Spatial analysis of environmental impacts of highway projects with special emphasis on mountainous area: an overview

Abstract

Mountains are progressively being invaded by highways for development and defence purposes. Environmental impact assessment (EIA) of highway projects in mountainous areas creates a challenging environment for data collection and impact prediction. Geographic information systems (GIS)-based EIA, using appropriate spatial analysis methods can sufficiently reduce the challenges created by mountain environments. The present study includes a review of articles, including research papers, government reports and EIA reports cutting across the inter-disciplinary nature of the topic. The paper identifies the spatial analysis methods, models and salient features of conducting GIS-based EIA of highway projects in mountainous areas. It is observed that spatial analysis of impacts of highway projects on the environmental attributes, especially air, noise, water and socio-economy in mountainous areas are largely unexplored.

Keywords:

• Environmental impact assessment
• geographic information system
• spatial analysis
• highway
• mountain
• multi-criteria decision-making

1. Introduction

Highways are essential for the development and security of a region. They bring prosperity in the form of continuous supply of goods and services, better transport facilities, economic development and better presence of military, especially near international border areas. However, from the perspective of the environment, broadening of highways will create environmental pollution due to increased traffic on road, landuse change, air and noise pollution, socio-economic changes and loss of biodiversity. Understanding of these environmental impacts is pivotal for unbiased and environmentally appropriate decision-making by the governmental agencies on the viability of such development projects. Environmental impact assessment (EIA) is the process of assessing the impacts of such development projects on the environment. However, conventional EIA is time-consuming, expensive and sometimes suffers subjective bias in assessment of the impacts of a project on the environment (Glasson et al. 2005; Takyl 2012). Geographic information systems (GIS) overcome the limitation of conventional EIA and provide an unbiased and easily interpretable EIA. GIS is a computer-based system for capturing, storing, querying, analysing and displaying geographically referred data (Chang 2008). For sound management of environmentally sensitive development projects like construction and broadening of highways in mountainous areas, it is essential to foresee the long-term impacts of such projects. GIS-based environmental modelling provides prediction and simulation of likely impacts of projects on the environmental attributes. GIS provide us with spatial information and maps for improved planning and decision-making at all levels and scales in a simple and unbiased manner (Skidmore 2002).

The GIS process involves several stages. Schematically speaking, this includes Data capture such as encoding or entering of real-world data in the digital system. The various data sources for capturing include conventional analogue map sources, reports and publications, aerial and satellite remote sensing, field data sources and existing digital map sources. This is followed by Database management in the form of data encoding and management of data, such as raster data encoding, georeferencing, projection, colour separation, vector data encoding, attribute data encoding, editing and linking of spatial and attribute data, to name a few. Thereafter, methods like database query, geospatial measurements, overlay analysis, network analysis, surface analysis and geostatistical analysis are used on geographically referred data for Spatial analysis. Inferences made from the spatial analysis are presented in the final stage of GIS process as Geovisualization. This is the process of presentation of spatial analysis in the form of maps, charts, reports, map layout and three-dimensional (3D) visualization (Figure 1).

Figure 1. An outline of geographic information process (GIS) and its relation with decision-making process.

A wide range of environmental attributes have been studied under spatial analysis of EIA for highway projects. However, most of these studies have been conducted in non-mountainous areas. Spatial analysis of air quality includes prediction of concentration of traffic-generated air pollutants in the impact area of highway, using computer-based mathematical models. These models base their calculations on inputs like traffic parameters, roadway conditions, meteorological conditions, emission rate and receptor location for the prediction of air pollution (Carpenter & Clemana 1975; Zimmerman & Thompson 1975; Eskridge & Thompson 1982; Benson et al. 1986; Benson 1989; Hertel and Berkowicz, 1989; Kono & Ito 1990; Goyal & Rama Krishna 1999; Carr et al. 2002). A similar method is used by traffic noise prediction models for calculation of noise level at various locations (FHWA 1998; Agarwal 2005; CISMHE 2008; NHAI 2013; Vij & Agarwal 2013). Data output from these models are spatially interpolated. Interpolated surface of air pollutant concentration and noise level is then reclassified according to certain criteria (like landuse classification based on ambient air quality by state environmental policies) to evaluate the impact (Murphy et al. 2006; Kurakula 2007; Janssen et al. 2008; Kumar & Foster 2009; Wang & Kockelman 2009; Tang et al. 2010; Eason 2013; Syafei et al. 2013; Taghizadeh et al. 2013; Halek & Kavousi-rahim 2014; Li et al. 2014). Mathematical models are also widely used in traffic-induced water quality impact assessment. Unfortunately, very limited work on spatial analysis of water pollution due to highway traffic and spatio-temporal change on water quality due to highway project exists (Agarwal 2005; Winkler 2005; Gajendra 2011). Spatial analysis of soil erosion impact assessment involves use of mathematical prediction models. Satellite imageries, digital elevation model (DEM), landuse map and rainfall data are used to generate input data for these models in GIS environment. Landslide susceptibility impact assessment involve a wide range of inputs covering soil properties, drainage network, road network, landuse pattern, vegetation type, hydrological and geological characteristics of the area. Satellite imageries, such as normalized differential vegetation index (NDVI), DEM, historical data of landslide inventory and precipitation are used in GIS framework for making landslide inventory maps. Studies show that climate, landuse and geological factors and area-specific set of conditioning and triggering factors create a unique set of criteria for landslide in an area. Change in landuse and geomorphic characteristics of the area due to highway project is the basis for measuring the impact of road construction or broadening on the landslide susceptibility and soil erosion (Dhakal et al. 2000; Agarwal 2005; Ramakrishnan et al. 2005; Lee 2008; Owusu 2012; Bardi et al. 2014). A range of spatial indicators have been identified for spatial biodiversity impact assessment (BIA) for highway projects. Geneletti (2003, 2004a, 2004b, 2005) has constructed these indicators based on criteria like rarity of the ecosystem, level of disturbance, core area and degree of isolation of ecosystems to name a few. Except for the only biotic indicator, i.e. rarity, spatial BIA mostly depends on abiotic indicators of the ecosystem for analysis. Moreover, due to ease of assessment, spatial BIA uses representative vegetation as a proxy for the ecosystem. As a result, it ignores transitional ecosystem which shares common features of both ecosystems. Such generalizations can lead to false conclusions. Cost-distance analysis, spatial interpolation, overlay analysis and map algebra have also been applied in socio-economic impact assessment in highway projects. However, it is worth mentioning that limited studies have been conducted on socio-economic impact assessment in highway projects (Geurs et al. 2009). GIS-based studies have shown that highway projects have significant impacts on the socio-economic attributes in mountainous areas (Brown 2003).

