**3.4 Statistical analysis**

This study included 24,819 respondents located using their residential coordinates. Using GIS, the presence/absence of each of the five characteristics within 300 m from each respondent was defined based on the criteria in Table 2. The working process for the assessment of the characteristics is described in *Results* below. Table 3 shows the percentage of the population living close (300 m or 100 m) to the different characteristics. Spearman´s rank correlation coefficient, appropriate for investigating associations between ordinal scales, was used to test associations statistically between the number of characteristics (0-5) present within 300 m or 100 m of the respondent's residence and ordered answers to the survey questions. P-values below 0.05, and equivalently 95% confidence intervals for odds ratios excluding unity in ordinal regression analyses with adjustments for a broad list of individual determinants of health, were regarded as statistically significant (Björk et al. 2008).


Table 3. Percentage of the population that has the different characteristics within 300 and 100 meters, respectively, distance from their residence (Björk et al. 2008). Note that all individuals that have a certain characteristics within 100 meters also have it within 300 meters distance from their residence.

Each respondent's coordinates are defined as the centre point of the complex in which he/she lives. That centre point can be quite far from the position of the person's home, making the 100 m distance incorrect. 300 meters is a fairly normal distance to walk to get to a nature area or a park.

The Agricultural Landscape for Recreation 231

Fig. 2. Time spent on moderate physical activities in relation to the number of recreational values (0–5) of the natural environment within 300 metres distance from the residence (from

The choice of criteria for elaborating each characteristic using GIS was in some cases a long process in which different combinations of classes of parameters for accessing each characteristic were elaborated successively. These trials generally started with a discussion among the research group members to develop a first set of criteria for identification. The first attempt at classification aimed at achieving an appropriate number of areas for the

As seen in Table 3, the scope of characteristics close to respondents on average differs

substantially. Culture and Lush are most common. Wild and serene are most rare.

**3.5 Results – Working process for the assessment of characteristics** 

concerned characteristic, not too many and not too few.

Björk et al. 2008).

One outcome of the study is that the objectively GIS-assessed availability of the five characteristics near one's residence (< 300 m) is positively associated with neighbourhood satisfaction (Figure 1), moderate physical activity (Figure 2) and, among tenants, low BMI. Thus, figure 2 suggests that individuals spend more time on average on moderately demanding physical activities the more characteristics they have within 300 m from home. This association remained after adjustment for individual (socioeconomic) factors.

The impact of the number of characteristics on BMI was less clear. After adjustment for individual factors associated with BMI, the beneficial effect of the characteristics was present among tenants but not among house-owners. The proportion of obese (BMI > 30 kg/m2) individuals among tenants was 17% in residences with zero characteristics within 300 metres compared with 13% in residences with at least one characteristics present (Björk et al 2008). No clear association between the number of characteristics and self-rated general health was detected after adjustment for individual factors.

The result for neighbourhood satisfaction among people living in flats is remarkable. If all five characteristics exist within 300 m, 70% of the tenants are satisfied with their neighbourhood, whereas a maximum of 50% are satisfied if only one or no characteristics are present. The corresponding figures for house owners are 83 % and 74 %. Consequently, house owners seem to be rather satisfied with having their own garden, while tenants' wellbeing is highly dependent on having good natural environments or parks within 300 m from home. This provides important input to the current debate on global warming and densification in urban planning. We are supposed to live more densely in cities to minimize commuting to our workplaces, but on the other hand, if we do not have sufficient nature and park qualities close to our homes in the cities, we will need to commute into the rural landscape for recreation, and that also has an impact on the climate effect.

Fig. 1. The relation between the number of recreational values (0–5) of the natural environment within 300 metres distance from the residence and the percentage reporting high neighbourhood satisfaction among house-owners (N = 13,930 answers) and tenants (N = 5,942 answers) (from Björk et al.2008).

One outcome of the study is that the objectively GIS-assessed availability of the five characteristics near one's residence (< 300 m) is positively associated with neighbourhood satisfaction (Figure 1), moderate physical activity (Figure 2) and, among tenants, low BMI. Thus, figure 2 suggests that individuals spend more time on average on moderately demanding physical activities the more characteristics they have within 300 m from home.

The impact of the number of characteristics on BMI was less clear. After adjustment for individual factors associated with BMI, the beneficial effect of the characteristics was present among tenants but not among house-owners. The proportion of obese (BMI > 30 kg/m2) individuals among tenants was 17% in residences with zero characteristics within 300 metres compared with 13% in residences with at least one characteristics present (Björk et al 2008). No clear association between the number of characteristics and self-rated general

The result for neighbourhood satisfaction among people living in flats is remarkable. If all five characteristics exist within 300 m, 70% of the tenants are satisfied with their neighbourhood, whereas a maximum of 50% are satisfied if only one or no characteristics are present. The corresponding figures for house owners are 83 % and 74 %. Consequently, house owners seem to be rather satisfied with having their own garden, while tenants' wellbeing is highly dependent on having good natural environments or parks within 300 m from home. This provides important input to the current debate on global warming and densification in urban planning. We are supposed to live more densely in cities to minimize commuting to our workplaces, but on the other hand, if we do not have sufficient nature and park qualities close to our homes in the cities, we will need to commute into the rural

This association remained after adjustment for individual (socioeconomic) factors.

health was detected after adjustment for individual factors.

landscape for recreation, and that also has an impact on the climate effect.

