A SUITABILITY MODEL WITH FUZZY LOGIC: Wine Industry Suitability Under Changing Climate

(1)

Magister thesis, 15 hp

Master’s Program in Human Geography with specialization in Geographical Information Systems (GIS) 60 hp

Spring term 2021

A SUITABILITY MODEL WITH FUZZY LOGIC

Wine Industry Suitability Under Changing Climate

Yasin Cepnioglu

(2)

ABSTRACT

The grapevine is one of the most cultivated crops in the world. It requires several environmental conditions besides the winemakers’ practices. The practices of the winemakers have a great impact on the quality of the product,

yet it is also depending on the pre-existing features of the cultivation’s sites.

Those features are categorised into 3 as atmospheric, topographic and soil conditions and each contains several criteria derived from earlier studies.

These conditions are generally not expected to change for centuries, but atmospheric conditions are subject to considerable change due to the climate

change. The categories are used to build the Vineyard Suitability Model with fuzzy logic and applied to the study area, Skåne, Sweden. The model is applied

in 4 versions which are according to the current and 3 scenarios of atmospheric conditions according to climate change projections for 2050. The

model results indicate that the study area currently possesses a low potential for the wine industry to emerge. However, the impact of climate change appears to turn a considerable size of cultivated area into high potential sites.

The results appear to be valid since the study area is just above the conventional wine production latitudes in the Northern hemisphere and

considered to be too cold for grapevine production.

(4)

1.INTRODUCTION

The wine industry is a traditional socio-economic activity for numerous locations. The oldest known cultivation site dates back to 8000 years ago, in an area in present day’s Georgia, in Eastern Europe (Grainger & Tattersall, 2016). The popularity of the industry is more than it has ever been especially in terms of customer demand and the product quality of the established wine regions is argued to be better than it had been in history (Grainger &

Tattersall, 2016). Nowadays, there are very few processed agricultural products that can compete in price with top quality wines (White, 2015). The historically established production areas such as many regions in France, Italy and Spain are expanding in response to the demand, besides, there had been emerging locations housing the industry. California is an example of a relatively new wine industry region that show good results with its wines that are regarded as high quality as the European competitors.

The geographical origins of the vineyard areas provide identities for their wines that are widely referred to as their terroir. Terroir is very important for the quality of the wine for technical reasons along with the practices of the winemaker (Dougherty, 2012; Grainger &

Tattersall, 2016; White, 2015). Besides, especially in western countries, vineyards create touristic interest in their vicinity due to their cultural popularity. Therefore, they provide an appealing image for their localities both for visitors and prospective future residents, as well as offering economic inputs. The majority of the larger wine-producing regions are situated in between the latitudes 30° and 50° in both hemispheres (Grainger & Tattersall, 2016). This area is called ‘temperate zone’ and regarded as advantageous for the grapes to satisfactorily ripen.

Ongoing climate change is already showing impacts on the existing vineyards. The observed temperature from 1950 to 1999 shows an overall increase and it appears beneficial for the wineries in that period, yet, if the warming effect continues as projections suggest, some of the high-quality wine regions are at risk of exceeding the upper threshold temperatures for preserving quality (Jones, 2007; White, 2015). Besides, it is expected for more arid regions such as those found in Spain and California to face difficulties with crucial irrigation sources since the competition for water sources is expected to be more challenging (Gerós et al., 2015; Jones, 2007). Therefore, winemakers have to consider the possible outcomes of the current trend for the future of the industry.

Climate change does not only mean negative impacts on the wine industry but also provides opportunities for formerly unfavourable regions (Nesbitt et al., 2018). A previous occurrence of climate change (referred to as the ‘medieval warm period’) is regarded as the golden age of wine for Britain between the dates 950 and 1200 (Grainger & Tattersall, 2016). It is debatable if there will be another golden age for colder regions but there are already emerging commercial vineyards on latitudes where it was not previously considered possible and there are few studies done for colder regions (Wanyama et al., 2020).

As White (2015) points out, determining suitable sites is only the first step of a long way

towards establishing a successfully functioning wine industry, yet it is important to avoid

costly mistakes. Geographical information systems are crucial aid on this matter since they

can reduce the costly trial and error processes by revealing high potential sites (Kurtural et

al., 2007). Moreover, the suitability conditions need to be considered together with climate

change projections for avoiding mistakes and spotting opportunities.

(5)

1.1.AIM AND RESEARCH QUESTIONS

This study uses geographical information systems to analyse the suitability of Skåne county in Sweden for an emerging wine industry. Sweden is well above the temperate zone since it is located in between the latitudes 60.13 and 18.64 in the northern hemisphere. However, there are already a number of vineyards already operating in Sweden for over a decade. A government-owned liquor store named Systembolaget is the only place where beverages with above 3.5% alcohol can be sold in Sweden and on their website, by May 2021, there are 48 white and 10 red wine options that are produced in Sweden (Systembolaget, n.d.). It is unclear how many vineyards are operating in the entire country but the southernmost county, Skåne, is currently housing vineyards that are producing 35 white wine and 7 red wine option, therefore, appear to be an important supplier area for commercial wine production.

The study aims to test a fuzzy logic model and to provide results showing the vineyard suitability state of the study area for today and the expected state of suitability according to climate change projections by 2050. The use of fuzzy logic for vineyard suitability appears new in academia, but it is promising since the suitability criteria are often hard to limit to certain values (Badr et al., 2018; Chrobak et al., 2020; Nesbitt et al., 2018). Modelling with fuzzy logic allows researchers to use a continuous range from 0 to 1 rather than evaluating the values with rigid categorical ranges. This study provides a model with criteria similar to earlier studies, but the model is configured to fuzzify those into a spectrum accordingly to the suitability conditions derived from a broader range of literature and combine them with overlay methods designed for fuzzy logic. The model may present a useful aid for future site selections, and also, can be used as a tool for testing the criteria’s themselves against observations on existing sites.

RESEARCH QUESTIONS

In terms of physical geography, does the study area offer favourable conditions for grapevine production?

Does changing climate have any impact on the spatial distribution of the vineyard suitability

conditions?

(6)

2.SUITABILITY CONDITIONS FOR EMERGING WINE INDUSTRIES

The methods of winemaking alone cannot replicate the quality of fine wines alone because the emergence of a successful wine industry requires suitable sites for vineyards to start with (White, 2015). The suitability depends on several major criteria of influence in the physical geography context. They can be considered in groups as atmospheric and topographical influences, and soil types (Jackson, 2008). Those factors can be manipulated in artificial ways which can be considered under winemaker practices and special breed grape varieties can provide additional resistance against physical geography obstacles. They can be desirable options depending on the case-specific pros and cons conditions, even though the additional winemaker practices would add to the costs of production, and special breed grape varieties may need to be tested by time for their demand in the wine market. Nevertheless, pre- existing suitable conditions are still regarded as valuable assets for commercial vineyards for high-quality wine production and referred to as crucially helpful for avoiding costly mistakes (Kurtural et al., 2007).

