Land value assessment based on an integrated Voronoi-geographically weighted regression approach in Makkah, Saudi Arabia

2.1 Conventional method of land value assessment

The conventional method for land value assessment is the on-site visual assessment technique adopted by local government. This method is time consuming, unreliable, and is not effective in terms of transparency and accuracy (Demetriou, 2017). These deficiencies have inimical effects on planning land valuation, as there is need for some research in order to examine land valuation factors and the quality of this traditional process. To address deficiencies in conventional land valuation process, there is need for the development of a new framework based on advanced spatial analysis techniques such as Multiple Regression Analysis (MRA) and Geographically Weighted Regression (GWR) within GIS environment (Demetriou, 2017). Shortcomings in the conventional method of mass appraisal can be solved by the use of Automated Valuation Models (AVMs) with Geographical Information System (GIS) (Demetriou, 2018). The conventional model depends on some key factors such as physical features, locational characteristics and economic conditions. Analytic Hierarchy Process can be used to produce criteria relative weights according to their influence on the discussed problem (Issa et al., 2019).

Land rent theory shows that unimproved land value is a result of the lands inaccessibility to goods, facilities, transportation, medical assistance, education and other services. In theory therefore, public transport provision which increases accessibility to services should in turn increase land values (Mulley et al., 2018). In land consolidation scheme, two components should be considered; these include the reallocation of land and the provision of rural infrastructure such as roads and irrigation networks.

Valuation is based on consumers’ behavior, which means how they want to spend their money based on their individual preference (Hamilton and Morgan, 2010) according to the principle of supply and demand, it also depends on buyers who are in competition with each other to ensure good utility as much as possible (Demetriou, 2017).

Land value assessment is a very important and risky process around the world, as it involves many aspects of socio-economic life in both the developing and developed countries, this is because it affects the value of real estate property each year which revolves around 1,000 billion euros (Weber, 2004). In land value assessment process, knowledge of neighborhood land and buildings plays a significant role in determining the value of a specific land. Due to incorrect neighborhood breakdown and incorrect land value accredited to neighborhood, there might be bias in the assessment of land value (Ruddock, 2014).

Although the drawbacks of standard valuation process have been pinpointed by some researchers (Jahanshiri et al., 2011), in the case of land consolidation, there are little or no research on land valuation (Yomralioglu et al., 2007) hence, there is need for relevant assessment studies which is focused on land valuation based on the spatial statistical analysis of a real full-scale project (Demetriou, 2017).

2.2 GIS and Hedonic methods

Hedonic model for property pricing was introduced by Rosen in 1974. Since it is based on the actual real behavior rather than intended one, it is considered as the revealed preference method. It does not require previous judgments, comparative prices, or income information as in the conventional methods, it needs some reasonable amount of pricing data in order to determine the value of property (Aladwan and Ahamad, 2019). Generally, the common procedure of hedonic model started with the selection of variables, which is usually a dependent variable and one or more independent variables, this is followed by applying statistical regression method to study property multiple parameters effects on its final price, or in other words the effects of each parameter on the final price, and finally, building the model which is needed to fit the sample data for further value estimations (Li et al., 2013).

There are two main hedonic methods used for regression analysis and estimating the relationships between variables namely: The Ordinary Least Square (OLS) and Geographic Weighted Regression (GWR). OLS is the dominant method in the implementation of regression analysis (Cebula, 2010; Kong et al., 2007; Lozano-gracia and Anselin, 2012; Ottensmann et al., 2008). OLS is considered as a global method that produces constant coefficient over the whole study area and does not take into account the autocorrelation and geospatial effects while, GWR is a local regression method enhances variation of coefficients in a given study area (Bujanda and Fullerton, 2017; Kinabalu et al., 2015; Radoslaw Cellmer, 2011). GWR has the potential of being implemented fully in a GIS environment (Lu et al., 2014). Wang and Li (2020) provide a systematic review of mass appraisal models for nearly two decades and identify a 3I-trend, namely AI-based, GIS-based and MIX-based models (Wang et al., 2020).

