Study area and characteristics of the livestock farms and tourism sector

An extensive description of the study area and study of livestock farms can be found in Calafat et al. (2015) and Gallego et al. (2019). The VC has a total of 542 municipalities spread across the three provinces; all of the municipalities of the VC have been included in the study.

 Otherwise, rural tourism, considered as tourist activity carried out in rural areas, is made up of the integrated leisure offer, characterized by contact with the natural environment. It also facilitates interaction with the local society through which it is managed (Generalitat Valenciana, 2006). In the VC, tourism in 2017 represented 13.2% of GDP and 14.4% of regional employment (Exceltur, 2018); the rural environment presents a great diversity of spaces rich in landscapes (22 natural parks, 4 marine reserves, 8 protected landscapes, 71 municipal natural parks and one natural monument (BOE, 1995), and potential for the practice of tourist activities.

 In the VC the influence of tourism depends on the location of the territory. Therefore, mass tourism concentrates in coastal areas (beach tourism) and have low livestock activity. In contrast, the interior areas of the study area have high livestock activity and tourism is based on the offer of rural houses and hostels in small municipalities (it is not mass tourism) where there are no hotels as in the coastal area. Therefore, this study considers only rural tourism in rural areas, which are the areas with livestock activity that do not contain hotels, which are linked to coastal areas with mass tourism where livestock activity is practically non-existent.

 Throughout the VC, rural tourism is concentrated in the province of Castellón, which has the largest number of rural houses and hostels. Valencia and Alicante, on the other hand, have seen the greatest increase in supply in the last 20 years (Table 1).

 The policies that mainly regulate the evolution of rural tourism in the VC come from three different areas, developed through the LEADER initiative (of the European Commission), the PRODER Rural Development Policy (of the Spanish Government) and aid from the Regional Administration, such as aid for the rehabilitation of rural housing for tourist accommodation. Two VC rural development programs co-financed by the European Agricultural Fund for Rural Development (EAFRD) have been developed, one for the period of 2007-2013 and the other for 2014-2020 (EC, 2015, 2017). In the last report, which presents the evolution of the areas included in the LEADER Programs, it is observed that practically all the areas included in the first period continue to be included in the second. Only a few larger municipalities, or those with positive demographic trends, leave the objective of the new LEADER Axis. This indicates that, despite the successive periods of public intervention (LEADER, LEADER II, LEADER PLUS and PRODER), the most depopulated territories of this region persist, even today, in this situation of demographic and economic crisis. For this reason, it is highlighted that the new development strategies to be defined should include proposals with greater innovative effort and the search for new approaches that allow for qualitative progress with respect to the starting situation and that allow for an increase in awareness of the importance of tourism together with a greater business dynamic.

 

 

Table 1. Evolution of the percentages of places in rural houses and hostels in the Valencian Community (VC)

(1) VC is divided administratively into three provinces: Castellón, Valencia and Alicante

 

Atmospheric dispersion models

 During the last few decades, research has been carried out to model the concentration of odours from livestock farms. Initially, only the different chemical components present in the odours were considered (Zahn et al., 2001; Janes et al., 2004), then environmental conditions such as wind speed, atmospheric stability classes and distance from the odour source were taken into account (Janicke et al., 2012; Lucernoni et al., 2016; Piringer et al., 2016; Guffanti et al., 2018; Oettl et al., 2018).

 One of the most relevant studies on odour modelling in livestock farms is the study by Hayes et al. (2006), which illustrates a dispersion model to determine the odour impact of intensive poultry production units and the evaluation of separation distances for new facilities. Some European countries (Austria, Germany, Switzerland, Netherlands, etc.) and some North American states or cities (Ontario, Illinois, Purdue, Iowa, etc.) have developed guidelines for the development of dispersion models for the calculation of separation distances. The Austrian model considers the most factors (Guo et al., 2004), including the number of animals, species, housing systems, ventilation systems, manure handling, feeding methods, land use and topography (Janicke et al., 2012; Schauberger et al., 2012b, 2013, 2014; Brancher et al., 2016). Since 1970, the U.S. Environmental Protection Agency (EPA) has developed a series of regulatory programs for the modelling of air quality. It groups air quality (or dispersion) models into different categories: Gaussian (non-reactive pollutants), numerical (if the source is of an urban type with reactive pollutants), statistical (if the chemical and physical processes do not have a clear scientific interpretation or a reliable database is not available), box (if pollutants emitted into the atmosphere are uniformly mixed in a finite volume), or physical (if it involves the use of tunnels, waterways or other means to model fluids).

