Full text of DOI:10.1093/forestry/73.2.173

Featuring and modelling forest
recreation in Italy
G. SCRINZI AND A. FLORIS
ISAFA, Forest and Range Management Research Institute, Trento, Italy
Summary
This paper reports on interim results of an ISAFA research project aimed at constructing a statistical
model to evaluate the level of visitor attendance at any given forest site, this parameter being
considered the best indicator of the recreational value of the forest.
The paper describes the main steps for defining and modelling the factors and variables involved
in this phenomenon: demand for recreation in woods at a national level, anthropic pressure on forest
resources at a territorial level, and the aesthetic–functional quality of forest sites. These phases
produced sub-models, described and discussed here, that can be useful even when used alone for
management purposes.
The method adopted for collecting the experimental data on actual attendance rates at 41 forest
sites in the Italian Alps is also described, as well as the attempts so far performed to carry out the
final model. The multiple regression approach failed, so an alternative elaboration method, based on
a better definition of some other variables such as accessibility, an increased quantity of experimental
data and neural networks, which at present is still in progress, is also briefly discussed.
Introduction
A set of indicators of this wider approach to
forest resources (including factors related to
The recreational value of forests is becoming one recreation, landscape, hydrogeologic safety and
of the main services provided by woodlands, naturalistic value) now needs to be worked out
especially in those countries where social and and tested for objectivity, ease-of-use and relia-
economic factors have reduced the importance bility so that it can be applied to the complex field
of timber production over the last decades of Forest Management or, at least at a preliminary
(Douglass, 1977; Davis and Johnson, 1987).
stage, as a support to the restricted pool of
Concepts, descriptions and methods taken specific know-how which most regional Forest
from the classic Forest Management approach, Administration rely on.
on which this discipline has been based for 300
Given the lack of worldwide agreed objective
years, are now quite unsuitable if management concepts for defining and measuring the indi-
strategies must take other factors than the pro- cators involved in non-material forest functions,
ductive value of forest into account, as demanded it is very important to define some standard con-
by the current multi-functional approach to forest cepts to measure accurately the demand for recre-
resources.
ational forest activities. In Table 1 and Figure 1
© Institute of Chartered Foresters, 2000
Forestry, Vol. 73, No. 2, 2000

174
FORESTRY
Table 1: Definitions and measure units of forest recreational use related behaviours
Forest: area covered by spontaneous vegetation consisting mainly of trees and shrubs (therefore including also
backwoods, heath, bush, etc.) as well as any unforested area (grasslands, routes, trails, rivers, etc.) situated
within or on the edge of forests.
Attendance at the following zones must not be considered cases of recreational use of forest:
• equipped zones dedicated to specific activities as golf-greens, ski tracks, campgrounds and similar, even though
they are located within a forest area.
• city gardens and parks (forests and parks in surroundings are indeed included in the survey).
Recreation and tourism event in forest (forest visit): circumstance in which an individual attends a forest area for
recreation purposes (tourism, sport, hobbies, rest, and so on) to enjoy entertainment possibilities offered by
forest environment.
Forest visit includes also forest attendance for hunting and fishing, and collecting generic wood products
(berries, mushrooms, etc.) even when these activities can bring an economic benefit to the user.
On the other hand, the following cases are not to be considered as visits in the sense of this survey:
• forest attendance for occupational reasons of whatever type (breeding, agriculture, forestry, surveillance, etc.);
• simple through-transits which do not include stays or long-lasting stops in forest or along in-forest routes;
• attendance at shorelines (at sea or lakes) even when these are provided with coast forests, if the attendance is
carried out uniquely for bathing or similar.
Stay time: time length expressed in hours from the moment of entrance in forest to the moment of leaving.
Recreational use: degree of attendance at a given forest site for the above defined recreation purposes. It can be
expressed in:
• number of users’ transits per surface unit and per time unit (usually per hectare per year, or per day);
or, when possible, in
• number of visitor days: sum of all the hours/visitor relevant to all the users’ attendance at a given forest site,
divided by 12 (Schreuder et al., 1975) or, alternatively, number of transits (visits) multiplied by the average
individual stay length.
the definitions and measurements used in this refers to the ‘open architecture’ of the model,
study are described. They are intended merely as making it amenable to modifications and
suggestions to the investigators concerned in this improvements, as mentioned above.
relatively young discipline, and will certainly be
modified and improved upon.
