleksa, et al., 2015), or stripes of pastures and ley that can harbour a rich variety of species (Carlsson, et al., 2015).

To evaluate the potential for improving the conditions for biodiversity trough these kinds of systems, it must be investigated in what landscapes improvements of the conditions are needed and likely to succeed.

1.1 A

The aim of this thesis is to map structural connectivity in agricultural landscapes in two major agricultural regions in Sweden and explore options for enhancing

connectivity by strategic perennialization. Objectives include:

1. Develop a model to map structural connectivity in the Swedish agricultural landscape

2. Identify landscapes where prerequisites for biodiversity can be improved by strengthening the structural connectivity

3. Based on the results from the model, investigate the potential to improve the prerequisites for biodiversity by introducing perennial crops in the

agricultural landscape

2 B ACKGROUND

2.1 C

Connectivity can be divided into either structural or functional connectivity. The structural connectivity is a measure based only on the structure of the landscape, whereas functional connectivity includes a specific species reaction to its

environment. These can however be further subdivided into additional categories, as described by Kindlmann & Burel (2008) in a review on connectivity measures.

Structural definitions include approaches such as corridors, distances, amount of habitat in a landscape and contagion or percolation. Corridors are strips of habitat that connect otherwise disconnected habitat patches and relies on the assumption that species avoid leaving their habitats. Those based on distance is founded in Euclidian distances, where connectivity is measured trough, e.g. shortest distance between habitat patches. Measures looking at the amount of habitat uses a buffer around

5 habitats to determine connectivity based on if surrounding patches fall inside of this buffer. Contagion or percolation approaches see habitats as connected if they in a two- dimensional grid joins with another habitat in at least one edge. Species are here assumed to be able to move between such patches (Kindlmann & Burel, 2008).

Some functional definition includes time spent searching for new habitat patches and immigration rates. Time spent searching for new habitat is based on the average

movement steps needed for an individual to reach a habitat patch. Immigration rate is based on the total immigration to a patch divided by the initial population. In this approach, a lower rate equals a more isolated patch and vice versa (Kindlmann &

Burel, 2008).

2.1.1 Structural & multi-species connectivity models

Since structural connectivity does not take into account any species-specific requirement, it could be questioned how well they correspond to functional

connectivity. This question has been raised in a number of studies, some of which are summarized here.

In a review of articles on connectivity, Fletcher et.al (2016) found that structural connectivity tended to have a positive impact especially on the distribution and diversity of species. Similar conclusions were made in a literature review on whether structural connectivity facilitate functional connectivity for species native to

Australia. It was found that corridors made a significant impact on functional connectivity by constituting habitats, as well as to some degree facilitate movement for species. Steppingstones were also found to be as effective, or even more, for facilitating movement (Doerr & Davies, 2010). A study investigating functional connectivity for fishing martens between protected areas found that corridors of structural likeness were the most important factor when individuals moved between areas (Stewart, et al., 2019). A similar conclusion was drawn in a study investigating the importance of structural connectivity for amphibian vertebrates, stating that the structural connectivity affects the richness of species pattern and that it is an

important factor for the presence of some species (Ribeiro, et al., 2011). Brodie, et al.

(2014) found that when combining the functional connectivity between a varied taxa of species to find important structural corridors in a landscape, their model yielded more accurate results when dividing them based on species classes, such as

carnivores or herbivores. Baguette & Dyck (2007) points out that structural

connectivity yields less precise result than functional and should be compared with empirical reference data sets over functional connectivity.

6

2.2 C

One possible way to model and measure structural, as well as functional,

connectivity is through the usage of circuit theory. Circuit theory is applicable to connectivity through its connection to “random walk theory”, which electrical currents correlate closely with (Mcrae, et al., 2008). A random walker in this context is an individual with no prior knowledge of an area, that base its movement on its immediate surroundings (Dickson, et al., 2018).

A random walker is here represented by an electrical current travelling between nodes. The surfaces on which it travels upon consist of resistors, where a higher resistance means a higher opposition to the flow of the current. The connectivity can be measured through resistance distance, which is a unit that describes the isolation between the nodes the current travels between. One of the strengths of this measure is that it incorporates multiple paths between nodes, as it reflects the minimum distance as well as cost and availability of different routes (Mcrae, et al., 2008). A measurement that can further be used to identify important corridors for random walkers is current density. A high current density indicates an important route, seeing as a current shows the flow of a charge in a circuit (Mcrae, et al., 2008).

2.2.1 Circuitscape

Circuitscape is a software in which connectivity can be modelled trough circuit theory. The input data to Circuitscape consist of a resistance surface and focal nodes or points. The resistance surface can be a raster map in which each pixel is assigned a certain resistance based on, e.g., land cover permeability or anthropogenic impact to evaluate how an electrical current – or random walker – moves in a landscape. Focal points are the locations between which connectivity is to be measured and can for example consist of a raster map that marks the habitat for a certain species, or an area with high biological values. These focal points are the places to which currents and grounds are connected. It simulates connectivity by connecting a 1-amp current between the focal nodes to create a map showing the cumulative current, the

aggregated current that flows along circuits after connecting each node to a current source, which represents the possible routes between the focal points. A higher current can be interpreted as a higher likelihood of that route being used by a random walker (Mcrae, et al., 2008).

7 2.2.1.1 Creating resistance surfaces

In searching the literature, no established method has been found for developing resistance surfaces for investigating structural connectivity. Those studies identified that seeks out to investigate structural connectivity have instead been based on multi-species functional connectivity, such as that of Compton, et al. (2007), who developed what they call a “resistance-kernel” model to investigate connectivity between breeding grounds for amphibians.