 The choice of the most appropriate model for a given application should be assessed on a case-by-case basis and considering several factors (Capelli et al., 2013). Stationary state models (Gaussian models) are applied when the aim of the study is to look for the most unfavourable condition. Models that are more sophisticated include parameterizations that are more complex and with a greater number of meteorological variables.

 In this study, the Gaussian model was applied to determine the most unfavourable situation. For this purpose, simulations of concentrations of atmospheric pollutants were generated both in the location of farms and in the surrounding area, generating an odour concentration point matrix.

 

Methodology for calculating farm odour nuisance

 The workflow followed for the calculation of odour nuisance of the VC livestock farms can be followed in Fig. 1.

 Data input and general procedure

 The impact of odour emission rates in the immediate surrounding of the farm depends on the magnitude of the emission rate of the facility, the proximity of sensitive receptors, the local topography and the prevailing meteorological conditions. To take into account all these aspects and determine the odour emission values generated by the activity, it was necessary to apply mathematical odour dispersion simulation models (Pagans et al., 2010).

 STEP 1: DATA INPUTS 

Previous work georeferenced the livestock farms of the VC (Gallego et al., 2019), allowing us to know the location of each, with the necessary precision. For geographic referencing, the ETRS89 (European Terrestrial Reference System 1989) was adopted (BOE, 2007).

 STEP 2: EMISSION AND SPATIAL ODOUR DENSITY OF LIVESTOCK FARMS

 First, the odour emission rate derived from each livestock farm has been calculated, using the emission factors by species and production orientation applicable to livestock farms in the VC, validated in Generalitat Valenciana (2008) (Table 2). Considered emission factors were obtained using samples analysed by dynamic olfactometry from different livestock facilities, which have provided an emission rate per animal unit. In this guide, the values that represent the reality of VC livestock and that are adapted to their production conditions were selected.

 The emission rate of each farm was obtained by mul-tiplying the odour emission rate, established for each type of animal, by the total number of animals of the same type in the facility. The normalized values, per unit area, were obtained by interpolation of the sam-ple data from the spatial density of the odour emission.

 

Figure 1. Workflow to obtain odour nuisance and its relationship with the evolution of rural tourism at municipal level.

 

Table 2. Odour emission factors by species and production orientation.

(1) FE Head: factor emission per animal head, measured in odour units per second(OU/s). Source: Generalitat Valenciana (2008).

 

 

 

 STEP 3: METEOROLOGICAL DATA

The meteorological information used was based on data of 46 weather stations over the last four years, provided by the State Meteorological Agency (AEMET) of the Spanish Ministry of Agriculture.

 The study was carried out on the dates of higher tempe-ratures, as the temperatures are the climatological factors that most influence the dispersion and concentration of odours and adopt the most unfavourable scenario possible. Specifically, the meteorological data of the summer mon-ths of the past four years has been used in the simulation. Data was used daily at 7, 13 and 18 hours regarding: wind direction, temperature, wind speed, low clouds and high clouds.

 For the characterization of different meteorological data, the mode for wind direction and the arithmetic mean for wind speed and temperature were calculated from the daily data. Fig. 2 shows an example of the wind roses obtained for each of the main directions used from one of the weather stations, which corresponds to the examples of concentration and dispersion shown in Figs. 3 and 4.

 

Figure 2. Example of wind rose (% of wind direction frequencies) from the same weather station used in Figures 3 and 4: (a) wind rose with directions of total hours used (%); b) rose of the winds with the directions in function of the speed of the total of hours used; c), d) and e) rose winds of the directions for 7:00, 13:00 and 18:00 hours respectively.

 

 STEP 4: CALCULATION OF ODOUR CONCENTRATION. GAUSSIAN METHOD 

The Gaussian dispersion model is one of the most widely used models to describe the movement of pollutants in the atmosphere (Schauberger et al., 2012a; Sommer-Quabach et al., 2014; Piringer et al., 2016).