A preliminary but essential stage of the work
was aimed at evaluating, in quantitative terms,
the role of the forest in Italian people’s recre-
ational activities, at the same time highlighting
the recreational use of forests in Italy (length of
events, type of forests concerned, kinds of in-
Methods and data
The whole research project is based on the forest recreation activities, etc.).
hypothesis that the intensity of users’ forest atten-
Among the different methods for estimating
dance can reliably be chosen as the best indicator recreation demand, including both direct counts
of the forest’s recreational value. It is necessary, made on input fluxes, path transits, numbers
therefore, to construct and test a model for esti- present at a given instant (Schreuder et al., 1975;
mating the level of forest attendance for recre- Karameris, 1982; Tosi, 1989; Tosi and Scrinzi,
ation purposes, taking into account both intrinsic 1994) and indirect procedures such as auto-
and extrinsic features likely to affect the people’s census points, questionnaires, interviews (Koch,
selection and use of forest areas.
1978; Tosi, 1989; Tosi and Scrinzi, 1994), a mail
Figure 2 shows the steps which comprise the poll by questionnaire was chosen as the inquiry
overall study, the variables involved and their method, because of its inexpensiveness and the
relations. The variable indicated as ‘others . . .’ opportunity to cover a large and widespread

FEATURING AND MODELLING FOREST RECREATION IN ITALY
175
very close to the minimum (700) suggested for
this kind of inquiry (Nöel, 1992) and their distri-
bution according to the above parameters was
also well-represented (Table 2), and therefore the
sample seemed to be sufficient to make reliable
estimates.
The questionnaire consisted of 16 questions
with suggested answers and an empty space for
additional remarks or suggestions.
Following the guidelines of a Danish inquiry
which laid down a basis for this kind of study
(Koch, 1978; Koch and Jensen, 1988), the whole
text was structured around the main question
concerning the time interval between the form-
filling and the latest forest visit for each inter-
viewed person. In this way people were not
Figure 1. Graphic illustration of the concept of recre-
ational use of the forest. In a population consisting requested merely to describe their generic habits
of six individuals, A, B, C, D, E and F, 66% of them
(4/6) are forest visitors in a given period (E passes
through by car and F stays outside the forest area).
During that time they go to the forest six times,
making a total of 10 visitor-hours of recreational
use. Individual A makes 50% of the visits (3/6); B,
C and D make 17% (1/6) each. User B, however, has
the most intensive recreational use of the forest with
50% (5/10) of the total visitor-hours. (Koch, 1978;
modified)
but instead to narrate their personal last experi-
ence, at the same time providing the researcher
with a frame of undistorted and substantiated
events which really happened in the more or less
recent past.
From a statistical point of view, the Danish
model did not fit our experimental data (differ-
ences between theoretical and actual frequency
distribution, evaluated through the ꢀ2 test, were
highly significant, in particular for intermediate
classes), probably due to quite different behav-
iours and propensities to a recreational use of the
forest for the two populations. Therefore it was
extensively modified to adapt it to the Italian situ-
ation. A stochastic correlation between the yearly
visit frequency of an individual and the time inter-
val since his latest forest visit had been hypothe-
sized, supposing that the individuals featured
with a high event frequency go to the forest at
time intervals shorter than individuals with a
lower event frequency.
Such a relationship was demonstrated by
observing 43 known-behaviour simulated popu-
lations, consisting of 240 subjects each, for which
it was supposed that the number of forest visits
followed, on long term bases, a Poisson distri-
bution of parameter c (average visit rate per time
unit). Using this assumption, the hypothetical
visits for each individual were distributed along a
10-year time scale. In each one of the 43 popu-
lations a ‘within the last decade’ moment was
randomly fixed, according to which the frequency
distribution relevant to the 240 simulated indi-
viduals were defined, using the same classes of
sample. Questionnaires were sent during the
period April 1992 to March 1993, and were dis-
tributed in 12 monthly mailings in order to
exclude the ‘seasonality’ component from the
estimates, due to the different levels of forest
attendance during different seasons.