2.1. ATMOSPHERIC AND TOPOGRAPHIC INFLUENCES

Atmospheric influences are concerned mainly with the conditions in regard to temperature, solar radiation, water and, to some extent, wind. Macroclimate is a very important factor because grapevines are vulnerable to high heat and cold during growing seasons.

Correspondingly, most of the vineyards are located between the latitudes 30 and 50 for both northern and southern hemispheres (Grainger & Tattersall, 2016). Besides the macroclimate, local weather conditions also greatly impact the suitability together with local topographies.

Milder temperatures (10°C - 35°C) may be observed largely between the given latitudes, yet it varies depending on the topography as well. For every 100 meters, the annual temperatures go lower by about 0.5°C (Jackson, 2008). Lower altitudes are commonly favoured, yet higher altitudes may be more favourable in areas with warmer climate conditions.

The concept of growing degree days (GDD) is widely used for evaluating the temperatures for vineyard suitability (Chrobak et al., 2020; Grainger & Tattersall, 2016; Jones et al., 2004;

Kurtural, 2007; Nesbitt et al., 2018; Wanyama et al., 2020). It is generally applied for the period between April and October in the northern hemisphere which is regarded as the growing season. To make the calculations, the concept uses 10°C as the base since grapevines are known to grow in milder temperatures as mentioned above. Every Celsius degree above the base comes as the score of each day. The days with lower temperatures are not counted.

Then, the scores for all the days in the growing season are summed up and the result gives the growing degree days value for the evaluated spot. Since the average temperatures fluctuate between years, the studies typically use long term weather observations.

Bud is a small part of plants that grows into a leaf or flower. Grapevine buds stay in dormancy during winter with a protective shell. The time after winter when the green leaves start to appear on the buds is called bud break. Freezing is dangerous for grapevines in growing seasons which starts with the bud break. Prolonged exposure to below 0°C after bud break can wipe out all the yield of a season (Grainger & Tattersall, 2016; Sgubin, 2018). White (2015) suggests that about 180 frost-free days are required for a healthy vineyard while Badr et al.

(2018) set a ‘critical range’ with a minimum of 150 frost-free days. Winter frost is favourable

(7)

because it is good for the plant health and may provide some natural pest and disease control, but severe winters which include periods with temperatures below -20°C can damage the plants (Jackson, 2008; Kurtural et al., 2007; Wanyama et al., 2020). On the other hand, warm winters are also not desired since they may cause double yield which shortens the life of the plant (Grainger & Tattersall, 2016).

One way to ease the negative impacts is to place the vineyards on slopes because cold air is denser than warm air, therefore, the air with freezing temperatures may flow down the slopes (Jackson, 2008). The desired angle of a slope is between 5%-15% (Badr et al., 2018). Below 5% may be insufficient to have any impact while above %15 may cause operational difficulties.

Besides, a slope within the given range provides natural water drainage (White, 2015). Slopes can also allow the vineyards to benefit from solar radiation reflection coming from near water bodies (Jackson, 2008).

Water is obviously needed for the grapevines to survive, but the roots of the plants can reach meters down in the soil and find enough water, therefore, it is often not a serious problem if the climate is not extremely dry. However, extensive water in the soil and humidity can cause more serious risks related to plant diseases. Therefore, climate conditions may mean requiring a drainage system or locating the vineyard on a slope with a sufficient angle (Jackson, 2008; White, 2015). Winds are not commonly considered crucial for vineyard site selection. In areas identified with particularly strong winds may cause damage, but there are various ways of overcoming such negative impacts, for example, ‘shelterbelt’ is a frequently used method both against wind and erosion (Jackson, 2008; White, 2015). One common application of shelterbelt is planting a couple of layers of trees on the side where winds are blowing towards the vineyards. Therefore, the damaging impact of strong winds gets broken, and the grapevines are protected.

Along with the localities’ temperature, the exposure to the solar radiation of the grape leaves and berries can have beneficial and damaging impacts. The exposure is mostly related to the topography of the sites and the increased availability of solar radiation generally provide nuances to the grapes that are regarded as high qualities but excessive exposure to sunlight can cause a damaging impact in arid environments. Typically, a vineyard that is located in an area with the lower end of temperature levels would benefit from maximising the solar radiation while the areas with higher temperatures would benefit more from milder exposures. For the areas with a colder climate, locating the vineyards on the slopes facing towards the equator is the most favourable for obtaining higher quality products (Badr et al., 2018; White, 2015). The latitude that the vineyard site is situated is also a considerable input since the length of the days in terms of time varies in different latitudes.

2.2. SOIL INFLUENCES

The roots of grapevines spend their entire lives in the soils which may host them for centuries

(White, 2015). There is no consensus about a perfect type of soil or bedrock suitable for

winemaking, and their impact on the wine flavour is also a debated topic. However, the areas

underlain by limestone are known to be hosting a considerable amount of successful wine

regions (Davis, 2010; White, 2015). Therefore, the limestone bedrock type is an appealing

feature for a new vineyard site amongst winemakers. Other famous bedrock types which host

(8)

successful terroirs are schist (Porto) and chalk (Chablis) (Jackson, 2008). Moreover, according to Dougherty (2012), geological findings show that the boundaries between different classification zones correspond to the underlying bedrocks’ borders.

Jackson (2008) argues that having the soil in a uniform composition in a site is more important than the composition itself because non-uniform soils can cause differing berry growth which negatively affects the quality. There are 4 main categories of soil according to the

‘International Scheme’ that concern the vineyards for the most areas (Jackson, 2008). Those are coarse sand, fine sand, silt and clay (respectively from lighter to heavier). There can be one soil type dominant in an area as well as multiple soil type mixes. Technically, grapevines can survive on each of them, yet clay dominant soils are not favoured since they are keeping the water in them more effectively (Dougherty, 2012) and higher water holding capacity may cause plant diseases. Besides, in clay soils, the heat is absorbed by the water during the day and on hot days this water evaporates, and the heat leaves the soil together with the water.

The soils with more sandy content have better drainage and the heat during the days are kept in the sandy particles rather than water content and it is reflected out during cold nights. This feature of soils with some sandy content is especially important against frost risk. However, if an area is dominantly covered with sandy soil, it may not hold enough water for the plants to thrive.

Another suitability indicator for soil quality may be the pH levels which impacts the mineral

availabilities (Jackson, 2008). The pH range between 5 and 8.4 is good enough for a vineyard

site (Badr, 2018). According to Jackson (2008), the exact pH level is not important as long as

it is not too low or too high. White (2015) suggest between 5.5 and 7.5 pH level as high

potential site feature, while Badr (2018) suggests between 6.7 and 7.2 as ideal pH range.