 The Gaussian atmospheric dispersion approach was used to assess the impact of the smell of the livestock facilities, in particular the model of Pasquill, obtained from the universal equation of dispersion turbulence and convective transport. Pasquill (1961) proposed a simple method of estimating atmospheric dispersion from continuous point sources, expressed in the equation:

 where C is the odour concentration (OU•m^-3) at the receptor location; Q is the odour emission (OU•s^-1); U is the horizontal wind velocity (m s^-1);and 𝜎_y  and 𝜎_z are the atmospheric dispersion coefficients (m). The Gaussian distribution determines the odour plume size on the leeward side, and this depends on the atmosphere stability and its own dispersion in transverse direction.

 It is acceptable to assume some simplifying hypotheses when the emission is on a small scale (up to a few kilometres): development of the model for a stationary state; the diffusion of mass is negligible in the direction of the x axis; the wind speed U is considered constant because the variations in the three coordinate axes are very small and can be looked over; the point source is located at x = 0 and the effective height of chimney H.

 Reduced expression of the Gaussian dispersion model corresponds to the level of the ground:

 where H is the effective emission height.

 Dispersion parameters are determined once the class of turbulence of the atmosphere is defined. This turbulence class defines the ability of the atmosphere to dilute more or less pollution that is being generated by a point source at a given time. The Pasquill stability classes (Table 3) def ine six stability levels (Classes A-C, D and E-F represent unstable, neutral and stable conditions, respectively). The more unstable the atmosphere, the greater the dilution. Stability classes are defined for different meteorological situations, characterised by wind speed and solar radiation (during the day) and cloud cover during the night. Once the stability class is defined the parameters of dispersion are determined, which also depend on the distance to the source.

 In the present study, the effective emission height is the source height, without considering any over elevation of the odour plume. The atmospheric dispersion coefficients were determined using Eqs. 3 and 4 (Ubeda et al., 2010), depending on atmospheric stability classes (A-F) (Table 4):

The methodology used in this study has been validated with field data in the case of a single source, in the same study area, the study of Ubeda et al. (2010).

 This way, the odour concentration matrix has been obtained around each farm (Fig. 3a).

Modelling was performed using meteorological data from the nearest meteorological station. Total odour con-centration around the farm was calculated as a sum of the concentrations of the resulting odour in each direction of the wind by spreadsheet software. 

 

Figure 3. Example of GIS representation of odour concentration and dispersion: a) odour concentration points grid; b) odour dispersion

 

 

Figure 4. Farm odour emissions (OU/s) and odour emission density (OU/km2): a) map of odour emissions by farm; b) map of odour emission density

 

Table 3. Pasquill categories of atmospheric stability.

(1) Categories: A, extremely unstable; B, moderately unstable; C, slightly unstable; D, neutral conditions; E, slightly stable; F, moderately stable. Source: Pasquill (1961). (2)Cloudy conditions refer to the extent of sky covered by clouds and is expressed in eighths. It is measured by observation, grouping all the clouds, and counting how many eighths (x/8) of the sky they occupy.

 

Table 4. Parameters used to determine atmospheric dispersion coefficients.

 

 

 STEP 5: MODELLING ODOUR DISPERSION IN GIS

The odour concentration matrix was implemented in GIS as a point layer (Fig. 3a). A continuous surface of the odour dispersion model of each farm was obtained by interpolation (Fig. 3b).

 STEP 6: MUNICIPALITIES' PERCEPTION OF ODOUR NUISANC

 To deduce that farms affect municipalities due to odours, the following odour concentration thresholds have been considered according to the population’s perception, according to standard (UNE, 2004):

-Detection threshold: The minimum concentration of odour that can be detected by 50% of the population. The detection threshold is 1 OU/m³.

-Recognition threshold: The minimum concentration of the odour at which 50% of the population can describe the odour. A recognition threshold accepted by the scientific community is defined as 3 OU/m³.

-Nuisance threshold: The concentration at which a small part of the population (<5%) manifests nuisance for at least 2% of the time. The most commonly used is 5 OU/m³.

 Odour is classifiable above 5 OU/m³ i.e., it can be identified and can be a source of complaint among the population. Odours are clearly recognizable when the concentration reaches 10 OU/m³, and complaints can also be received.

To analyse this situation in the study area, an overlap was made between the odour dispersion layers and the layer of urban centres, which allows us to know how many farms can affect each urban centre and the maximum odour density value with which it can be affected. In addition, it is possible to know how many municipalities can be affected by each farm, and for each urban centre, how many farms can affect it and, therefore, the maximum concentration of odour that can be perceived.