Fig. 1. The relation between the number of recreational values (0–5) of the natural

= 5,942 answers) (from Björk et al.2008).

environment within 300 metres distance from the residence and the percentage reporting high neighbourhood satisfaction among house-owners (N = 13,930 answers) and tenants (N

Fig. 2. Time spent on moderate physical activities in relation to the number of recreational values (0–5) of the natural environment within 300 metres distance from the residence (from Björk et al. 2008).

### **3.5 Results – Working process for the assessment of characteristics**

The choice of criteria for elaborating each characteristic using GIS was in some cases a long process in which different combinations of classes of parameters for accessing each characteristic were elaborated successively. These trials generally started with a discussion among the research group members to develop a first set of criteria for identification. The first attempt at classification aimed at achieving an appropriate number of areas for the concerned characteristic, not too many and not too few.

As seen in Table 3, the scope of characteristics close to respondents on average differs substantially. Culture and Lush are most common. Wild and serene are most rare.

The Agricultural Landscape for Recreation 233

The best areas cannot be drowned in an overly generous classification, which is a general problem in most landscape analysis. Therefore, this buffer zone was rejected from our classification. Despite this rejection, Lush was classified close to a large percentage of the

Beaches, dunes, and sand plains, bare rock, sparsely vegetated areas, burnt areas, natural grassland, moors and heath land, all forests larger than 25 ha. Plus slopes more than 10 degrees (creating viewpoints), farmland pointed out in a preservation plan and coastal zone

In our first attempt, we used minimum size criteria: Forest 100 ha; natural grassland 20 ha; heath 50 ha. This resulted in very few areas. One observation was that many open space areas are small, but together form large open spaces. Therefore the size criteria for grassland and heath were rejected in our classification. Moreover, forests of different categories are in reality rather small, but together form large forest areas, giving the impression of "entering another world". To address this, we reduced the size criterion for

In our first attempt, we also used noise level criteria for roads 250 m from the residence, but rejected these as being too high in many cases. At first we also used 800 m minimum distance to wind power aggregates as a criterion for obstructing a feeling restfulness, but

Non-urban parks, farmland pointed out in a preservation plan, areas of national interest in

In our first attempt, we also included national interests in recreation. Such areas often overlap national interests in cultural preservation, and are in these cases included anyway. However, they also include large areas for recreation without meeting criteria for the Culture characteristic, but instead for some of the other characteristics, so that dataset was rejected from Culture. Culture was classified as being close to a relatively large percentage

When conducting a regional analysis of environmental perception, there are a great many pitfalls to consider. Regional studies require existing datasets mainly produced to show objective data, but perception is a subjective interpretation of the real environment. People´s perceptions also differ from individual to individual due to their previous experiences. This, however, is a general issue independent of the scale of the study. For these kinds of studies, the elaboration of classification criteria always has to be done by representatives of the local

population.

**3.5.4 Space** 

forest to 25 ha.

**3.5.5 Culture** 

**4. Discussion** 

community.

later rejected this as well.

of the population (see Table 3).

preservation in a national plan.

Excluded are areas with a noise level over 40 dB(A).

cultural preservation, nature reservation areas.

The resulting maps were examined by the project staff and compared to their own experiences of certain areas in the studied region. After critical studies of the resulting maps, other classification criteria were tested in a second attempt, and so on in further attempts until the final classification versions were established – those found in Table 2. At the end of the paper, you can see the five maps of the final classification.

### **3.5.1 Serene**

The effort to develop the classification model can be illustrated to some extent by looking at the work with the characteristic Serene. For Serene, we first used a minimum size criterion: Areas should have a minimum size of 20 hectares to be classified as Serene. The reason for this was that an area must be relatively large to be perceived as Serene. From the beginning, we also claimed that the average noise level should not exceed 40 dB(A)24 h, and that a 250 m buffer along roads and dwellings was required. Moreover, we required a buffer of 800 m around large wind power generators and shooting ranges.

Later on, the area size criterion for Serene proved to be useless, as a single large forest often consists of several serene land use areas grouped together, each of which is smaller than 20 hectares. Therefore we rejected that size criterion. We also rejected the minimum distance criterion to roads and wind power generators. The noise criterion, however, was sharpened to 30 dB(A)24h. This is perhaps the reason for the relatively rare presence (Table 3) of Serene near the residents.

### **3.5.2 Wild**

Forests, thickets, bare rock, mires and wetlands, lakes and rivers – all larger than 15 ha, if not closer than 1 kilometre from villages and towns. If closer, there is no minimum size for the area, because children seem to sense the Wild characteristic even in very small areas. Also areas steeper than 10° are classified as Wild.

Excluded are areas with a noise level over 40 dB(A), and areas with wind generators within 800 m.

In our first attempts, we set the minimum size of areas to 30 ha and to 5 ha within 1 km from villages and towns. But that reduced the amount of Wild too drastically.

### **3.5.3 Lush**

Mixed forest, marshes and mires, beaches, dunes, sand plains and bear rock. Plus all registered "key biotopes", certain inventories of pasture land, biodiversity areas, bird biotopes of regional interest, also Nature 2000 objects and national parks.

Many of the mixed forests are very small areas, down to 25 x 25 m integrated in the open agricultural landscape, and these are of great importance to biodiversity from an ecological as well as recreational point of view. At our first attempt we included a buffer zone of 100 m. The reason was that fringe zones between different biotopes are very species rich, and we wanted to capture that by including a buffer zone in the biotope area concerned. However, this resulted in areas that were far too large to make the classification useful. One hundred meters is an overestimate of specious rich fringe zones, they are narrower than this.

The best areas cannot be drowned in an overly generous classification, which is a general problem in most landscape analysis. Therefore, this buffer zone was rejected from our classification. Despite this rejection, Lush was classified close to a large percentage of the population.