The people’s sample was made up of 3000 indi-
viduals from 104 Italian municipalities, randomly
drawn from municipal registry lists, representing
the adult and active Italian population aged
18–75 years. It was composed according to the
Italian population distribution in 1989 (Istituto
Nazionale di Statistica, ISTAT, 1990), according
to sex, age class, geographic area of residence and
class of demographic size of the municipalities,
trying to bring out sub-populations with custom-
ary topic-related behaviours. No cultural or econ-
omic feature was taken into account in choosing
the sample, such as the educational level or occu-
pation, as the inclusion of personal questions
could discourage some people’s response.
Although the return rate of replies all over the
country was lower than expected (670 replies
from 3000, or 22.3 per cent), their number was

176
FORESTRY
Figure 2. Flow-chart of the general model for estimating the recreational value of the forest.
Table 2: Distribution of Italian population according to four main categories and comparison between the
stratification of the selected sample (questionnaires sent) and the final sample (replies)
Questionnaires sent Questionnaires replied
–––––––––––––––––– –––––––––––––––––– Reply rate
Population (%)
No.
%
No.
%
(%)
Sex
Male
Female
49
51
1560
1440
52
48
396
274
59.1
40.9
25.3
18.6
Age (years)
18–25
26–40
41–55
56–75
17.7
29.7
34.2
18.4
127
260
184
99
18.9
38.9
27.4
14.8
Geographic
area
North
Central
South
45.8
19.5
34.7
1426
619
955
47.5
20.6
31.8
392
105
173
58.5
15.7
25.8
27.4
16.9
18.1
Demographic
size of the
municipalities
(inhabitants)
<5000
5000–30 000
30 000–
18.9
34.8
19.5
531
1122
522
17.7
37.4
17.4
136
258
93
20.3
38.5
13.9
25.6
22.9
17.8
100 000
>100 000
26.9
825
27.5
183
670
27.3
22.1
22.3
Total
3000
time distance from the latest visit of the real and time distance from the latest event, the distri-
inquiry. butions within each population were summarized
In order to elaborate a model for the interpre- by the median interval t from the latest visit and
tation of the relationship between visit frequency the average number of visits n (for time units) for

FEATURING AND MODELLING FOREST RECREATION IN ITALY
177
the subjects of each population. Such method above data, in fact, concern only the direct
allowed all the data from any simulation to be ‘exploitation’ of the recreation opportunities
represented by the same model.
offered by forest resources, and cannot account
This relationship was processed by regression for all other (unquantifiable but undoubtedly very
analysis in order to demonstrate the stochastic important) factors related to scenery, landscape
model linking these two variables. As expected, and environment having a basic function in recre-
the average individual number of visits n for time ation activities linked to forests as an irreplace-
unit was negatively correlated with the median able element (mountain tourism, winter sports,
time distance t from the last visit (Figure 3).
After the model application to the real sample,
visits to lakes and coasts, etc.).
Taking into account the extent of forest in Italy
the expected average number of visits n0 was (8.7 million ha: source, National Forest Inven-
calculated from the sampled median time from tory; MAF-ISAFA, 1988), Italian forests receive
the latest visit t0 = 66.2 (days), in turn calculated an average recreation and tourism ‘pressure’ of
from the experimental distribution. In this way, 19.4 visits per hectare per year. In Table 3 some
from our sample we calculated that each Italian comparisons with data collected in other coun-
goes to the forest for recreation an average of tries are shown (Koch, 1978; USDA, 1990;
4.03 times per year. The 95 per cent confidence Benson, 1991; Jacsman, 1991).
interval for this value was 4.03 1.1.
Protected forest areas receive a recreation and
Comparing these data with statistics concern- tourism pressure (68 events/ha/year) over five
ing other free-time activities (ISTAT, 1993), forest times greater than other forest areas (13
attendance seems to be of secondary importance, events/ha/year). Approximately 70 million forest
covering no more than 2.5 min/day, on average. visits (40 per cent of the total number) concern
This value is lower than 1 per cent of the average natural parks or reserves, which in 1994 covered
individual’s daily free-time (4.7 h) and about 5 7 per cent of the overall national forest area. Even
per cent of the free-time dedicated exclusively to if these data are considered positive from an edu-
outdoor recreation activities (0.8 h/day).
cational and cultural point of view, they could
It would be wrong, however, to unjustifiably have worrisome implications from a naturalistic
underestimate the social importance of forest as a and ecological point of view, thwarting the main
recreational and tourism-attracting asset. The reasons that some parks were created for.