It is evident from the above that a wide range of physical, biological and socio-economic attributes have to be considered for spatial analysis of environmental impacts of highway projects. Thereby, multi-criteria decision-making (MCDM) process is used for weighing these attributes. Based on location and nature of problem, a wide range of MCDM is used in spatial analysis of EIA. MCDM-based attribute weight in association with overlay analysis is used for ranking of alternative scenarios for highway projects (Saaty 1980; Malczewski 1999; Wang et al. 2006; Dey 2010; Zhu 2011; Zolfani et al. 2011). There is a growing body of literature in MCDM-based GIS studies which are exploring the application of uncertainty and sensitivity analysis of spatial models on EIA (Crosetto & Tarantola 2000, 2001; Lilburne & Tarantola 2009; Feizizadeh et al. 2014). A summary diagram on the possible role of GIS in EIA is illustrated in Figure 2.

Figure 2. Application of spatial analysis and GIS in EIA process.

Most of the studies on spatial analysis of EIA for highway projects are restricted to urban areas with plain terrain. On the contrary, a substantial fraction of ecologically sensitive and culturally unique areas reside in the mountainous regions of the world. The innate complexity and remoteness of these areas lead to a limited and superficial level of EIA. The difficult terrain, unpredictable weather and poor logistics lead to a limited number of sampling points during the preparation of baseline database and impact prediction stages of EIA. GIS-based spatial analysis, like spatial interpolation, helps in generating interpolated surfaces of the environmental attribute using the sample points. Moreover, overlay analysis, GIS-based statistical modelling, process modelling and index modelling bring out the spatial dimension of environmental impacts, which conventional EIA fails to highlight. GIS uses maps, charts, 3D-simulation and spatio-temporal simulation for data interpretation. Thereby, use of GIS in the interpretation of EIA makes it much more understandable for decision-makers and stakeholders, which eventually leads to well-informed decision-making.

The purpose of this paper is twofold: (a) to elaborate the role of GIS and spatial analysis in impact assessment of highway projects; and (b) to discuss the state of art in spatial analysis of impact assessment of highway projects in mountainous areas, especially in India. The rest of the paper unfolds as follows. In section two, method and methodology used for the literature review is briefly discussed. Section three elaborates the role of spatial analysis in roadway projects. The subsections under it discuss the data required, determinants, techniques and models used and gaps observed in the spatial analysis of individual environmental attributes (such as air and water quality) impact assessment. The study shows that, GIS-based spatial analysis can be an effective tool in various stages of transport EIA in mountainous areas, by overcoming the complexity of terrain, weather and remoteness of human habitations.

2. Method and methodology

The review includes research articles from a wide range of journals cutting across the interdisciplinary nature of the topic. It also includes review papers, research thesis, conference reports, and reference books related to EIA and GIS. Also, government evaluation and analysis reports, EIA reports and manuals were referred to.