This method is based on kernel estimators, which uses two-dimensional data to create three-dimensional surfaces that represent an estimated underlying probability distribution from, e.g., a dispersal source of a studied species, thus creating a

resistance surface. The resistance weighting was based on expert opinion in lack of sufficient empirical data (Compton, et al., 2007). The resistance kernel model was used by Andersson, et al. (2012; 2014) to investigate the structural connectivity in different North American regions. In these studies, more generalized weighting schemes were used based on the assumption that requirements for general movement are less sensitive than that between breeding grounds.

Another method was developed by Koen, et al. (2014), also aiming to investigate multi-species connectivity. Resistance was assigned based on assumptions of the impairment of different land covers to movement. A resistance of 1000 ohms was assigned to covers deemed highly impermeable, 100 ohms to unnatural but

permeable surfaces, and 10 ohms to areas deemed natural or highly permeable. The resulting maps were then validated with data over movement of herptiles as well as fishing martens, which led to the conclusion that the model was better suited to estimate the connectivity of herptiles. This model was later used by Bowman &

Cordes (2015) to evaluate differences in wildlife connectivity in the Great Lakes Basin and used the exact values and method from Koen et.al (2014).

Brodie, et al. (2014) created a multi-species resistance layer trough sampling of species abundance. Resistance was calculated by inverting the abundance together with Euclidian network analysis. Single-species resistances where standardized and calculated to create combined resistance surfaces for carnivores, herbivores and a total for all species.

Dickson, et al. (2016) created a resistance surface based on previously mapped

“Human Modification” in the western USA as described by Theoblad (2013),

together with slope percentage for all surfaces, with an extra penalty for steep slopes and a specific resistance for bigger rivers close to the models’ maximum resistance.

8 2.2.1.2 Calculation modes

Circuitscape has four modes of calculation; pairwise, advanced, one-to-all and all-to- one.

When using the pairwise mode, the current is connected to one pair of focal nodes at the time, one of which is ground, until all pairs have been calculated to create

cumulative current and voltage maps.

In the advanced mode, any number of current sources connected to any number of grounds, defined by the user, is activated at the same time. In the advanced mode, the user defines the current strength and can also give different sources different current strengths. Likewise, the grounds can have any user defined resistance.

The one-to-all mode resembles the pairwise mode, but instead of connecting one pair at a time, one node connects to all other nodes simultaneously. This is then repeated, using a 1-amp current source for each node, to create cumulative current and voltage maps.

The all-to-one mode is like the one-to-all, only in reverse. Here, all nodes connect to one ground node with a 1-amp current repeatedly for all focal nodes to create current- and voltage maps (Mcrae, et al., 2013).

3 M ETHOD

To investigate the structural connectivity in the Swedish agricultural landscape, the software Circuitscape (Mcrae, et al., 2020) was used. For this purpose, a resistance surface was created to measure the current on, as well as focal points to measure it between. Two production areas with some of the highest agricultural production in Sweden was used in the investigation.

Focal points were created from the locations of value cores, previously recorded areas with high biological values (Bovin, et al., 2017).

Resistance surfaces was created based on human impact as well as structural differences from value cores in terms of height and cover.

A detailed description of the method can be found below in section 3.1 – 3.3.

9

3.1 S

Two agricultural production areas, parts of Skåne plains and Västra Götaland plains, were selected for the investigation. They were selected rather arbitrarily, based on having among the overall highest agricultural production in Sweden (SCB, 2018).

Vector data for these areas were produced in a parallel ongoing research project and created using tabular data on production areas for blocks in the 2016 block database, provided by the Swedish Agricultural Agency. Due to significant computing power demands from the Circuitscape software, a further smaller area was used to evaluate the model as it was developed to avoid unnecessarily long processing times to yield a result from the larger production areas. The testing area was also used to investigate the potential impact from strategic

perennialization, something that could not be done for the production areas within the timespan of this thesis. The areas are shown in Fig. 2.

Since currents are measured between focal points there is a risk that the result becomes misleading near the borders of investigated areas, since there may be focal points in close proximity outside of borders, between which a strong current could otherwise exist. To eliminate this potential risk, a 20 % buffer was added around the investigated areas (Koen, et al., 2012).

To be able to efficiently work with all the different data that went into the model, all layers for all types of GIS input data were clipped, using the production area vectors with added buffers. Buffers was added using the QGIS plugin “Buffer by

percentage” (Dugge, 2018).

Figure 2. The investigated production areas. Västra Götaland plains in red, Skåne plains in orange and the testing area in blue.

Skåne plains consists of an additional smaller area in northeastern Skåne. That area was however not included.

10

3.2 C

3.2.1 Resistance Surfaces

Resistance surfaces were created for each area and forest types in two steps. First, resistance attributed to the human impact was estimated, to which resistance attributed to structural difference afterwards was added

A detailed raster land cover data set over Sweden divided into 25 different land cover classes with a 10x10 m resolution was used as a foundation to build the

resistance layers from (Ahlkrona, et al., 2019), to which additional layers were added to create a more complex surface. Forest value cores, which consist of formally

protected or unprotected areas with high biological values (Bovin, et al., 2017), was rasterized using the QGIS algorithm “Rasterize (Vector to Raster)” (QGIS

Development Team, 2020) and assigned a value of 500 to be easily identified. A dataset over pastures and ley that were produced in a parallel ongoing research project and created using tabular data on production areas for blocks in the 2016 block database provided by the Swedish Agricultural Agency, was assigned the values 2000 (pastures), 3000 (pastures) and 4000 (ley) using the GRASS GIS tool

“r.reclass” (GRASS Development Team, 2019). The values were selected in order to make it possible to easily distinguish them later in the process. Pastures were divided into two different categories in the original dataset. This was maintained in the

model in case a need to distinguishes between them would arise, which it however did not.