(9)

3.METHODOLOGY

3.1. ASSESSING SUITABILITY BY USING GEOGRAPHICAL INFORMATION SYSTEMS (GIS)

The suitability criteria above are crucial for assessments of possible vineyards sites and there have been research applications for academic purposes. Using GIS in this context appears rather new in academia but there are also a growing number of publications emerging in the last decades (Badr et al., 2018; Chrobak et al., 2020; Jones et al., 2004; Kurtural et al. 2007;

Nesbitt et al., 2018; Wanyama et al., 2020). The studies typically build suitability assessment models which consider each suggested criterion within themselves first, and then, combine the outputs by applying an overlay to provide an overall assessment throughout the studied areas.

Overlay analysis on GIS is a method that is used to combine multiple data layers on a subject area. It is advantageous for researchers because today’s computational capacity is sufficiently high on affordable hardware costs, therefore, researchers are able to analyse large areas with vast amounts of data. Besides, there are several software that are developed for GIS applications which are constantly optimized to run fast and accurate and provide an increasing number of tools for more fine-tuned analysis.

An earlier attempt to investigate the possibility of terroir is done by Jones et al. (2004) for Umpqua Valley of Oregon, USA. The researchers base their analysis on four suitability categories which are topographical, soil, land-use and climate suitability and they also have subcategories. Topographic suitability has the sub-categories for elevation, slope and aspect;

soil suitability has “water holding capacity”, pH level, drainage and “depth to bedrock”.

Climate suitability is assessed according to growing degree days calculations and then further refined by adding the precipitation data and the data about frost free days as well as last frost in spring and first frost in fall. Land use data is used to find the areas that are actually available for agricultural production.

They use the weighted overlay method to overlay the subcategories of topography, soil and climate to provide a unified raster evaluated according to vineyard suitability. The method includes a classification process for the subcategories. Each subcategory values are classified into a scoring schema according to their suitability status and classified raster layers are produced in a unified form for weighted overlay. The weighted overlay method allows users to combine the classified raster with assigned percentages of influence which must sum up to 100. Besides the level of influence, users may restrict any area with an undesired value such as urban areas. After evaluating each category individually, the categories are overlaid together to build a raster surface for the overall vineyard suitability index of the study area.

The results show that most of the existing vineyards are situated within favourable areas yet there are still large areas with higher suitability scores that can be used by existing and future grapevine producers.

Kurtural et al. (2007) utilize a similar approach with categories for macroclimate, mesoclimate

and land attributes. The macroclimate category includes subcategories for growing degree

days calculations and the number of days that extreme cold (below -26) is observed between

the years 1969 and 2002. Mesoclimate includes slope, aspect and elevation. Land attributes

include the drainage level of the soil, organic matter percentage of the soil and current land

use. The researchers prefer the weighted sum method to overlay these data layers. This

(10)

method does not include a prior classifying process and simply multiplies each raster data value with a weight value that is given by the user. Therefore, the subcategories are reclassified into suitability score raster dataset before the overlay process. The researchers design the weighted sum to have the highest possible score as 100. Up to 40 points may come from macroclimate, up to 40 points may come from mesoclimate and up to 20 points may come from land attributes.

The overlay produces a vineyard suitability index for their study are which is scored from 0 to 100. They categorise their results into best, good, fair, risky and unsuitable areas. They consider the areas with above 85 points to be the best while the ones with below 55 points are unsuitable. The analysis showed that 9% of the vineyards are situated in the best areas while 75% are on good and 16% are on fair areas. These results provide a guide for new vineyard projects but also indicates some assurance for the accuracy of the model since most of the existing vineyards are located either in the best or good category areas. However, those rigid ranges of suitability scores can be misleading since the difference between an area with 84 points and another area with 85 points may not be meaningful, even though, they are in different categories.

A recently published study that is conducted by Wanyama et al. (2020) also uses similar categories as Jones et al. (2004) and the weighted sum method for assessing land suitability for grapevine production in Michigan, USA. They state that Michigan has a rapidly growing wine industry that dates back to the 1970s. However, the precipitation variables are divided into 2 groups as the precipitation in growing days (early season) and precipitation in the late season. This nuance is added because the precipitation in early times of the season is desirable but during the late season, it brings an increased risk of low yield. Moreover, average temperatures from March to June (‘spring temperatures’) to further elaborate the model for higher potential areas, and topographic depression areas (‘sinks’) cold air and water gets collected which may cause significant difficulties for vineyards. Furthermore, the researcher classified growing degree days calculation into two categories according to the suitability for white wine grapes and red wine grapes. Their study shows the current impact of warming temperatures on the wine production potential in Michigan and the use of GIS for modelling those impacts.

The weighted sum method for creating multi-criteria suitability models provide fruitful results as the examples above presents. However, it is also argued that the relative impacts of exact values of the criteria are rather debated and often vaguely defined (Van Leeuwen et al., 2006).

For example, the slope gradient percentages from 5 to 10 are regarded as the most suitable

in the literature and scored as 10 in a scoring range of 0 to 10 while the slope percentages

from 1 to 4 are marked with the score 5 (Kurtural, 2007; Wanyama et al. 2020). The difference

in the scores are significantly different in these groups but in reality, for example, there may

not be as dramatic difference observed between a place with a slope percentage of 4.95 and

another place with 5.05. We would rather expect quite similar effects. To address this

limitation, the fuzzy logic may be an accurate aid since it offers a continuous classification to

deal with uncertainties (Badr et al., 2018) as the example given above. Instead of setting rigid

boundaries between the different classes, fuzzy logic assigns a value in a continuous range

between 0 and 1. The users are allowed to set the shape, direction and spread of the curve

on a mathematical basis. Larger or smaller values can be favoured, and upper and lower limits

can be set if needed. Middle range values can be also set as high ranked with a Gaussian

(11)

function. Fuzzy classification with gaussian function assumes a bell shape curve and rewards the values that are close to the predetermined middle point. Each criterion gets reclassified into a fuzzy format and then fuzzy overlay may combine those layers through mathematical or logical operators.

Chrobak et al. (2020) conducted one of the studies that attempt to include fuzzy logic to GIS- based vineyard suitability analysis and site selection. Besides the criteria that are resembling the above-mentioned studies, they point out that there are also other relevant but non- countable information about possible sites is present and it could be utilized by using fuzzy logic. They also suggest that high-quality data may not always be available or can be too costly to obtain, fuzzy logic also provides an alternative approach that can increase the accuracy of analysis even when the available data is below the needed levels for an analysis in ‘classical’

type logic, such as weighted overlay or weighted sum methods.

Another study from recent years, published by Nesbitt et al. (2018), applies a multi-criteria suitability analysis with fuzzy logic for England and Wales. As aforementioned, the areas suitable for wine production are conventionally expected to be located in between the latitudes 30 and 50, and neither England nor Wales is within those borders. However, climate change appears to stretch the borders above those borders. Consequently, vineyards are blooming with growing investments which sums up to a 246% increase in terms of area (722 to 2500 ha) in England and Wales during the period between 2004 and 2017. However, the researchers suggest that the site selection processes have not been through a systematic spatial comparison towards finding the optimum areas.