60
50
–b
N = a ꢁ T
a = 215.933
b = 0.941
40
30
20
10
0
0
50
100
150
200
250
300
Median time (days) passed since the last event
Figure 3. The model carried out using the 43 simulated populations.

178
FORESTRY

FEATURING AND MODELLING FOREST RECREATION IN ITALY
179
People’s travel expenses to reach the forest • measurement of the dependent variable (QEF)
entrance are also worth mentioning, as they
amount to 2350 billion lire/year. This amount
through the calculation of the sample mean
judgement for each stimulus;
refers exclusively to means of transport and does • identification of a set of variables (25 elemen-
not include any other expenses (e.g. for food or
equipment), which cannot be calculated. By way
of comparison, note that the national annual
tary descriptors) featuring each forest pattern
submitted to people’s appraisal and assumed to
be correlated to acceptance;
output of timber in 1993 did not exceed 770 • experts’ estimate of each picture according to
billion lire (ISTAT, 1995).
A more detailed description of the statistical
numeric values (on a standardized 1–5 scale) to
be given to each elementary descriptor;
features of the model can be found in Scrinzi et al. • grouping together the elementary descriptors in
(1995). An English version is available if requested
by e-mail (isafa.asfor@tqs.it).
After the preliminary estimate of the demand
for forest recreation in Italy, which also provided
six independent variables, called operative
descriptors, referring to the main features likely
to affect the aesthetic–functional evaluation
(Table 4);
other information later used in the model, mul- • elaboration of the model for the interpretation
tiple linear regression was chosen as the suitable
elaboration method for estimating the attendance
degree of a given forest site, and some explana-
tory variables were defined, at the same time
defining the dependent variable as the experi-
mental value of transits per hectare and per year
observed in 41 forest sites in Trentino (east alpine
area).
The first explanatory variable taken into con-
sideration was the aesthetic–functional quality
(QEF) of the forest, defined as the whole of the
factors involved in and affecting the level of
attraction that differently featured forest sites
can exert on recreational users (Daniel and
Boster, 1976; Schroeder, 1984). The procedure
for determining this variable (Scrinzi et al.,
1996) can be summarized in the following main
steps:
of the relationship between QEF and operative
descriptors by means of multiple linear regres-
sion analysis (Tables 5 and 6).
Concerning people’s preferences, users seem to
prefer ‘pseudo-natural’ forests. On the other hand,
however, these must be relatively open, not
Table 4: The independent variables of the QEF model
Operative descriptors
STAND
Elementary descriptors
size
slope (inclination of stems)
branching
integrity
distribution uniformity
size variability
density
epiphyte species
chromatism
• arrangement of a photographic archive with 50
forest views (visual stimuli) representing the
variety of Italian forest patterns, from coastal
to alpine forests, at the same time illustrating
as much as possible the differentiation in the
forest parameters (stand, understorey and
ground morphology);
• selection of a people’s sample representative of
the overall Italian population (the same as
involved in the previously described inquiry);
• display of the selected visual stimuli to the
sample individuals and recording of all the aes-
thetic–functional quality judgements for each
stimulus, on a standardized scale from 1
(minimum acceptance) to 10 (maximum
acceptance);
coverage discontinuity
shrubs
tall herbaceous species
herbage
UNDERSTOREY
CUTTINGS
chromatism
dryness/humidity
scattered residuals
collected residuals
scattered products
collected products
stumps
GROUND
slope
roughness
PHOTOGRAPHY
SEASON
background quality
lighting
seasonal suitability

180
FORESTRY
Table 5: The model for estimating QEF: results of the regression analysis
Variable
Coefficients
Standard error
STD coefficients
t-test
Constant
Stand
Understorey
Cuttings
Season
–3.811
0.676
0.856
1.178
0.205
0.886
0.209
0.182
0.144
0.060
0.000
0.256
0.370
0.590
0.248
–4.299
3.237
4.703
8.193
3.430
R2 = 0.780.
Standard error of the estimate = 0.616.