The forest value cores, pastures and ley were added to the foundational layer using the GRASS GIS tool “r.series” with the aggregate operation “sum”, which as the name suggests sums the value of all overlapping layers. All layers overlapped with each other, leading to the pastures, ley and value cores receiving a value

corresponding to the sum of their own value plus the foundational layers value, which ranges from 2 – 128. For example, ley got values ranging from 4000 – 4128.

All above-mentioned values do not relate to the resistance but are merely used for identification so that it can be certain which values corresponds to a certain land cover when actually assigning resistance values.

3.2.1.1 Resistance from human impact

The resulting map with added pastures, ley and value cores were used to create a resistance surface with resistance from human impact in accordance with Table 1 using the GRASS GIS tool “r.reclass” (GRASS Development Team, 2019). These

11 values were largely decided upon by using Landscape Development Intensity (LDI) coefficients, a measure based on emergy that takes in to account all the non-

renewable energies that are consumed for a certain land use per ha and year, as described by Brown & Vivas (2005), scaled up to the range 1 – 100 unless otherwise indicated in Table 1. The exception is water, which was assigned the highest total resistance after structural resistance was added to reflect the fact that water often make out a significant barrier for many species (Koen, et al., 2014; Dickson, et al., 2016).

Table 1. The resistance attributed to human impact for different land covers and the values they are based on. As the

Circuitscape software can only take values from 1 and over, a value of 1 had to be added to all LDI coefficients when scaling it up from 0 – 10 to 1 – 100.

Land cover type Resistance attributed to human impact

LDI coefficients from Brown & Vivas (2005)

Value cores 1 0.00, natural system

Deciduous forest 1 0.00, natural system

Mixed deciduous &

coniferous forest

17 1.58, Pine plantation

Coniferous forest 17 1.58, Pine plantation

Wetlands 1 0.00, natural system

Pastures 21 2.02, Woodland pasture

(with livestock)

Ley 29 2.77, Improved pasture

(without livestock)

Arable land 46 4.54, Row crops

Vegetated or non-vegetated other land

19 1.83, Recreational / open

space – low-intensity Artificial surfaces, with or

without buildings, without roads

100 10.00, Central business

district (average 4 stories)

Roads 84 8.28, Highway (4 lane)

Inland & marine waters Highest final resistance after added structural differences

The decision to treat all areas dominated by deciduous forest types as natural, while all mixed or coniferous dominated areas as pine plantations, was based on the fact that mainly coniferous forests are used for silviculture in Sweden. On a national

12 level, deciduous forests make out 6,4 % of the productive woodlands, or 22,3 % in Skåne county and 9,2 % in Västra Götaland county (SLU, 2016).

Roads and artificial surfaces were set to the highest LDI coefficient for their

respective category so as to not underestimate them since the different types of these surfaces are not accounted for in the land cover dataset.

3.2.1.2 Resistance due to structural difference

Further resistance for structural differences was created by adding resistance based on how much each pixel deviated from the average height and cover (Ahlkrona &

Jönsson, 2019) within the value cores. One extra ohm of resistance was added for each meter or cover percentage they differed, with exception of value cores which got no extra resistance. For this purpose, the height and cover data – which is divided into separate data sets for object heights between 0,5 – 5 m and 5 – 45 m – were added together to create layers containing the total object cover- and height, using the GRASS GIS tool “r.series” (GRASS Development Team, 2019) with the aggregate operation “sum”. These layers furthermore had their null values replaced by “0”, since areas such as agricultural fields or waters do not have any height or cover in the original layers. To calculate the average values inside value cores, these data were extracted using the QGIS tool “Clip raster by mask layer” (GRASS

Development Team, 2019), for which the value core vector layers were used as masks. The average height- and cover values were then identified using the QGIS algorithm “Raster layer statistics” and can be found in Table 2.

Table 2. The average height and cover of coniferous and deciduous value cores in the investigated areas.

Height, deciduous value cores (m)

Height, coniferous value cores (m)

Cover,

deciduous value cores (%)

Cover, coniferous value cores (%)

Skåne Plains 20 17 59 64

Västra Götaland Plains

21 21 74 77

Testing area 21 21 55 55

Depending on what forest type that was being investigated, a final adjustment was made in which the investigated forest type received 50 % of the calculated resistance, mixed forests 75 % and the opposite forest type 100 % (e.g., if connectivity between deciduous value cores was being investigated, deciduous forests received 50 % of the

13 resistance, mixed forests 75 % and coniferous forests 100 %). This was done to reflect that different species often are adapted for different types of vegetation (Berglund, et al., 2018). The underlying assumption in this model is that differentiating between different forest types simulates these differences.

The structural difference was added to the resistance attributed to human impact using logical functions in the GRASS GIS tool “r.mapcalc”, mainly if (a, b, c,), which means “if a, do b, otherwise c”. An example can be seen in Equation 1. In the

equation, the resistance surface for connectivity between deciduous value cores is calculated:

𝑖𝑛𝑡( 𝑖𝑓( 𝑑𝑒𝑐𝑖𝑑𝑢𝑜𝑢𝑠 𝑓𝑜𝑟𝑒𝑠𝑡𝑠, ( 𝑎𝑏𝑠(( ℎ𝑒𝑖𝑔ℎ𝑡 𝑜𝑢𝑡𝑠𝑖𝑑𝑒 𝑣𝑎𝑙𝑢𝑒 𝑐𝑜𝑟𝑒𝑠 – 𝑎𝑣𝑔. ℎ𝑒𝑖𝑔ℎ𝑡 𝑖𝑛 𝑣𝑎𝑙𝑢𝑒 𝑐𝑜𝑟𝑒𝑠) + ( 𝑐𝑜𝑣𝑒𝑟 𝑜𝑢𝑡𝑠𝑖𝑑𝑒 𝑣𝑎𝑙𝑢𝑒 𝑐𝑜𝑟𝑒𝑠 – 𝑎𝑣𝑔. 𝑐𝑜𝑣𝑒𝑟 𝑖𝑛 𝑣𝑎𝑙𝑢𝑒 𝑐𝑜𝑟𝑒𝑠 ) ) + 𝑏𝑎𝑠𝑒 𝑟𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒)