The study they conducted fills this gap by providing a model for systematically assessing the vineyard sites in their study area based on GIS with fuzzy logic. Their results indicate risks and opportunities associated with the existing vineyards and the other potential areas. The study considers only arable, horticulture or grassland as potentially suitable areas, and it should also be addressed that the higher potential areas may not be preferred because of the lack of marginal advantage compared to other possible uses of those areas. They suggest their model may provide guidance for future initiatives towards establishments with higher yield prospects.

3.2. VITICULTURE IN SWEDEN

The history of viticulture in Sweden is relatively young; it reaches only back to the 1990s. It is

likely that there had been grapevine cultivations during milder the climate era (between the

years 950 and 1200), yet it is not possible to refer to an established Swedish wine production

culture today since the climate has been unfavourable for many centuries. However, there

had been growing interest in wine production at the southern end of the country and this can

also be related to the warming impacts of climate change in recent decades. Especially in

Skåne county, there are already over 20 vineyards operating, and the trend is likely to spread

to other counties as well as within Skåne. It is important to provide a robust method of

analysing the suitability of the cultivatable landscape for avoiding mistakes. Moreover,

measuring the change of viticultural suitability according to the current projections in Sweden

contributes to the body of knowledge in human geography research.

(12)

This study follows a similar methodology to the works mentioned above (Badr et al., 2018;

Chrobak et al., 2020; Nesbitt et al., 2018) by applying a multi-criteria suitability analysis based on GIS and fuzzy logic. The evaluations are done both for the existing state of suitability and the expected state according to the climate change projections in 2050. The study aims to provide a data-driven assessment model for the possibility of wine industry development in Sweden as well as capturing the expected impacts of climate change on Swedish viticulture.

3.3. METHODS

ArcGIS Pro version 2.7.3 (ESRI, 2021) is used for data processing and analysis.

3.2.1. ASSESSMENT CRITERIA

Following the prior studies, the criteria that are used for the analysis in this work are grouped into 3 categories which are atmospheric conditions, topography, soil conditions. The criteria under the categories and the evaluation methods are applied following the earlier studies’

frameworks (Badr et al, 2018; Chrobak et al., 2020; Wanyama et al., 2020). The analysis is done by using 4 sub-models that are, in essence, one model in 4 pieces. The first 3 models are for evaluating the subject area under each category of criteria and the final model is to combine the results of those.

The first model is named as Atmospheric Conditions Model (ACM). It analyses the subject area according to the precipitation and air temperature under 4 criteria which are growing degree days, yearly precipitation, growing and harvest period precipitations. Criteria such as cold days (below -20°C) and the number of frost-free days in the growing season are not added to the model since they have no impact on the results (further explained below). Then, the criteria values are fuzzified by using the fuzzy membership tool of ArcGIS Pro.

The second is Topographic Conditions Model (TCM). The form of land in the subject area assessed by the model according to the percent angle of the slopes, the azimuthal angle (aspect) and topographic depression (sink). Criteria data is generated by using a digital elevation model, and then, their values are fuzzified. The third one is Soil Conditions Model (SCM). It evaluates the soil types according to their drainage quality, bedrock types and the depth of soil. The soil types are reclassified on a 0-10 scale prior to fuzzifying since the drainage conditions data is not available itself. Bedrock types are also reclassified instead of fuzzified. They are reclassified as 1 if they are Limestone, Schist or Chalk otherwise classified as 0. Bedrock is considered as an addition to soil drainage criteria in the model. It increases the suitability score of the soil considerably if they are available while their absence decreases the suitability to a very limited degree. After processing or fuzzifying criteria data, all three models above uses the fuzzy overlay tool of ArcGIS Pro for combining the output datasets.

TCM and SCM provide a single raster for each model that carries values between 0 and 1 for

each cell. The ACM provides several rasters depending on the number of climate change

scenarios applied.

(13)

The final model is named Vineyard Suitability Model (VSM), and it does the overlay for the results of the first 3 models. The model results are combined in an interrelated fashion in equal weights by using the ‘Gamma’ fuzzy overlay type with a gamma value of 0.5. The overlay type is further explained below, together with the results of VSM. Table 1 gives a summary of the criteria and the fuzzification methods. The sources of the data that is used to feed the models for this study is also referred to below, in the results section.

Table 1. Criteria and fuzzifications summary table.

Fuzzy Membership Type Settings

Atmospheric Conditions Model (ACM) Growing degree days (GDD) in 4 scenarios

Near Midpoint: 1300

Spread: 0.000005

Growing season precipitation Near Midpoint: 63.5

Spread: 0.005

Harvest season precipitation Small Midpoint: 74

Spread: 0.005

Average yearly precipitation Near Midpoint: 762.5

Spread: 0.00002 Topographic Conditions Model (TCM)

Slope Near Midpoint: 10

Spread: 0.015

Aspect Near Midpoint: 180

Spread: 0.0001

Sink Linear (inverse) Minimum: 10

Maximum: 0 Soil Conditions Model

(SCM)

Soil drainage (Soil types) Linear (Manually reclassified prior to fuzzification)

Minimum: 0 Maximum: 10 Bedrock type Linear (Manually reclassified prior to

fuzzification)

Minimum: 0 Maximum: 10

Soil depth Linear Minimum: 0.5

Maximum: 6

3.2.2. CULTIVATED AREAS AND EXISTING V INEYARDS IN SKÅNE COUNTY IN 2021

The models evaluate the suitability for the entire Skåne county in Sweden, there are many

different kinds of land use that are hosted in a county; therefore, it is needed to define the

eligible areas of grapevine farming for evaluating the results and building discussions. For this

study, only the cultivated areas in Skåne are considered eligible and the polygon feature

classes showing the cultivated areas are obtained from Lantmäteriet (2019), the Swedish

mapping, cadastral and land registration authority in Sweden. Furthermore, the suitability

conditions in cultivated areas are measured exclusively for vineyards, therefore, relative

advantage of vineyards compared to other possible activities at the sites are not covered by

this study.

(14)

As aforementioned, Sweden is above the latitudes that are conventionally considered as suitable for grapevine production and that is mainly because of the insufficient air temperature level. Skåne is the closest county to the favoured latitudes and the average air temperatures are significantly higher compared to most of the other counties in Sweden which also show the potential (Figure 1). Moreover, Sweden, in general, is promising for research in this context since the quantity and quality of publicly available and geographically referenced data is favourable.

Figure 1. Average yearly air temperature ranges in Sweden for the time period 2000-2010 in 50m resolution (Data Source: Meineri & Hylander, 2017)

(15)

However, data about the existing vineyards seem to do not exist. It is useful to obtain the data about existing vineyards in the county for it is used to show the impacts of climate change on their lands’ suitability status, and also, it is used for testing the models for errors. Therefore, it is produced for this study by applying the following process. First, the company names and addresses are found by making an online search which provided 23 companies’ information.

Then, their locations are found on the map by using Google Maps(n.d.). Names of the companies and coordinates of points within the vineyards are recorded in an excel table.