Table 6: The model for estimating QEF: analysis of variance
Variation source
Sum of squares
Degrees of freedom
F-ratio
Estimates of variance
Regression
Residuals
Total
60.343
17.079
77.422
4
45
49
39.748
15.086
0.380
–
–
–
intricate, with good visibility on the ground, low multiplicative structure, such a mean reflects the
density of stand, and even vertical stems of inter- relative value of single variations, appropriately
mediate or large size. Whether the trees are identifying all the situations where the extreme
broadleaved trees or conifers does not seem to values of the descriptors play a major role in the
play any role in affecting preference. Forestry QEF estimate.
interventions have undoubtedly an unpleasant
From a statistical point of view, the absence of
impact (at least when their consequences are error in the values of the independent variables
evident) as do also any natural event (weather- (Xi) is one of the most obvious requirements for
beating or damage by pathogenic agents) affecting the use of linear regression. Nevertheless, an
the stand integrity. Unexpectedly, ground slope experimental check up of the error in the
and roughness do not seem to play a significant variables is often impossible in regression analy-
role in judgements, and this fact would suggest sis. Therefore, its application in a context in
that the aesthetic senses could be prevailing on which independent variables are affected by a
functionality, at least as far as the morphological certain degree of instability (evaluations based on
features of the site are concerned. In Figure 4 the judgements rather than on actual measurements)
three most and the three least liked pictures are was thought of as reliable considering that the
shown.
consequences of unfixed Xi pertain mainly to
Having fixed the criteria for the basic features the testing procedures or, more generally, to
scaling from 1 to 5 (e.g. the degree of acceptance the statistical inferences based on them, affect-
increases with a decrease in low branching, pro- ing less significantly the predictive use of the
gressively increases with tree size, improves with model.
lesser ground slopes, and so on), four forest
Provided that the features correlated to the aes-
experts evaluated each visual stimulus in relation thetic–functional judgements expressed by the
to all the elementary descriptors according to sample pertain to the six components: stand,
their influence on QEF. The mean value of understorey, cuttings, ground, season and pho-
the four evaluations was taken as its ‘actual’ tography, the final equation is:
value. The 25 elementary descriptors were then
grouped into six independent variables through QEF = –3.811 + 0.676 stand + 0.856 understorey
the geometric mean. In fact, due to its
+ 1.178 cuttings + 0.205 season

FEATURING AND MODELLING FOREST RECREATION IN ITALY
181
Figure 4. The three best visual stimuli (top panels) and the three worst visual stimuli (bottom panels)
according to sample judgements. Numbers indicate the mean ratings.
adj R2 = 0.780
(in particular slope) to forest site attendance, it
should be considered separately in the final
An analysis of the structure of the model high- model.
lights the great importance of the descriptor cut-
The second explanatory variable considered is
tings followed by the descriptor understorey: the the Tourist–Recreational Pressure (PTR) on the
contribution of these to the model exceeds even territory surrounding a given forest site. It
the importance given to stand. It is evident, there- depends on structural features (tourism-related
fore, that QEF is not a fixed and unchanging facilities) and from the demographic pressure of
attribute of a forest area, as it can vary remark- the territory to which the forest site belongs
ably with the seasons or because of human inter- (Scrinzi et al., 1997).
ventions (cuttings or other factors).
This variable has been determined by means
The model does not include the variables of a probability model which considered terri-
ground and photography because both the step- tories at the level of province or region, sub-
wise test and the partial parameter significance divided into sub-units called ambits, each one
test F suggested that these operative descriptors covering a small number of municipalities
do not significantly contribute to the evaluation homogeneous with respect to demographic and
of the dependent variable QEF. The insignifi- tourist features. Flows were estimated between
cance of the photography variable shows that the each ambit (the so-called emitter) and each of
people’s sample formulated their aesthetic evalu- the others (the so-called receiver). Each ambit
ation without taking into account the picture’s generates a number of forest recreation events,
quality, but only on the basis of the ‘forest and it is at the same time emitter of events
content’ of sites, as required by the question- towards other ambits (and itself) and receiver of
naire. The main hypothesis about the insignifi- events generated in other ambits.
cance of the descriptor ground is that pictures
For the model arrangement, four types of sta-
may be unable to display the actual ground fea- tistical data have to be considered at ambit level:
tures, such as roughness and slope. Given the territorial, demographic, tourist and forest data.