∗ 0.5,

𝑖𝑓( 𝑚𝑖𝑥𝑒𝑑 𝑓𝑜𝑟𝑒𝑠𝑡𝑠, (𝑎𝑏𝑠(( ℎ𝑒𝑖𝑔ℎ𝑡 𝑜𝑢𝑡𝑠𝑖𝑑𝑒 𝑣𝑎𝑙𝑢𝑒 𝑐𝑜𝑟𝑒𝑠 – 𝑎𝑣𝑔. ℎ𝑒𝑖𝑔ℎ𝑡 𝑖𝑛 𝑣𝑎𝑙𝑢𝑒 𝑐𝑜𝑟𝑒𝑠) + ( 𝑐𝑜𝑣𝑒𝑟 𝑜𝑢𝑡𝑠𝑖𝑑𝑒 𝑣𝑎𝑙𝑢𝑒 𝑐𝑜𝑟𝑒𝑠 – 𝑎𝑣𝑔. 𝑐𝑜𝑣𝑒𝑟 𝑖𝑛 𝑣𝑎𝑙𝑢𝑒 𝑐𝑜𝑟𝑒𝑠 ) )

+ 𝑏𝑎𝑠𝑒 𝑟𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒) ∗ 0.75,

𝑖𝑓( 𝑐𝑜𝑛𝑖𝑓𝑒𝑟𝑜𝑢𝑠 𝑓𝑜𝑟𝑒𝑠𝑡𝑠, 𝑎𝑏𝑠(( ℎ𝑒𝑖𝑔ℎ𝑡 𝑜𝑢𝑡𝑠𝑖𝑑𝑒 𝑣𝑎𝑙𝑢𝑒 𝑐𝑜𝑟𝑒𝑠 – 𝑎𝑣𝑔. ℎ𝑒𝑖𝑔ℎ𝑡 𝑖𝑛 𝑣𝑎𝑙𝑢𝑒 𝑐𝑜𝑟𝑒𝑠) + ( 𝑐𝑜𝑣𝑒𝑟 𝑜𝑢𝑡𝑠𝑖𝑑𝑒 𝑣𝑎𝑙𝑢𝑒 𝑐𝑜𝑟𝑒𝑠 – 𝑎𝑣𝑔. 𝑐𝑜𝑣𝑒𝑟 𝑖𝑛 𝑣𝑎𝑙𝑢𝑒 𝑐𝑜𝑟𝑒𝑠 ) ) + 𝑏𝑎𝑠𝑒 𝑟𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒, 𝑖𝑓( 𝑣𝑎𝑙𝑢𝑒 𝑐𝑜𝑟𝑒𝑠, 1,

𝑖𝑓( 𝑤𝑎𝑡𝑒𝑟, 𝑚𝑎𝑥𝑖𝑚𝑢𝑚 𝑓𝑖𝑛𝑎𝑙 𝑟𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒,

𝑎𝑏𝑠(( ℎ𝑒𝑖𝑔ℎ𝑡 𝑜𝑢𝑡𝑠𝑖𝑑𝑒 𝑣𝑎𝑙𝑢𝑒 𝑐𝑜𝑟𝑒𝑠 – 𝑎𝑣𝑔. ℎ𝑒𝑖𝑔ℎ𝑡 𝑖𝑛 𝑣𝑎𝑙𝑢𝑒 𝑐𝑜𝑟𝑒𝑠)

+ ( 𝑐𝑜𝑣𝑒𝑟 𝑜𝑢𝑡𝑠𝑖𝑑𝑒 𝑣𝑎𝑙𝑢𝑒 𝑐𝑜𝑟𝑒𝑠 – 𝑎𝑣𝑔. 𝑐𝑜𝑣𝑒𝑟 𝑖𝑛 𝑣𝑎𝑙𝑢𝑒 𝑐𝑜𝑟𝑒𝑠 ) ) + 𝑏𝑎𝑠𝑒 𝑟𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒))))))

(Equation 1) It should be noted that the calculation had to be run twice, once without water to receive a maximum final resistance, and once more to manually insert resistance for water. The average height and cover in value cores also had to be calculated

separately and inserted manually.

3.2.2 Focal points

Focal points were created using a vector layer from previously recorded value cores, which consist of formally protected or unprotected areas with high biological values (Bovin, et al., 2017). Value cores with an area of ≥0,4 ha were used for this purpose, a size that was decided upon based on the importance of relatively small habitats in the agricultural landscape (Brunet, et al., 2019). The value cores of the right size were extracted using the QGIS “field calculator” (QGIS Development Team, 2020). An already existing column listing the size in ha for each individual polygon was used for this purpose, which could be selected using the expression if (area => 0.39, 1, 0).

14 The selected polygons were then assigned centroids using the QGIS algorithm “point on surface” and assigned unique values using the algorithm “Add autoincremental field”. The Circuitscape software requires that each focal point has a unique value wherefore the “autoincremental” field was used as a burn-in value when rasterized.

3.2.3 Adjustments of input data

When the input data was finished, the resolution was converted to 20x20 meters using the GDAL warp-tool. Here the “average” resampling method was used in order to maintain a representative resistance in relation to the 10x10 m resolution.

This was done due to limitations in computing power. An analysis of the differing outcome from a lower versus higher resolution was carried out on the testing area (see section 4.3 – 4.3.3.2).

The Västra Götaland plains production area had to be split in two before running in the Circuitscape software, again due to limitations in computing power. This was made using the mask created for this area, which was divided in a straight line between the coordinates X: 388525.5106886456, Y:6439002.872520123 and

X:391860.1703163984, Y:6511031.520479583. To avoid any effects from focal points located close by where the map was split, a buffer of 85 % was added to both halves using the QGIS Plugin “Buffer by Percentage” (Dugge, 2018), set to 185 %.