Then, the excel table is turned into a CSV file to use in the ArcGIS Pro project. Then, a new point feature class is produced by using XY to point tool with the recorded coordinates. Then, the satellite imagery provided in ArcGIS Pro is used for drawing the polygons for the vineyards.

Google Maps street view is also used when there is hesitation about where to start and end the polygons.

4. RESULTS

4.1. ATMOSPHERIC CONDITIONS MODEL (ACM)

Atmospheric conditions are evaluated under 4 criteria with a model build for this study (Appendix, Figure 7). The criteria are growing degree days (GDD), growing season precipitation, harvest season precipitation and average yearly precipitation. The frost status in the study area is also an important criterion that may severely damage the plants, but the initial analysis for the study area show that the dangerous level of frost is rarely present, and all the study area appears with maximum suitability in terms of frost risk. Consequently, the frost risk does not change the suitability status in any way for the study area, therefore, excluded from the model. The criteria for GDD is calculated by the following formula:

(Average daily air temperature of 1^st growing season day - 10°C) (Average daily air temperature of 2^nd growing season day - 10°C) (Average daily air temperature of 3^rd growing season day - 10°C)

…

(Average daily air temperature of last growing season day - 10°C) Growing Degree Days (GDD) Value

The days with below 10°C air temperature are not counted, therefore, there are only positive

values for each day that determines the GDD. The data for daily air temperatures are derived

from 17 observation stations located in Skåne from the website of the Swedish

Meteorological and Hydrological Institute (SMHI, n.d.). The data included the air temperature

values for each day between April 1

^st

and October 31

^st

at 17 weather stations throughout the

county. While some of the records are covering more than 100 years, some others are limited

to a few years. For this study, the mean air temperature on each growing season in the last

10 years are considered as a daily average. The growing season also includes the harvest

period and covers the time between April and October.

(16)

Besides the historical data, the projections for 2050 are derived from SMHI (2015). According to the study, a 1-2°C increase in air temperatures is expected for Skåne county by 2050. For addressing the impact of climate change in air temperatures, 3 alternative versions of the GDD calculations are applied with 1°C, 1.5°C and 2°C higher daily air temperature levels. The additional air temperatures increased the number of days above 10°C, which in turn significantly increased the levels of GDD values. Four GDD calculations are made for the current status and 3 possible scenarios of the aforementioned amounts of increase in air temperatures.

After setting the GDD values, the inverse distance weighted (IDW) interpolation tool (in spatial analyst tools of ArcGIS Pro) with 10mx10m cell size is used to propagate the values since the stations are situated at certain points. The observation station points also have height values which already addresses the elevation related temperature differences). The method spreads GDD values throughout the county, the assigned value is derived from the surrounding observation stations and the impact of each station to the cell value is weighted according to how near the station is. A separate raster dataset is produced for each GDD calculation which resulted in 4 raster datasets for different air temperature scenarios.

To assess the suitability in relation to precipitation, it is analysed with 3 sub-criteria. First, a yearly precipitation feature class is derived from the geodatabase produced by SHMI (2014).

The feature class shows the precipitation values in millimetres for both current status and 2050 projections on grids covering 16km

²

. Yearly precipitation values between 635mm and 890mm (Goldammer, 2018) are regarded as suitable for vineyards. The study area has no grid with a lower value than this range, yet there are several areas with higher values that can cause damage due to overwatering.

In the model, the grids of current conditions and 2050 projections are turned into 2 separate rasters. Then, the raster values are fuzzified (fuzzy membership) with ‘Near’ function with the midpoint of 762.5 and spread of 0.00002 which rewards the precipitation values between 635mm and 890mm with the highest score (1.00 or very close) and decrease with the increasing distance from the midpoint for both higher and lower precipitation values.

Yearly precipitation gives an overall picture for the subject areas but the rainfall during the growing and harvesting season also important to check separately because higher amounts of precipitation during growing and especially harvest period could decrease the quality and yield. Besides, more rainfall also signals more cloudy weather which decreases solar exposure.

In colder and more rainy climates, the areas with lower precipitation in growing and harvest periods are strongly desirable.

Following the methods of Wanyama et al. (2020), 2 more criteria for precipitation are added to the model by using monthly precipitation data from 23 observation stations located in Skåne county (SHMI, n.d.). The values are the mean of 10 years (2010-2020) observations for each month. The growth period criterion uses the precipitation value means belonging to June, while the harvest period criterion uses the mean of August, September and November.

Then, these values are laid on all the county by using IDW interpolation with 10mx10m cell

size. Then the resulting raster for growth period values is fuzzified by using fuzzy membership

type ‘Near’ with the midpoint of 63.5 (mm) and spread of 0.005. This setting rewards the

average precipitation values for June. The raster for harvest period values is fuzzified by using

(17)

fuzzy membership type ‘Small’ with the midpoint of 74 (mm) and spread of 0.005. This setting rewards the lower values of precipitation with the higher suitability score.

Then the 3 sub-criteria of precipitation are combined by using the fuzzy overlay method with the ‘And’ overlay type which only takes the lowest value of the cell, therefore, the negative impacts of excessively rainy or arid climates are reflected on the resulting raster datasets. This overlay type is relevant since the sub-criteria are likely to correlate. The overlay is done 2 times separately for current and projected precipitation status in 2050. In the end, the model once more applies a fuzzy overlay process to combine the precipitation and GDD conditions to reach the overall atmospheric conditions suitability raster. Again, to avoid correlation related errors, the ‘And’ overlay type is used. The current precipitation conditions are combined with current GDD conditions, while the expected precipitation for 2050 is combined with the GDD calculations of increased air temperature scenarios.

Figure 2. Atmospheric Conditions Model results.

ACM provides the resulting maps shown above (figure 2). Average suitability throughout the county appears only 42.6% while the best possible spot is 64.1%. Therefore, the model suggests that the current atmospheric conditions are less than ideal in the study area.

However, the increasing air temperature scenarios change the image considerably. 1°C

increase in the temperatures most dramatically and changes the average suitability to 56.8°C

and the suitability of the best spot in the county reaches to 84.6% suitability score. The other

scenarios with higher temperature values further increase the average as well as the

maximum suitability percentage for the best spot. This proportionally larger impact is mostly

propelled by the number of days that is currently just under 10°C. 1°C increase in air

temperatures allowed those days to be included in the GDD calculations, and in turn,

increased the overall GDD level marginally further.

(18)

It is visible on the maps that the areas with higher suitability scores are spreading with higher air temperatures. However, the spread does not take place equally through the county because the precipitation values remain high and even increases at some locations. Especially, on the north-western corner of the county, precipitation does not only prohibit the increase in suitability conditions but also spreads the lower suitability values to a larger area.