undoubted contribution of ground components Specific data concern the following:

182
FORESTRY
• total area (ha);
covered within one day to the chosen recreational
• population (resident);
• total forest area (ha);
• high-forest area (ha);
area was fixed at 125 km. Therefore, the spatial
mobility model was achieved by grouping forest
visitors into seven classes of distance (previously
• utilization rate of reception facilities (expressed defined in the questionnaire) covered from their
by the ratio between number of bed-places and domicile or temporary stay resort to the entrance
average number (over several years) of of the chosen forest site. The probability distri-
overnight stays;
bution of events in these seven classes is different
• distance between each ambit and all others and for residents and for tourists (Figure 5). In order
to the neighbouring provinces (road distance to allow the spatial model to have a circular shape
and straight-line distance).
(a necessary condition to simplify calculations)
we had to transform road distances into straight-
The integration of holiday-related national sta- line distances, by means of a regression model
tistics (ISTAT, 1994a,b) with the ISAFA data built on real data between pairs of localities in the
(Scrinzi et al., 1995) allowed us to distinguish the territory.
different habits in forest attendance between
(2) the ambit extent, which in this model acts
tourists and residents. In this way we calculated as the partitioning criterion for the events con-
the index of forest-addressed events per normal cerning ambits equidistant from the emitting one.
day (FENDI) relevant to residents and the corre- The larger the ambit, the higher the number of
sponding index per holiday day (FEHDI) relevant events addressed to it.
to tourists; such indices amount respectively to
(3) the Forest Quality Index (FQI), which
0.00863 and 0.09097. The ratio between these further distributes all events concerning ambits
two indices reveals that the tourists’ tendency equidistant from the emitting one. FQI is the ratio
towards forest-addressed activities is 10.5 times between high-forest extent and total forest extent
greater than residents’.
(high forests + coppices). The comparison
In the arrangement of the model, these data between the FQI of each ambit and the FQI
allow us to calculate the number of yearly events average value for all the ambits belonging to the
generated from each ambit as well as from any
neighbouring province:
residents’ events = number of residents ꢁ
FENDI (0.00863) ꢁ 365
tourists’ events = bed places ꢁ
mean annual utilization rate of bed places ꢁ
FEHDI (0.09097) ꢁ 365
Events are distributed from an emitting ambit
in the forest areas of other receiving ambits
emitting
ambit
center
BELT NUMBER
1
2
6
7
4
5
3
according to three criteria of partition:
(1) the distance between the barycentres of the
emitting and receiving ambits. This plays a major
role in regulating the amount of recreational and
tourist flows between the two localities. A lesser
tendency towards long-distance movements
along with the increase in the number of alterna-
tive areas makes the event probability per surface
unit (related to visitors from a given place)
decrease rapidly and consistently as the distance
increases.
Figure 5. Scheme of the territorial belts with differ-
ent probability of events constructed around each
emitting ambit. Numbers indicate the main par-
ameters featuring each belt.
Looking at the answers in the above described
questionnaire, the maximum road distance

FEATURING AND MODELLING FOREST RECREATION IN ITALY
183
same class of distance (weighted with their SYSTAT (Wilkinson, 1989), starting from the
extent), allows us to adjust the distribution of hypothesised model:
events. Obviously, this redistribution does not
alter their total amount.
ATT = ƒ (QEF, PTR, S, A)
Knowing both the number of total events
The main statistical parameters of the
generated by each ambit and the number of attempted model (Figure 6) show that it does not
excursional events, i.e. those generated in a given work properly. In fact, the coefficients in the
ambit i and addressed towards the forest Pearson correlation matrix are quite low, as also
resources of other ambits, we may calculate local are the adjusted SMR and the F-ratio. We
events (defined as those generated within a given attempted to change the model by adding or
ambit i and addressed towards the forest removing variables, but the results, although
resources of the same ambit) by subtracting the slightly better, did not improve the model signifi-
latter from the former. The PTR on each ambit is cantly.
the sum of its local events plus the excursional
events generated in other ambits and addressed
to it.