3.2.4 Circuitscape settings

The calculation mode in Circuitscape was set to “all-to-one”, in which one focal point is connected to the ground whereas all other focal points connect to a 1-amp current source. This process is then repeated for each focal point to generate a cumulative current density map (Mcrae, Shah, & Mohapatra, 2013).

3.3 C

The buffer of all areas was removed after running them through the Circuitscape software by using the QGIS “Clip Raster by Mask Layer” algorithm, using the unbuffered production areas as a mask.

In the analysis of the cumulative current maps, the current density was limited to 98

% of the actual value. This delimitation was made since the highest 2 % made out a negligible part of the total current density, while including them also made the result harder to interpret and possibly misleading. The maps for the respective forest types are presented in quantiles with 20 classes and shows the relative connectivity for each type.

15 3.3.1 Creating total connectivity maps

To create a map where the total connectivity for all forest types in each area can be seen, the connectivity for each type was classified on a scale from 1 – 10 and added together with a weighting based on their number of focal points. The classification was made using Equation 2:

(𝑐𝑢𝑚𝑢𝑙𝑎𝑡𝑖𝑣𝑒 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 𝑚𝑎𝑝 / 𝑚𝑎𝑥𝑖𝑚𝑢𝑚 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 𝑑𝑒𝑛𝑠𝑖𝑡𝑦) ∗ 10

(Equation 2) The weighted combination of the maps for each area was made using Equation 3:

((𝑐𝑙𝑎𝑠𝑠𝑖𝑓𝑖𝑒𝑑 𝑑𝑒𝑐𝑖𝑑𝑢𝑜𝑢𝑠 𝑚𝑎𝑝 ∗ 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑑𝑒𝑐𝑖𝑑𝑢𝑜𝑢𝑠 𝑓𝑜𝑐𝑎𝑙 𝑝𝑜𝑖𝑛𝑡𝑠) + (𝑐𝑙𝑎𝑠𝑠𝑖𝑓𝑖𝑒𝑑 𝑐𝑜𝑛𝑖𝑓𝑒𝑟𝑜𝑢𝑠 𝑚𝑎𝑝

∗ 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑐𝑜𝑛𝑖𝑓𝑒𝑟𝑜𝑢𝑠 𝑓𝑜𝑐𝑎𝑙 𝑝𝑜𝑖𝑛𝑡𝑠)) / (𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑑𝑒𝑐𝑖𝑑𝑢𝑜𝑢𝑠 𝑓𝑜𝑐𝑎𝑙 𝑝𝑜𝑖𝑛𝑡𝑠 + 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑐𝑜𝑛𝑖𝑓𝑒𝑟𝑜𝑢𝑠 𝑓𝑜𝑐𝑎𝑙 𝑝𝑜𝑖𝑛𝑡𝑠)

(Equation 3) 3.3.2 Identification of potential areas for improved connectivity

A further analysis was made to identify where the structural connectivity potentially could be improved by introducing perennials, such as energy forest. This was done by identifying the agricultural fields that were among the areas with the highest 10 % of the total current density. In cases where currents of this strength did not exist, the highest 10 % of the current density in agricultural fields alone were used. These delimitations were chosen arbitrarily to provide a simple overview of where the agricultural fields with the strongest connectivity are located. The areas with the strongest connectivity was chosen for this purpose since they are assumed to support some level of connectivity which can be improved. Areas with lower connectivity are here thought to not support any significant connectivity and would thus not benefit from perennialization, since there is no connectivity to improve.

To investigate what impact energy forest may have for improving the structural connectivity in the agricultural landscape, the testing area resistance layer was modified within the identified fields in accordance with how an energy forest plantation may look. The investigated fields received the resistance from human modification of ley with a height of 7 m and a cover of 100 % to represent a fully grown energy forest of salix sp., one of the most commonly grown energy forest types in Sweden (Jordbruksverket, 2013). The Agricultural fields that contained any

significant presence of high current density were used for this purpose and were

16 arbitrarily chosen. The resistance for the modified fields becomes 88 in total for both deciduous and coniferous value cores.

4 R ESULT

In the section below the result from the application of the model are presented and analysed. Each area is presented based on forest type as well as weighted together.

4.1 S

The connectivity in the plains of Skåne is presented in Fig. 3 and statistically summarized in Table 3.

Figure 3. The modelled connectivity in the Skåne plains Production area, after value core forest type and in total. The legends in the current density maps specify the cumulative current densities presented as quantiles. A land cover map with focal points are also included.

17

Table 3. A statistical summary of the values resulting from the modelling of the Skåne plains.

Skåne Plains, coniferous Skåne Plains, deciduous

Max current density: 1082 19747

≤ 98 % current density: 35 654

Focal points: 214 1176

The connectivity here is the greatest around square C-16, F-9, H-7, K-4, M-5 and Q-3.

This is likely due to a higher concentration of focal points in these areas, which mainly consists of forest rich in value cores. Interestingly, an area with strong

connectivity can be found around the intersection of K-3 and L-4 as well, although no focal points exist in these areas, which are dominated by agricultural landscapes. The strongest connectivity in these places can be found in patches of trees and to some extent vegetated or non-vegetated other land.

Some of the strongest connectivity in these two areas can be found in the towns Staffanstorp and Skivarp, which can be explained by that they are dominated by residential buildings with high densities of trees and vegetated or non-vegetated other land, see Fig. 4. The situation is much like this in square F-3 as well, where several focal points and value cores also can be found. An area with weaker connectivity is around square Q-4.

The area is similar to the areas K-3 and L-4, but the weaker connectivity can likely be explained by an absence of focal points other than in square Q-3. A similar pattern also exists around square E-17 and J-9.