4.2. TOPOGRAPHIC CONDITIONS MODEL (TCM)

Topographic conditions are evaluated with 3 criteria that are slope, aspect and sink in a model named as Topographic Conditions Model (TCM). The slope refers to the percentage of the inclination of the slope, while the aspect refers to where the slopes are faced to. Sinks are also known as topographic depressions where cold air and water gets collected

(Wanyama et al., 2020). Figure 8 in the appendix show the model in diagram form. A height raster dataset with 2m x 2m cell size obtained from Lantmäteriet (2019) to use as a digital elevation model (DEM). Then, the DEM raster dataset is aggregated into 10 m x 10 m cell size (by taking mean values of the contained smaller cells). The new cell size still provides a high detail DEM which gives accurate results for this study areas scale. The model is tested in much smaller areas with both cell size alternatives to verify the accuracy.

The raster datasets for slope and aspect is generated by using surface tools in the 3D analyst toolset in ArcGIS Pro (Esri, 2021). The sink raster dataset is generated by, first, using the ‘Fill’

tool in spatial analyst hydrology toolset and then using the ‘Minus’ tool in ‘Math’ toolset to decrease the height values of DEM from the fill raster, therefore, generating a raster dataset that only contains the sink areas.

After generating the criteria raster datasets, the cell values are fuzzified. For slopes, the fuzzy membership type ‘Near’ is used with a midpoint of 10 and a spread of 0.015 is used which rewards 5-15% slopes with higher membership scores. For aspect, again, ‘Near’ is used with a midpoint of 180 (South) with a spread of 0.0001 which rewarded the south-facing slopes with high scores while the others have lower depending on how far their value of the angle is compared to 180. Flat areas are assigned with the value 100 (corresponds to east) to receive an average score (about 0.6) since flat areas cause neither advantage nor disadvantage in terms of aspect. For sinks, a linear membership type is used with a minimum value of 10 and a maximum value of 0. This membership type rewarded the very shallow sinks areas with higher scores and deeper sinks with lower scores. The sinks as deep as 10 meters or more receive the lowest fuzzy membership score, 0.

Finally, the model puts these 3 criteria together with the fuzzy overlay type ‘Gamma’ which brings a result out of interrelations between the subject raster layers rather than bringing the value of one of them as ‘And’ or ‘Or’ types do. Gamma is a combination of 2 other fuzzy overlay types, product and sum. Product type multiplies 3 raster cell values to find the overlay score which would never be more than the minimum value, but it can be very low. Sum is the opposite; it uses the following formula:

Fuzzy Sum Value = 1 – (1 – cell value 1 x 1 – cell value 2 x ……)

. This formula provides a value that can never be lower than the highest but can get very high.

Gamma overlay type combines these two types’ results with a gamma rate which is in

between 0 (Product) and 1 (Sum). In this model, the gamma value is set as 0.5 which brings a

value in the middle of ‘product’ and ‘sum’ calculations. The output raster dataset shows the

topographic suitability of each cell location with values between 0 and 1.

(19)

Figure 3. Topographic Conditions Model results.

The results are shown in figure 3. The areas with yellow colour tones are the highest suitability areas which are combinations of southern face slopes with above 5% and without topographic depression (sink). The distinctively blue areas are the topographic depression areas with more than 10 meters depth or too steep areas with above 30% slope. The flat or nearly flat areas bring forward the dominant, green, colour with average (lighter green) to lower (darker green) suitability conditions. The areas that are covered with water appear on the digital elevation model as flat areas, thus, they appear as if they are having an average suitability condition. To avoid misleading results, water-covered areas are excluded from the topography conditions model results and appear as empty (white) in figure 3.

4.3. SOIL CONDITIONS MODEL (SCM)

High-quality wines are known to grow in various soil conditions and there is no consensus on

the best soil type as aforementioned. But there are some favourable features that are widely

accepted as increasing the likelihood of high quality and yield. Soil Conditions Model (SCM)

(Appendix, figure 9) analyses the soil suitability for vineyards with 3 criteria, drainage, bedrock

type and soil depth. The data is obtained from the Geological Survey of Sweden (SGU) in

polygon feature class form for bedrock (SGU, 2017) and soil type (SGU, 2014), and in raster

dataset form for soil depth (SGU, 2017).

(20)

Naturally, the grapevine roots need some soil depth which should be at least about 0.9 meters and it is ideal to have 6 meters or more (Gladstones, 2011). Thus, the ‘Linear’ fuzzy membership type is used with a minimum value of 0.5 and a maximum value of 6 to assess soil depth suitability and provide a raster dataset with the fuzzified values. The lowest value is set as 0.5 instead of 0.9 because it is possible that the plants grow above half a meter of soil, yet it is less suitable, and this fuzzy membership type assigns just a lower score for those areas instead of zero.

The bedrock type’s impact on suitability is not agreed upon by all, but there are certain types of bedrocks that are widely observed in high-quality areas. Limestone and schist two of those that are present in Skåne. Since there is no exact value that can be referred to, both of the bedrock types are assigned with the highest membership value (1) by reclassification and all the other areas are reclassified into the middle value as 5. Then, a linear fuzzy membership function applied where the minimum value is 0 and the maximum is 10.

The soil criterion value has also been difficult to determine since the available data about the soil is hard to match the suitability evaluations according to the literature. But some types of soils are known for better drainage than others, especially clay-rich soils known to have poor drainage. The soil types are scored as it is shown in the table below (Table 2), and then, the polygon feature class of the soil types are turned into a raster dataset with those values. Then, the resulting reclassified raster is fuzzified linearly with a minimum value of 0 and a maximum of 10.

Table 2. Reclassifications of the Soil Types.

Soil Type Reclassification Value

Rock 7

Glaciofluvial sediments 10

Clay - Silt 5

Moraine 8

Moraine - Clay 7

Postglacial sand -Gravel 10

Peat 8

After that, the soil type and bedrock suitability rasters are combined with fuzzy overlay type

‘Gamma’ with a gamma value of 0.6. Therefore, the areas with one of the suitable bedrock types received higher soil suitability score if not already highest while the areas without the favourable bedrock types received lowered scores to a limited extent. In the end, another fuzzy overlay is applied to combine the soil and bedrock suitability with the soil depth. ‘And’

overlay type is used to make sure no shallow soil areas or clay-rich areas are assigned higher

scores. The resulting raster dataset shows the soil conditions raster with fuzzified values.

(21)

Figure 4. Soil Conditions Model results.

The SCM results in a map (Figure 4) that favourable areas are more determined by the presence of the advantageous soil types in terms of drainage. Yet, the bedrock types of limestone and schist are increasing the suitability values noticeably stretching from southeast to middle region. The region with high suitability on the middle eastern side also partly receives the higher value from the bedrock type of limestone. The areas with less than 0.5- meter soil depth or covered with water bodies receive zero suitability value. The blue colour symbolizes the areas with lower than 30% suitability. Most of those areas are without soil (or water-covered) but there are also large areas on the northwest corner where the suitability conditions are very low due to shallow soil.