Results and discussion
The calculation procedures used in this model
can be almost completely automated by using an The results of this research can be considered
appropriate work-sheet (Scrinzi et al., 1997; from two points of view: the know-how acquired
Floris et al., 1998).
by means of the intermediate steps and the final
Two other explanatory variables have been forecast model.
taken into account: slope and accessibility. The
The first task has been properly completed,
first one was already included in the ground vari- allowing us to estimate the forest recreation
able of the QEF model but it was rejected from demand and defining its main qualitative features.
the stepwise analysis, probably due to the diffi- This inquiry could certainly be repeated in the
culties of evaluating it on pictures away from near future, increasing the sample and improving
both users and experts. Accessibility was defined some aspects of the methodology. Besides, now
as the easiness for the user to reach the site and we have access to a model for QEF evaluation
was represented by the ‘out of path’ distance that can easily be applied at a local scale for
between the last point on trail and the site, management purposes.
expressed in metres.
Moreover, a probability model of the anthropic
The experimental dependent variable necessary pressure on forest resources for recreational pur-
for the hypotheses model, i.e. the ‘actual’ poses was also carried out, which could be a
tourist–recreational attendance expressed in tran- useful tool for decision makers, administrators
sits/ha/year, was surveyed on 41 forest sites in and politicians, informing them on tourist
Trentino (alpine region in north-eastern Italy). A development of forest and wild areas. It can also
sort of auto-census method was carried out, using be used to compare attendance levels on sites rel-
record stations (Lucas, 1985), similar to post evant to different territories with each other.
boxes placed within wood stands and not visible
The second and main purpose of the study
from the path. Users were requested to compile a cannot be considered to have been attained so far.
short and very simple form recording the date, In fact the regression model which attempted to
time of their transit and the number of persons relate the attendance at sites (number of events)
present (Agatea, 1993). For each site at least six to the previously described explanatory variables
survey periods were considered (each 1 week has not supplied statistically reliable explanations
long), distributed during the period April– of the experimental data. For this reason, and
October of the years 1992 and 1993. For those also considering the contributions of the explana-
sites at lower altitudes, some winter sessions were tory variables, complex and probably non-linear,
also surveyed. All data were referred to per we are currently trying another approach using
hectare and per year.
neural networks to explain the phenomenon from
The general forecast equation has been calcu- a mathematical point of view (Hagan et al., 1996;
lated using the multiple regression tool of Pattie and Haas, 1996).

184
FORESTRY
Figure 6. The multiple regression model to estimate the attendance degree of a given forest site: main
statistical parameters and analysis of variance.
Nevertheless, it seems that the study will con- Acknowledgements
tinue to investigate such topics as:
The research project described here, carried out under
the direction of Gianfranco Scrinzi throughout several
• control of the mensuration modalities of the
years, has so far involved some other researchers, Vit-
experimental data concerning transits in forest
torio Tosi, Paola Agatea, Tiziana Flamminj, Anna
areas (auto-census methods require good co-
operation from users and probably need longer
periods of observation and some kind of check
on the reliability of the collected data);
• the amount of experimental data necessary for
the model (although this kind of data is very
expensive to collect, maybe 40 observations is
too small a number);
• type and number of explanatory variables to
take into account (for example, in a wider
meaning of the variable accessibility more
factors could be considered, such as the dis-
tance by road, car parking availability, pres-
ence of eating and hotel facilities and so on).
Cutrone and Maria Fulvia Zonta. Giuseppe Cappa-
longa gave precious help in translating the paper.
References
Agatea, P.M. 1993 Come prevedere l’attrattiva turistica
di un bosco? Agric. Ric. 147/148, 25–38.
Benson, J.F. 1991 Value of non-priced recreation on the
forestry commission estate in Great Britain. J. World
For. Res. Manage. 6, 49–73.
Daniel, T.C. and Boster, R.S. 1976 Measuring Land-
scape Esthetics: the Scenic Beauty Estimation
Method. USDA Forest Service Research Paper RM-
167, Rocky Mountain Forest and Range Experiment
Station, Fort Collins, CO.

FEATURING AND MODELLING FOREST RECREATION IN ITALY
185
Davis, L.S. and Johnson, K.N. 1987 Forest Manage- MAF-ISAFA 1988 Inventario Forestale Nazionale
ment. 3rd edn. MacGraw-Hill, New York.
Douglass, R.W. 1977 Forest Recreation. 2nd edn.
Pergamon International Library, New York.