The connectivity between deciduous value cores in Fig. 3 largely follows the pattern of its coniferous counterpart, with some differences. A key difference is that the connectivity is significantly stronger between the deciduous value cores, likely due to the higher number and concentration of focal points. In relative terms the

connectivity is also weaker around square D-25, F-3 and Q-3 due to a lower density of focal points in these particular areas.

The total connectivity between all focal points can be seen in Fig 3. The connectivity mainly mirrors the one seen between deciduous value cores. The impact from the coniferous connectivity can be seen in the form of a slightly stronger connectivity in squares D-15, F-3, Q-3 relative to the deciduous connectivity. The relative

Figure 4. The town Staffanstorp and the current densities within it. Interestingly, the current density is higher inside the town than around it.

18 connectivity compared to the coniferous connectivity can however instead be seen to be dampened in the same squares, as well as in R-5 and H-17. This is due to the much higher abundance of deciduous focal points relative to coniferous focal points.

4.1.1 Agricultural connectivity and potential areas for perennialization Below, in Fig. 5 the agricultural connectivity is presented.

A pattern for both coniferous and deciduous connectivity is that many of the areas with the strongest connectivity is here absent, due to them being

dominated by forests rather than agricultural fields. Areas with strong connectivity nonetheless remain, where the top 10 % are marked in green in the figures. The common denominator among these areas is that they are located a) in or between areas with a large quantity of focal points, such as for along the border between I-7 and L-5 or b), in areas that have few or no focal points but are located in an area where focal points exist outside the area and that contain patches of land cover types that support a higher current density, such as around the intersection between squares K-3 and L-4 for the deciduous connectivity.

Figure 5. The current densities in agricultural fields in the Skåne plains. All land covers aside from arable land is hidden.

The green colour represents the highest 10 % of current densities in the production area.

19 The areas assumed to have the highest potential to increase the structural

connectivity by the introduction of perennial crops are the top 10 % current densities marked with green. The main difference between the deciduous and coniferous potential is that the coniferous is overall slightly lower. This can be observed

particularly in the areas between squares K-3 and L-4, and along the border between I-7 and J-4. The exception is around square D-16 where the relative connectivity is overall stronger for coniferous value cores. Seen to the total connectivity the highest potential follows the pattern for deciduous connectivity. Interestingly, many of these areas exist between residential areas with focal points in their proximity, the pattern in these areas follows that of the case with Staffanstorp and Skivarp.

4.2 V

G

P

The coniferous connectivity is presented in Fig. 6 and statistically summarized in Table 4.

Table 4. A statistical summary of the values resulting from the modelling of the Västra Götaland plains. The summary includes the total area as well as the separate parts that resulted from the splitting of the area.

Västra Götaland Plains, coniferous

Västra Götaland Plains, deciduous

Max current density: 33836 48882

≤ 98 % current density: 2161 2966

Focal points: 3189 3988

Western max current density:

42865 60657

Western ≤ 98 % current density:

2302 2375

Western focal points: 1960 2307

Eastern max current density:

13641 24039

Eastern ≤ 98 % current density:

1081 1899

Eastern focal points: 1577 2180

20

Figure 6. The modelled connectivity in the Västra Götaland plains production area, after value core forest type and in total.

The legends in the current density maps specify the cumulative current densities presented as quantiles. A land cover map with focal points are also included.

Some of the areas with the strongest connectivity can be seen along the left border of the map from square C-9 to F-16. This is likely because these areas are dominated by forests with few interruptions and rich in value cores. Another area with high

connectivity is along the waterline in squares F- 13/14/15. Some interesting points here is that strong connectivity can be observed some way out in the water between the mainland and some small islands in square F-16, likely due to a high concentration of focal points in a quite small area, which overcomes the

otherwise high resistance of the water. In the other Figure 7. Coniferous current density and resistance layer.

21 squares along the water there are however fewer focal points, with some in square F- 14 and a cluster higher up towards the inland, see Fig. 7. The stronger connectivity here is likely due to a generally low resistance following high structural similarity to the average value core.

In and around the intersection of square E-9 and F-8 an area consisting of mostly forest and wetlands with a multitude of focal points is located, with a

correspondingly strong connectivity.

In and around square I-8 a large area can be observed that on the contrary, has a generally weaker connectivity than many other areas in

the map. The land cover here mostly consists of arable land with overall few focal points. Some exceptions can be found in squares H/I-7 where some focal points are located and patches of forests that support a stronger connectivity. Streaks with stronger connectivity can be found elsewhere in the area as well, mostly

corresponding with land covers other than arable lands, especially forests and to a certain degree along a stream which is belted by some vegetation, see Fig. 8.

From K-6 to K-9 a sharp contrast can be seen in the

current density. This is the precise same place where the western and eastern part of the resistance- and focal point maps overlapping ended and is most likely the reason for the difference in current seen here. This is likely explained by the western map having an overall higher current density than the eastern part, possibly because of a higher number of focal points in the former.

In the eastern part of the maps, several areas with especially weak connectivity can be observed, with examples including around squares L-6 and P-5. The common denominator between these is that they are large areas with a lower density of focal points with few land cover types that efficiently can support connectivity.

The deciduous connectivity can be seen in Fig 6, for which the pattern largely follows that of the coniferous connectivity. The connectivity is strongest along the left border from square C-9 to F-16, as well as along the shoreline in square F-13/14, in and around the intersection of square E-9 and F-8, and along the waterline in squares K- 11/12 and L-12. Although stronger than that of the coniferous, the deciduous

connectivity is weakest in the fields in and around squares I-8 and with exceptions

Figure 8. The current density in a large agricultural area, with some stripes of higher densities in forested areas.