4.4. VINEYARD SUITABILITY MODEL (VSM)

The final model’s (Appendix, Figure 10) purpose is simply to combine the outputs of the earlier

3 models. 4 fuzzy overlays are applied since there are 3 climate change scenarios in addition

to the current climate. SCM and TCM models’ outputs are added to all the fuzzy overlays as

they are, but each climate scenario version of ACM is added to a separate overlay process. All

the overlay types are set as ‘Gamma’ with gamma value 0.5. 70% overall suitability is selected

as the threshold for favourable cultivation sites only for presentations of the results of this

application and does not reflect a common practice derived from the literature.

(22)

The VSM results in figure 5 show the final outputs of the models. The suitability in current conditions appears quite low overall. The results show that majority of the cultivated areas in the county has lower than 50% suitability which goes even below 30% on the north-west. This particularly low conditions are largely coming from the high precipitation and unfavourable soil conditions in the northwest. On the other hand, the southwest offers comparatively higher potential. However, the portion of the cultivated areas with favourable conditions (70% and above suitability rate) corresponds to only 0.13% of the total with 0.58 km

²

area.

Figure 5. Vineyard Suitability Model results.

The scenarios with increased air temperatures together with the precipitation projections show a substantial increase in the overall suitability picture. 1°C increase in air temperatures provides much more area compared to current conditions which increase the total area with favourable conditions to 21.01 km

²

that corresponds to 4.46% of the total area. 1.5°C increase nearly doubles this impact. The favourable conditions seem to spread further with increasing air temperature levels, but the marginal increase goes lower. In terms of increasing the rate of favourable areas, the difference between the increase from 1.5°C and 2°C is much lower compared to the increase from current conditions to 1°C higher. Moreover, the impact of higher air temperatures does not spread through the county evenly because of the prohibiting factors coming from the aforementioned high precipitation and unfavourable soil conditions.

This prohibiting impact is observed in large areas on the northwest.

(23)

5.DISCUSSION

The emergence of a wine industry depends on atmospheric, topographic and soil conditions alongside the winemaker knowledge and practices. However, climate change comes with opportunity and challenges by shifting the suitability status for areas (Wanyama et al, 2020).

Skåne appears somewhat in short with favourable conditions for wine production in current conditions, yet it also appears on the favoured side of climate change since there is a considerable size of area turning favourable for grapevine production.

The areas with less than 30% suitability are not strictly unsuitable but they are not optimal according to input data that are presented under the 3 categories introduced above. The ranges of suitability values are taken from the literature. Still, the values suggested in the literature are accorded for the conventional wine production conditions which use the well- known and cultivated Vitis Vinifera varieties such as Cabernet Sauvignon, Syrah, Sauvignon Blanc, Chardonnay, Merlot or Riesling, but there are thousands of other grapevine varieties that may fit better to different local conditions.

Figure 6 show the locations of the vineyards currently operating in Skåne. The suitability status of the whole county is already quite low, and the northern side of the region is comparatively lower. However, there are plenty of vineyards in the north. The average suitability on the areas of 5 vineyards appears less than 30% which seems to be a contradiction for the validity of the model results. However, it appears there are 2 grapevine varieties widely used in the county, namely, Piwi and Solaris. Both of these varieties are known for their resistance against plant diseases which are the main reason for the prohibiting impact of high precipitation. Solaris variety is also known for ripening earlier which is an advantage for areas where warm days are present for shorter periods. These are examples of specially bred grapevine varieties for below optimal conditions; hence, the presence of them can explain the contradictory results and supports the model’s expectations instead.

The specially bred varieties seem to be accurate choices since only 0.58 km

²

area (0.13% of total cultivated areas) comes out as high potential sites for the whole county in current conditions. But they are not the only choice according to model prediction for 2050. The model results expect the high potential areas to increase into about 5% to 9.5% of the total cultivated areas. Those areas offer more flexibility with grapevine varieties and more suitable for conventional winemaking, therefore, it may become a competitively more advantageous agricultural investment alternative.

The new vineyards may take advantage of the impacts of climate change. Starting and

operating a vineyard is costly and it can take years to first harvest. Moreover, there is

considerable interest from touristic visitors and some of the current winemakers already have

facilities for wine tourism guests where they offer set menus and vineyard tours (Mullen,

2019). Therefore, it would be expected from new vineyards to also employ such side practices

and that means even more cost for investment. Consequently, it is important to know the

suitability of the potential site for grapevine farming. The VSM model offers a systematic and

robust method of assessing suitability. The results may provide crucial guidance before the

costly mistakes take place.

(24)

Figure 6. Existing vineyards’ locations on the Vineyard Suitability map for current conditions.

5.1. LIMITATIONS

The model is designed according to the criteria suggested in the literature but there are several limitations in the results which are related to the input data. In the atmospheric conditions model, the raster dataset for air temperatures is generated by applying an interpolation process. The input data is from 17 station spots and taking the average of the last 10 years’ air temperature values. Interpolation may give a good overall picture but may not reflect the conditions on specific sites. Therefore, the model with current input only shows a generalized picture on a regional scale. Input data with better quality and resolution of climate data can provide more nuanced results. Moreover, mean values of 10 years may not be enough to determine the current average, however, taking longer time periods into account may also give an inaccurate result considering the climate change’s impact. Besides, climate change may not only mean changing patterns of air temperatures and precipitation but also increasing extreme weather events that may be triggered by climate change.

Vineyards are highly vulnerable to weather extremes such as hailstorms.

(25)

There are also several limitations to address in the soil conditions category. PH values of the candidate areas’ soils are often included in the evaluations to add further nuance. It is not added in this study because of the lack of a data source. However, pH values, in general, do not present a crucial advantage or disadvantage according to the literature. A more serious limitation is the difficulty of classifying the soil and bedrock types into suitability scores. The difficulty comes, firstly, from the fact that the soil types’ drainage quality is usually referred to in categorical, non-numeric, values such as good and bad. Furthermore, even those values for soil types are mostly classified according to the clay silt and sand content of the soils is not the form of classification in the data available.

Another limitation is related to the cultivated areas in the records which may change over time. Some of the cultivated areas may turn to, for example, urban areas or some of the forest may open for cultivation. Some of the areas with higher suitability conditions are actually on the areas which are classified in land-use records as forest, therefore, changes in those classifications over time may also change the vineyard suitability of conditions of the cultivated areas overall.

6. CONCLUSION

The wine industry is popular in both cultural and economic means around the globe.

Producers of grapevine in the largest scale are China, Italy, the USA, Spain and France respectively and it had been thought that further northern regions are too cold or rainy for wine production (Wanyama et al., 2020), but these conditions may have changed or will change in future with the impact of climate change. However, there are few studies done for colder regions, and Skåne county is just above the conventionally favoured latitudes.

Applying a suitability model for grapevine production in the study area adds to the research by both assessing the viticultural suitability in colder regions and also testing a geographical information systems model based on fuzzy logic. The results for Skåne show that climate change is turning a considerably large size area favourable for a wine industry in the conventional sense, even though, the current conditions seem to reflect fairly low suitability.