Different functional specifications of hedonic equations were found in literatures. The simplest mathematical linear regression was employed by (Kong et al., 2007; Lozano-gracia and Anselin, 2012; Bujanda and Fullerton, 2017, and Zhang et al., 2015). The semi log regression was used by (Cebula, 2010; Kong et al., 2007; Ottensmann et al., 2008; Lehner, 2011, and Randeniya et al., 2017). Alternatively, studies by (Lehner, 2011), and (Paper and Paper, 2013) used the log-log regression. Among these methods, the semi log regression is commonly used in building the computerized hedonic model because of its simplicity in interpreting results, as well as its ability to accept categorical variables (Cebula, 2010; Kong et al., 2007; Ottensmann et al., 2008; Lehner, 2011, and Randeniya et al., 2017).

Some researchers have established the use of Hedonic price method for pricing houses. Zhang et al. (2004), Armstrong and Guez (2006) examined the relationship between different traffic convenient degree and house price (Armstrong and Guez, 2006; Zhang et al., 2004) while, (Wang and Zheng, 2007) analyzed the degrees at which urban public service accessibility influences housing price (Wang and Zheng, 2007). Visser et al. (2008) employed Hedonic price modelling to derive different models of pricing property from which the contribution of the characteristics of the residential environment were estimated (Visser et al., 2008). Meanwhile, city economics researchers mainly focused on the application of Hedonic method to study the external environment influence in pricing houses (Liu et al., 2010). Also, GIS spatial interpolation method was used to get spatial distribution map and contour map of housing price (Mei and Li, 2008). Cellmer et al., (2019) developed an application of Hierarchical Spatial Autoregressive Models for Land Value Maps in Urbanized Areas. (Cellmer et al., 2019). According to

The use of GIS along with the Hedonic approach appeared in work presented by (Schaerer et al., 2008). They employed the Zurich and Geneva Geographic Information System in valuing natural land uses in urban areas of Geneva and Zurich (Schaerer et al., 2008). In the study carried out by Cavailhès et al. (2009), hedonic price models was combined with a GIS-based geographic model to evaluate the price of landscapes of houses in the urban fringe of Dijon, France (Cavailhès et al., 2009). Li et al. (2013) used GIS and Hedonic in the modelling of spatial variation of housing price in Xiamen city, China. They discovered that location characteristic, neighborhood characteristic and building characteristics had significant influence on housing price (Li et al., 2013). Korea et al. (2015) presented a framework that combines Geographic information System (GIS) with Hedonic pricing method in order to improve analysis and determining the value of green spaces with price of houses in Subang Jaya, Malaysia (Korea et al., 2015).

The use of hedonic approaches with GIS has been proved effective for land value assessment. More specifically, the use of the GWR method in a GIS environment enhances the consideration of spatial variations of phenomena that influence prices of land. The GWR method provides only a discrete and local regression equation that corresponds to each input data. However, in order to estimate the price of land in an area of interest, there is need to find a method to apply alongside with local regression equations continuously in the whole area. The selection of GWR method compared with other common methods is due to the advantage of GWR in the perception of the spatial variability of the independent variables.

3.1 Recall on GWR and Voronoï Diagrams

• Geographically Weighted Regression (GWR)

Spatial nonstationary is a condition where a basic 'worldwide“ model cannot clarify the connections between certain arrangements of factors. The idea of the model must change over space to mirror the structure contained in the information. Geographically weighted regression attempts to capture this variety by aligning a numerous regression model which permits various connections to exist at various spaces focused (Brunsdon et al., 1996). Geographically weighted regression (GWR) has been broadly used in comprehending the spatially varied connections between socio-economic results and policy investment (Wang and Chen, 2017).

The utilization of Geographically Weighted Regression (GWR) philosophy with prices of land as the dependent variable is clarified by spatially characterized independent variables used in determining the value of land.

Based on geographically weighted regression, parameter evaluation can be made for all the geographical data input characterised by its coordinates, independent variables and dependent variable as shown in the formula below.