Italiano. Sintesi metodologica e risultati. Ministero
dell’Agricoltura e delle Foreste, Istituto Sperimentale
per l’Assestamento Forestale e per l’Alpicoltura,
Trento.
Floris, A., Scrinzi G., Barrel, I. and Nuzzi, M. 1998 La
pressione turistico-ricreativa sulle risorse forestali in Nöel, E. 1992 Aggiornamenti sull’idea di caso. Bollati
Valle d’Aosta. Sherwood 2, 19–25. Boringhieri, Torino.
Hagan, M.T., Demuth, H.B. and Beale, M. 1996 Neural Pattie, D.C. and Haas, G. 1996 Forecasting wilderness
Network Design. PWS Publishing Company, Boston,
MA.
recreation use: neural network vs. regression. AI
Appl. 10(1), 67–74.
ISTAT 1990 Popolazione e movimento anagrafico dei Schreuder, H.T., Tyre, G.L. and James, G.A. 1975
Comuni, anno 1989. Annuario ISTAT n. 2, Roma.
ISTAT 1993 Indegine multiscopo sulle famiglie ’87–
’91. L’uso del tempo libero in Italia, Roma.
ISTAT 1994a Statistiche del turismo, anno 1992.
Annuario ISTAT n.7, Roma.
Instant- and interval-count sampling: two new tech-
niques for estimating recreation use. For. Sci. XXI(1).
Schroeder, H.W. 1984 Environmental perception rating
scales: a case for simple methods of analysis.
Environ. Behav. 16, 573–598.
ISTAT 1994b Indagine sulle vacanze. Le vacanze degli Scrinzi, G., Tosi, V., Agatea, P. and Flamminj, T. 1995
Italiani nel 1993. Notiziario ISTAT Serie 4, Foglio 41
n. 10, Roma.
ISTAT 1995 Statistiche forestali, anno 1993. Annuario
ISTAT n.46, Roma.
Jacsman, J. 1991 Methodische Ansätze zur Ermittlung
der Belastung von Wald und Landschaft durch die
Erholungsnutzung. Schweizeris. Z. Forstwes. 142(2),
85–108.
Karameris, A. 1982 Analyses und Prognose der Erhol-
ungs-nachfrage in Waldern als forstlicher Beitrag zur
Raum-planung. Forstl. Forschungsb. Munchen 50.
Gli Italiani e il bosco. Coordinate quali-quantitative
dell’utenza turistica in Italia. Comunicazioni di
Ricerca ISAFA 95/1, Trento.
Scrinzi, G., Floris, A., Flamminj, T. and Agatea, P. 1996
Un modello di stima della qualità estetico-funzionale
del bosco. Comunicazioni di Ricerca ISAFA 95/2,
Trento.
Scrinzi, G., Flamminj, T., Cutrone, A. and Floris, A.
1997 Un modello di valutazione quantitativa della
pressione turistico ricreativa sulle risorse forestali.
Comunicazioni di Ricerca ISAFA 96/4, Trento.
Koch, N.E. 1978 Forest recreation in Denmark. Part 1: Tosi, V. 1989 I servizi turistico-ricreativi dei boschi:
the use of the country’s forests by the population. esperienze nel Triveneto. Annali ISAFA, X, 101–265.
Statens Forstl. Forsogsvaesen Kobenhavn 35(3), Tosi, V. and Scrinzi, G. 1994 Turismo nel Parco Natu-
289–451.
rale delle Dolomiti di Sesto: un’analisi quali-quanti-
tativa dell’utenza. Comunicazioni di Ricerca ISAFA,
94/1, Trento.
USDA (US Department of Agriculture) 1990 An Analy-
sis of the Outdoor Recreation and Wilderness Situ-
ation in the United States: 1989–2040. General
Technical Report RM-189, Fort Collins, CO.
Koch, N.E. and Jensen, F.S. 1988 Forest recreation in
Denmark. Part IV: The preferences of the population.
Forstl. Forsogsvaes. Dan. 41, 243–516.
Lucas, R.C. 1985 Visitor Characteristics, Attitudes, and
Use Patterns in the Bob Marshall Wilderness
Complex, 1970–1982. USDA Forest Service, Inter-
mountain Research Station, Ogden (UT), Research Wilkinson, L. 1989 SYSTAT: The System for Statistics.
paper INT-345. SYSTAT Inc., Evanston, IL.