 STEP 7: RELATIONSHIP BETWEEN ODOUR NUISANCE AND THE EVOLUTION OF RURAL TOURISM AT THE MUNICIPAL LEVEL.

 The municipal evolution in VC of rural tourism was based on the variation in the number of places in rural houses and hostels in the period from 2006 to 2017, a period in which the LEADER and PRODER programs were applied.

 This analysis is performed in order to assess whether tourism in rural areas is increasing or, on the contrary, decreasing. In addition, the study examines if there is an influence between the evolution of rural tourism between municipalities near each other, i.e. if the increase of the rural tourism in a municipality spread to the nearest municipalities; it was carried out by means of an analysis of spatial autocorrelation (Anselin et al., 2006). The study shows if the evolution of rural tourism in a municipality has an overflow effect in neighbouring (showing a positive spatial autocorrelation), favouring the concentration in a given geographical area. Otherwise, if the growth of rural tourism is concentrated on a municipality at the detriment of neighbours’ municipalities, it produces the spatial hierarchy phenomenon (negative spatial autocorrelation).

 The spatial correlation (or dependence) is calculated from the spatial weights’ matrix (also called contact matrix or spatial proximity matrix), which represents the influence relationships of some observations and indicates which spatial units are neighbouring and which are not. It is symbolized with W, it is a matrix square N×N (where N is the number of spatial units, in this case N being the number of municipalities in the VC), not stochastic, whose elements wij reflect the intensity of the interdependence between each pair of municipalities i,j (Moreno & Vayá, 2000). If wij = 1 the two municipalities are contiguous and if wij = 0 otherwise. The elements of the diagonal are 0, since no municipality can be its own neighbour. Subsequently the matrix is normalised in rows. There are different contiguity criterions in practice (Rook criterion if there is a common border; Bishop Criterion if there is a common vertex; Queen Criterion if there is either a common border or vertex). Queen’s Criterion was used as a contiguity criterion in this study, due to make up for spatial contiguity by incorporating both the rook and bishop relationships into a single measure (Anselin et al., 2006). Table 5 presents a hypothetical example of a special weight matrix with 5 municipalities, as well as a representation of the Queen's contiguity criteria. The spatial weights matrix allows generate an array of space delay of the explanatory variable (WY), from multiplying the W matrix by a vector of variables, in N×1 order.

 Moran's Index values range from -1 to 1, where -1 means a negative spatial autocorrelation and 1 means a positive spatial autocorrelation. A value of zero indicates a random spatial pattern. The software package used for this index was GeoDa 1.12, developed by Anselin.

 Moran’s Index measures spatial autocorrelation using the following equation:

where N is the number of municipalities, Xi is the number of places in rural houses or hostels in i, Xj is the number of places in rural houses or hostels in j, x^- is the mean number of places in rural houses and hostels and Wij is a matrix of spatial weights.

 Local Spatial Autocorrelation (LISA) indicates which municipalities have greater weight in index global Moran's Index and which ones are not. LISA map represents the local significant values (Anselin, 1995; Chasco & Fernández, 2008), and classifies the spatial clusters such with positive autocorrelation (high-high and low-low) and negative autocorrelation (high-low and low-high). A positive autocorrelation cluster indicates that the development of rural tourism has been the same in all the municipalities that are neighbours (high-high or low-low), instead, a negative correlation indicates that if in a municipality the rural tourism has increased, the opposite will have happened in the neighbouring municipalities (high-low).

Subsequently, the bivariate spatial correlation was analysed. In this case, it was examined whether there is autocorrelation among rural tourism variations and odours from the livestock farms. This correlation was held using the concept of global and local spatial bivariate correlation. The number of livestock farms in a municipality according to the emission of odours (OU/m³) was considered as the explanatory variable in this study and the spatial delay matrix was calculated on the variation of the rural houses in each municipality (WY). In this case it is the values of the global moral bivariate I range from -1 to 1, where -1 means that the evolution of the number of rural houses and hostels and the number of livestock farms in a range of odour evolve at opposite rates (high-low indicates that the houses and hostels have increased in an municipality and in neighbouring municipalities there has been a decrease in the number of farms in a range of odour; low-high indicates that there has been a decrease of rural houses and hostels in an municipality and the number of farms in a range of odour have increased in neighbouring municipalities). A spatial autocorrelation equal to 1 means a positive spatial autocorrelation and the evolution of the number of rural houses and hostels and the number of livestock farms in a range of odour evolve in the same sense.