With the GWR method, every observation is treated as an individual input at a particular geographic location as required in spatial analysis (Du and Mulley, 2012). A key advantage of GWR modeling is the perception of the spatial variability of the independent variables. Indeed, ordinary regression models (such as Ordinary Least Square) do not take the spatial variability of independent variables into consideration, which may lead to ineffective models (Densmore and Mulley, 2012).

• Voronoi Diagram:

Voronoi diagram is a data structure which portrays spatial contiguous relations, it was proposed by Russian mathematician Georgy Feodosevich Voronoi in 1908 (Zhang and Li, 2009).

Voronoi diagram (Fig. 1) is originally defined for a set of sites which divides a plane surface into a number of regional planes where each site corresponds to one region (Yaagoubi et al., 2015). These sites may be points, lines or polygons, with (or without) a corresponding weight (Yaagoubi et al., 2015). Voronoi diagram represents the geometric spatial arrangement, the unpredictable spatial distribution of socio-economic variables (for example focal city, rancher's market, stockroom, street limit, Land centroid) may be considered as sites distribution inside Voronoi diagram (Zhang and Li, 2009). Each point on a Voronoi edge is equidistant from the two sites whose corresponding regions share the same boundary. Also, each vertex is equidistant from the nearest three sites that generate Voronoi regions.

Close

Fig. 1

Illustration of A Voronoi Diagram (Yaagoubi et al., 2015)

Export to PPT

3.2 Factors influencing the value of land in Makkah

There are several factors that may influence the value of land in Makkah. These factors are classified into socio-economic factors, land policy factors and natural factors. Based on an exhaustive literature review and interviews with experts in land valuation in Makkah, a list of 28 factors were determined. Then, a questionnaire was designed in order to choose the most critical criteria based on the experience of 30 participants (professionals working in the area of Land Valuation in Makkah). The analysis of the questionnaire highlighted 20 factors that appear relevant for land valuation in Makkah as shown in Table 1. A value of 5 corresponds to a strong positive effect on land value, a value of 0 is neutral, a value of −5 corresponds to a strong negative effect. After that, an OLS analysis was carried out based on these 20 factors as independent variables and land price for each input data (11 locations with the corresponding land price) as the dependent variable. OLS is considered as global method that produces constant coefficient over a study area and does not consider the problems of autocorrelation and spatial effects. Through OLS analysis, a detailed report can be gotten on the strength of each factor and the possibility of it being a main factor. Also, all factors that have similar influence are to be considered. Based on the OLS analysis, five factors were considered as statistically significant, these are: Distance from the Holy Mosque, Distance from Masha’er Mina, Distance from sewage network, Distance from main roads, and Distance from shopping centers.

3.3 Implementation of the Voronoi-GWR approach

The importance of the GWR method was presented in section III.a (Recall on GWR and Voronoï Diagrams). This method takes into accounts the spatial distribution of input data, where each input will have its own regression equation. In order to apply these equations for land value assessment over the entire study area, the area should be subdivided based on Voronoï regions defined by each input data (Fig. 2). The regression equation corresponding to each input data will be applied to the corresponding Voronoï region.

Close

Fig. 2

Voronoï regions generated from land value input data in the study area.

Export to PPT

Based on the OLS analysis that was used to determine the main factors influencing the price of land in Makkah section 3.2 (Factors influencing the value of land in Makkah), the GWR method was applied on all input data considering only the selected criteria, these include:

• Criterion 1: Distance from the Holy Mosque.

• Criterion 2: Distance from Masha’er Mina.

• Criterion 3: Distance from sewage network.

• Criterion 4: Distance from main roads.

• Criterion 5: Distance from shopping centers.

The information in Table 2 summarized the intercept, residual, and the weight of the five criteria for each land value input data.

The equation calculated by the GWR method for each land value input data was applied to the corresponding Voronoi region as shown in Fig. 3.

Close

Fig. 3

Land Value assessment within a Voronoï region.

Export to PPT

After estimating the value of land in each region of Voronoï, these regions were aggregated in order to obtain a global estimate of the price of land in Saudi Ryals (SAR) per meter square over the entire study area as shown in Fig. 4.

Close

Fig. 4

Land Value assessment in Saudi Riyal (SAR) per meter square within the study area.

Export to PPT