22 around H/I-7, as well some streaks where other land cover classes than arable land can be found. As for coniferous, another area with especially weak connectivity can be found in square L-6.

Several differences occur as well. Although generally stronger, the deciduous connectivity is weaker around squares G-17/18, varying between absolute and relative terms. The explanation for this is likely a higher resistance due to a higher structural difference to value cores for deciduous value cores compared to that of coniferous value cores, see Fig 9.

A clearly stronger deciduous connectivity can be observed in square L-6 than that of coniferous, which for the latter is one of the areas with the weakest connectivity.

An interesting area to compare the deciduous and coniferous current densities is in squares M/N- 9/10. Both areas have high current densities, but with differing

characteristics. The deciduous current is a mixture of some strong current corridors with a high current density “spread out” around them, while the coniferous current

has some larger strong areas and corridors with less connection between them. Both are rich in focal points as well as having similar resistances in the area, but where the deciduous has several, smaller and more spread out value cores, the coniferous has larger value cores located more closely together. In square N-9 the city of Skövde is located, where the deciduous current is significantly stronger inside as well as closely

around the city, likely due to having more focal points located in square O-9, located directly to the east of the city. An overview of the area can be seen in Fig. 10.

In squares K-7 to 10 a clear contrast could be seen where the resistance- and focal points map had been divided for the coniferous map. The same area in the deciduous map appears to be seamless. This might be because the deciduous focal points are more evenly distributed across the area.

As for Skåne plains, the weighted total connectivity in Västra Götaland plains largely resembles the deciduous map due to a higher number of deciduous focal points.

Figure 10. Deciduous (left) and coniferous (right) current densities in the area surrounding the city of Skövde.

Figure 9. The coniferous (right) and deciduous (left) resistance around squares G-17 & G-18.

23 Some of the clearest differences, relative to the deciduous, are in squares G-17/18 where the connectivity is slightly stronger, N-14/15 which has slightly stronger connectivity as well, and a slightly weaker connectivity in square L-5/6.

4.2.1 Agricultural connectivity and potential areas for perennialization The agricultural connectivity is presented graphically below in Fig. 11.

Seen to the strongest 10 % of the

connectivity, little to none can be found on arable lands. To get any significant visible results, the highest 70 % must be included. This is likely due to the overall significantly higher current density in the Västra Götaland plains, with highly connected forest with corridors that usually do not require the crossing of any agricultural fields.

Areas with relatively strong connectivity can be seen in different parts of the production area. These areas include the left border along squares C-8 to E-14, in squares L/M- 12, M-10 and in squares E/F- 8/9. The pattern for all these areas is that they are located in areas surrounded by forests rich in value cores. This can

especially be observed in squares E/F- 8/9, which is located between two areas dominated by forests, most of which is made out of value cores. The areas with the weakest agricultural connectivity can instead be found in the middle of the map

Figure 11. The current densities in agricultural fields in the Skåne plains. All land cover aside from arable land is hidden.

In this area, none of the highest 10 % of current densities could be found within arable lands.

24 surrounding squares I- 7/8/9 and J- 8/9, which has a significantly lower density of focal points and are dominated by arable land with few land cover types that efficiently support connectivity.

The coniferous agricultural connectivity largely follows that of the deciduous with some exceptions. The main differences are, relatively speaking, a slightly stronger connectivity between the western border along squares C-10 to E-14 and the

waterline along squares E-10 to F-12, as well as along the waterline in squares I-11/12.

A weaker connectivity compared to some of the strongest for deciduous can instead be found in the areas in squares L-11/12 and M-10. A generally weaker connectivity can also be observed I- 7 to 9 and J- 8/9, as well as in the agricultural areas all along the eastern border. The weaker connectivity in these areas can likely be explained by fewer focal points in the close vicinity to agricultural areas, especially in areas O/P- 8/9.

The total connectivity has patterns from the deciduous as well as the coniferous present. In the area along the right border from squares M-7 to P-9, the connectivity is slightly weaker than that of the deciduous, following the areas weaker coniferous connectivity. The same pattern can be seen in the fields around squares I-7/8/9. On the opposite, squares G-15/16, as well as J-13 shows a somewhat higher current density than that of the deciduous, resulting from the higher coniferous connectivity in these areas. Notably, although more strongly affected by the patterns of the

deciduous current density, the impact from the coniferous current density relative to that of the deciduous is overall greater than that which could be observed in the Skåne plains, due to a smaller difference in the numbers of focal points.

4.3 T

In the section below, the results from the testing area is presented. It starts with a general description of the different results following different resolutions. This is followed by separate description of the connectivity between different forest types and their different resolutions. For this area, the potential for strategic

perennialization has also been modelled and described. A statistic summary of the area is presented in Table 5 and graphically in Fig. 12.

25

Table 5. A statistical summary of the values resulting from the modelling of the testing area.

Testing area, coniferous Testing area, deciduous

Max current density, 10x10: 20 762

≤ 98 % current density: 1,39 113

Focal points: 32 206

Figure 12. The modelled connectivity in the testing area, after value core forest type in total and after resolution. The legends in the current density maps specify the cumulative current densities presented as quantiles. A land cover map with focal points are also included.

26 4.3.1 General differences between different resolutions

When comparing the maps with different resolution in general, the clearest difference is just that: different resolution. The maps with a resolution of 10x10 m have better details, which might be an advantage when investigating areas with sharp contrast in land cover classes, such as along streams in agricultural fields which are often lined with vegetation that are only a few pixels wide. An example of this can be seen in Fig. 13. Other examples

include lines of trees in agricultural fields, also pictured in Fig. 13.

The importance of a higher resolution might be more important for areas with fewer focal points, such as for the coniferous value cores.

When comparing the difference between resolution for the coniferous maps, further differences that are more pronounced appear.