The quality of the results of the model greatly depends on the quality and the resolution of

the input data. The data used for this work appears sufficient, still, future studies checking

the validity of these results may add further nuances. Moreover, the model parameters for

the suitability ranges of the criteria can be rearranged according to the requirements of the

special grapevine varieties or winemaker practices. In other words, the model can be

accorded for the suitability conditions that are more relevant for relatively new wine

production methods and grapevine varieties such as Rondo, Solaris and Piwi.

(26)

APPENDIX

Figure 7. Diagram of the Atmospheric Conditions Model (ACM).

Figure 8. Diagram of the Topographic Conditions Model (TCM).

(27)

Figure 9. Diagram of the Soil Conditions Model (SCM).

Figure 10.Diagram of the Vineyard Suitability Model (VSM).

(28)

REFERENCE LIST

Badr, G., Hoogenboom, G., Moyer, M., Keller, M., Rupp, R., & Davenport, J. (2018). Spatial suitability assessment for vineyard site selection based on fuzzy logic. Precision Agriculture, 19(6), 1027–1048.

https://doi.org/10.1007/s11119-018-9572-7

Chrobak, K., Chrobak, G., & Kazak, J. (2020). The Use of Common Knowledge in Fuzzy Logic Approach for Vineyard Site Selection. Remote Sensing (Basel, Switzerland), 12(11), 1775–.

https://doi.org/10.3390/rs12111775

Davis, J. (2010). From oil exploration to winemaking. The Sydney Morning Herald, October 12

(https://www.smh.com.au/lifestyle/john-davis-from-oil-exploration-to-wine-making-20101012-16h80.html).

Derived on 09/04/2021.

Dougherty, P. H. (2012). The Geography of Wine: Regions, Terroir and Techniques (2012th ed.). Springer.

https://doi.org/10.1007/978-94-007-0464-0

ESRI, (2021). ArcGIS Pro (Version 2.7.3) [Computer software].

https://www.esri.com/en-us/arcgis/products/arcgis-pro/resources

Gerós, H., Chaves, M. M., Gil, H. M., Delrot, S., & Delrot, S. (2015). Grapevine in a changing environment: A molecular and ecophysiological perspective. ProQuest Ebook Central https://ebookcentral.proquest.com.

Gladstones, J. (2011). Wine, Terroir and Climate Change. Wakefield Press.

Goldammer, T. (2018). Grape Growers Handbook: A Guide to Viticulture for Wine Production (3rd ed.). Apex Publishers.

Google. (n.d.). Retrieved April 30, 2021, from https://www.google.com/maps/

Grainger, K., & Tattersall, H. (2016). Wine production and quality (Second edition.). Wiley Blackwell.

Jackson, R. (2008). Wine science: principles and applications (3rd ed.). Academic.

Jones, G., Snead, N., & Nelson, P. (2004). Geology and wine 8. Modeling viticultural landscapes: A GIS analysis of the terroir potential in the Umpqua Valley of Oregon. Geoscience Canada, 31(4), 167–178.

Jones, G. V. (2007). Climate change and the global wine industry. Proceedings from the 13^th Australian Wine Industry Technical Conference. Adelaide, Australia.

Kurtural, S., Dami, I., & Taylor, B. (2007). Utilizing GIS Technologies in Selection of Suitable Vineyard Sites.

International Journal of Fruit Science, 6(3), 87–107. https://doi.org/10.1300/J492v06n03_07

Lantmäteriet. (02.07.2019). GSD-Property Map, Vector (6.2.5) [Dataset]. Retrieved from https://maps.slu.se/

Lantmäteriet. (11.06.2019). GSD-Elevation data, Grid 2+ (2.5) [Dataset]. Retrieved from https://maps.slu.se/

Meineri, E., & Hylander, K. (2017). Fine‐grain, large‐domain climate models based on climate station and comprehensive topographic information improve microrefugia detection. Ecography (Copenhagen), 40(8), 1003–1013. https://doi.org/10.1111/ecog.02494

Mullen, T. (2019, January 14). Sweden’s Blooming Wine Scene Is Challenged by Regulations More Than Climate.

Forbes. https://www.forbes.com/sites/tmullen/2019/01/14/swedens-blooming-wine-scene-is-challenged-by- regulations-more-than-climate/?sh=53acff14167a

Nesbitt, A., Dorling, S., & Lovett, A. (2018). A suitability model for viticulture in England and Wales:

opportunities for investment, sector growth and increased climate resilience. Journal of Land Use Science, 13(4), 414–438. https://doi.org/10.1080/1747423X.2018.1537312

(29)

Sgubin, G., Swingedouw, D., Dayon, G., García de Cortázar-Atauri, I., Ollat, N., Pagé, C., & Van Leeuwen, C.

(2018). The risk of tardive frost damage in French vineyards in a changing climate. Agricultural and Forest Meteorology, 250-251, 226–242. https://doi.org/10.1016/j.agrformet.2017.12.253

Geological Survey of Sweden. (05.02.2014). JORDARTER 1:25 000-1:100 0000 (1) [Dataset]. Retrived from https://maps.slu.se/

Geological Survey of Sweden. (16.11.2017). JORDDJUPSMODELL (1.1) [Dataset]. Retrieved from https://maps.slu.se/

Geological Survey of Sweden. (20.12.2017). BERGGRUND 1:50 000-1:250 000 (1.1) [Dataset]. Retrieved from https://maps.slu.se/

Systembolaget. (n.d.). Retrieved May 27, 2021, from

https://www.systembolaget.se/sok/?categoryLevel1=Vin&country=Sverige

Swedish Meteorological and Hydrological Institute. (n.d.). Ladda ner meteorologiska observationer | SMHI.

https://www.smhi.se/. Retrieved April 20, 2021, from https://www.smhi.se/data/meteorologi/ladda- nermeteorologiska-observationer/#param=airtemperatureInstant,stations=all

Swedish Meteorological and Hydrological Institute. (2014). Nerladdningstjänst för SCID. Https://www.smhi.se/.

Retrieved April 20, 2021, from https://www.smhi.se/pd/klimat/rcp_scenario/scid/

Swedish Meteorological and Hydrological Institute. (2015, June). Sveriges framtida klimat - Underlag till Dricksvattenutredningen. https://www.smhi.se/publikationer/publikationer/sveriges-framtida-klimat- underlag-till-dricksvattenutredningen-1.89524

Van Leeuwen, C., & Seguin, G. (2006). The concept of terroir in viticulture. Journal of Wine Research, 17(1), 1–

10. https://doi.org/10.1080/09571260600633135

Wanyama, D., Bunting, E., Goodwin, R., Weil, N., Sabbatini, P., & Andresen, J. (2020). Modeling Land Suitability for Vitis vinifera in Michigan Using Advanced Geospatial Data and Methods. Atmosphere, 11(4), 339–.

https://doi.org/10.3390/atmos11040339

White, R. E. (2015). Understanding Vineyard Soils Second Edition (p. 29). New York, NY: Oxford University Press.