In squares T/U-3/4 some current density can be seen, which largely is absent in the lower resolution map. Many more finely vegetated areas nearly disappear as well, such as the stream in squares J-7 to 8 and lines of trees along roads and in

agricultural fields to mention a few. It should however be noted that the general picture is coherent between the resolutions and that areas with higher densities are overall present with both resolutions.

The 20x20 resolution is overall blurrier, which partly is natural with the lower resolution, but which also is an effect that might be reinforced by the method with which the resolution is changed. In several areas the lower resolution seems to have a

“shine” around areas with high current density. This is because the resistance layers were resampled by assigning each pixel the average value of the surrounding pixels, which gives a “transition zone” between areas of differing resistance as well as current density. This can be seen in, e.g., square O-15

4.3.2 Deciduous value cores 4.3.2.1 10x10 meters resolution

The higher resolution deciduous connectivity can be seen in Fig. 12. Some of the strongest connectivity can be found in squares M-14 along the shoreline to Q-11, as well as right above in squares N-14 to R-13. Further corridors stretch from here to the top of the map where it curves west- and south to another lake, which is largely

Figure 13. A comparison of resolution for deciduous connectivity. The higher resolution can be seen to the right, with the lower resolution to the left.

27 circled by an area with strong connectivity. Similar patterns can be seen stretching south- and east in a curve from R-11 to T-8, as well as along the southern border from squares H-3 to N-3. The general pattern is that this connectivity exists in forested areas, connecting to focal points, either at the ends of these corridors or within them.

Examples of areas where the focal points are located at the ends of corridors are in square R-11 to T-12, around the lake from square H-13 to E-13 and between squares K-4 to N-3. All along the eastern border, an area with a generally weak connectivity can be found. Some focal points exist in this area, but these are fairly isolated in large areas of agricultural fields that does thus not efficiently support connectivity.

Nonetheless, some corridors can be identified along thin strips of vegetation in the agricultural fields. In the bottom right corner, an area with some more pronounced corridors can be seen. This is the place where the town Sjöbo is located, in which vegetated areas such as parks and churchyards exist, which enables connectivity.

4.3.2.2 20x20 meters resolution

The lower resolution map overall follows the pattern of the higher resolution, but some significant differences can be seen. Along the waterline from square M-14 to Q- 11 where for the higher resolution some of the strongest connectivity was found, the lower resolution map instead displays a connectivity that is not much stronger than that which can be found in some agricultural fields. This is surprising, as the area largely consists of value cores, meaning it has the lowest possible resistance for both resolutions. The area with the low resistance is in places somewhat thinner, in which the interruptions that exist are slightly larger, but not to a degree that seems likely to explain the degree of difference between the resolutions. The same thing can be seen in squares P-17, along the waterline in G-12 to H-13 and G-14 to F-14, S-12 and M-3 to M-4. No satisfying explanation has been found for the differences in these areas.

Otherwise, the overall pattern for the lower resolution map is that the areas with the strongest connectivity are somewhat more pronounced, likely resulting from that land cover with higher resistance become “blurred” from the rescaling method so that their resistance becomes lower. As in the general description, areas with thin strips of vegetation carrying currents over land covers with generally high resistance largely disappears in the lower resolution map, such as along the eastern border.

4.3.3 Coniferous value cores

The coniferous maps overall differ significantly from the deciduous ones. This can be explained by a much lower number of value cores and a significantly weaker

28 connectivity, likely resulting from a generally higher resistance due to a higher

structural difference to its value cores, as well as smaller value cores.

4.3.3.1 10x10 meters resolution

The coniferous current density is the highest all around the lake in square G-13, squares G- 8 to 13, squares M- 9/10, N/O-9 and O/P-8, K-4/5/6 and along the western border in squares B-8, 9 and 17. The common denominator between these squares is that they are the places that either contain focal points or are located closely between focal points. The relatively strong connectivity below the lake around square N-8 is likely explained by that it mostly consists of coniferous forest with high structural similarity to its value cores. Much of the same can be said about the other mentioned squares, with the difference that they consist of much smaller patches of trees that do not connect well to surrounding areas.

4.3.3.2 20x20 meters resolution

The areas with the strongest connectivity are the same areas as for the higher resolution, with the difference that – as for its deciduous counterpart – they are slightly more pronounced due to a “blurriness” in the resistance layer, resulting in that land covers with higher resistance is softened and areas with lower resistance become slightly lower if located near pixels with lower resistance.

Again, as for deciduous, the lower resolution map loses most thinner stripes of current in more isolated landscapes.

4.3.4 Agricultural connectivity and potential areas for perennialization

The agricultural connectivity for deciduous and coniferous value cores can be seen in Fig. 14.

Figure 14. The current densities in agricultural fields in the testing area. All land cover aside from arable land is hidden. The green colour represents the highest 10 % of current densities within agricultural fields, rather than the highest 10 % of the entire area.

29 None of the agricultural fields for any type of value core reached a connectivity strength within the top 10 % range for the total area. Instead, the top 10 % of the current density found within the actual agricultural fields are here marked with green. The fact that none of the fields was within the top 10 % is possibly explained by a poor presence of vegetation intercepting fields to any larger extent.

As in the case for areas with stronger connectivity in general, the areas marked with green for the deciduous value cores are located in areas adjacent to focal points.

Additionally, the green areas all appear in areas and corridors with strong connectivity outside of where the agricultural fields are located.

For coniferous value cores, only a negligible area with a connectivity within the strongest 10 % is found. The ones that yet exist are primary located in square B-8/17.

As per usual, these are located between focal points in between streaks of stronger connectivity directly outside of the fields.

4.3.5 Modification based on strategic perennialization

The fields selected for perennialization can be seen to the left in Fig. 15. The result from the perennialization is presented to the right in the same figure.

Figure 15. The deciduous and coniferous areas that were modified in accordance with perennialization, marked in green, and the result from perennialization.