•	During the review, Boolean search with keywords was used with the following strategy:

•	The literature review was conducted from May 2015 to February 2016. Initially, the search was based on more broad criteria, focusing on GIS-based EIA studies. The objective of this initial search was to identify the GIS tools for EIA. The second stage of search included EIA in mountainous areas for identification of characteristic features of such studies. Based on this information, the necessary depth was achieved to conduct a more objective search. The search strategy was to focus on selection of research articles which met a higher degree of combination of these keywords. The selection criteria for review was based on the degree of coherence of the content of the article with the review title and research objectives under which this literature review has been undertaken.
•	Different combinations of GIS tools such as spatial interpolation method, geostatistical methods, and spatial analysis methods like buffering and overlay analysis were used in combination with other keywords to refine the search. Similar search strategy was used with MCDM methods. During the search, the majority of articles supplied by the search engines were based on an urban backdrop. As a result, the search process grossly ignored the criteria of mountainous areas. To negate this, ‘urban’ keyword was included in the search as ‘without the word’.
•	The various academic databases consulted during the review were Google Scholar, ScienceDirect, JSTOR, Springer Search, Academic.edu, ResearchGate.net, Taylor & Francis Online, Directory of Open Access Journals, and British Council online library. Cross-referencing was also done. Some of the authors were requested to send their copy of research articles for the review. The PhD thesis of Agarwal (2005), work of Geneletti (2003, 2004a, 2004b, 2005), Barber et al. (2011), Antunes et al. (2001), and Saaty (1990) were extensively studied.

3. Spatial analysis of EIA for highway projects in mountainous areas

GIS is used in several stages of EIA. In the pre-processing stage, GIS can act as a spatial database management tool, where data can be imported, exported and stored for query for EIA. Moreover, it can be used for preparing and displaying baseline maps. GIS can contribute in the processing stage of EIA by spatial analysis viz., overlay analysis, buffering, corridor analysis, cumulative effects analysis, and viewshed analysis for impact prediction. It can be used in association with other statistical, mathematical and computer-based simulation sub-models. It can also be integrated with expert system to support MCDM for ranking of project alternatives. In the post-processing stage, GIS acts as an easy and effective visualizing tool for communication of the results of EIA (Eedy 1995; Bachiller & Wood 2001; Yadav & Mishra 2014). Also, technical, aesthetic, social and environmental data have been used in MCDM integrated GIS-based spatial analysis for EIA (Warner & Diab 2002). Data inputs viz., topographic map, vegetation map, landuse/landcover map, satellite images and ground truthing techniques have been used in GIS framework for EIA (Abbas & Ukoje 2009).

GIS has been used in the EIA of various roadway projects. GIS works as a flexible and effective information technology tool in road planning process. It has been used in proposing several road corridors as an improvement measure during road planning stage (Moskvitina 1999). Spatial impact assessment methodology (SIAM), a GIS-based EIA model, had been used for the central Portugal highway. The model uses overlay analysis and indices of air pollution, water resources and biological resources to demonstrate the capabilities of SIAM in EIA (Antunes et al. 2001). Agarwal (2005) provides a GIS-based methodology to prepare EIA. He emphasizes that most of the environmental attributes are spatial in nature and justify the inclusion of GIS as a suitable spatial database system in EIA process. An integrated approach to EIA of transportation projects has been used in the state of Florida using remote sensing technology, GIS and spatial modelling to present a comprehensive analysis of vulnerability of the environment near the road areas (El-Gafy 2005). Alagan (2007) explores the potential of participatory GIS (PGIS) in EIA, emphasizing the role of PGIS in informed decision-making, the ability to empower communities in that decision-making and contribute towards democratic environmental decision-making. GIS and analytic hierarchical process (AHP), a type of MCDM, have been used in road network planning in temperate forest mountain environments, by considering slope, soil type, geology, hydrographic network, aspects, tree volume, tree type and elevation maps. The study demonstrates the use of AHP and GIS as an appropriate and suitable method in the forest road network planning (Samani et al. 2010).

3.1. Spatial analysis of air quality impact assessment

Vehicular emissions due to traffic in highways are a major cause of degradation of air quality. Harmful health effects of air pollution are dependent on the type of pollutant source, the strength of the source, and the behaviour of the individual potentially exposed (Litchfield et al. 2010). Geographic factors like downwind proximity to highway, smaller geographic area and high population density facilitate traffic induced air pollution health hazards (Jerrett et al. 2005; Brugge et al. 2007; Song 2008; Svendsen et al. 2012).

For the ease of prediction, various roadway traffic-induced air pollutant dispersal models have been developed. These include CALINE 4 (CAlifornia LINE, source dispersal model – version 4), HIWAY (highway air pollution model), HYROAD (hybrid roadway model), IITLS (Indian Institute of Technology – line source model), ROADWAY (Carpenter & Clemana 1975; Zimmerman & Thompson 1975; Eskridge & Thompson 1982; Benson et al. 1986; Benson 1989; Hertel and Berkowicz, 1989; Kono & Ito 1990; Goyal & Rama Krishna 1999; Carr et al. 2002). Some studies have shown that CALINE 4 has been a poor predictor of air pollutants (Anjaneyulu et al. 2008; Gupta et al. 2011; Sharma et al. 2013). CALINE 4 and IITLS have been widely used in EIA studies in India. IITLS has performed better than CALINE 4 in mountainous terrain in India (CISMHE 2008; NHAI 2013). Factors such as seasons, meteorological conditions, transport characteristics and terrain affect the performance of these models (CISMHE 2008; Gupta et al. 2011; Ilic et al. 2014).

These line source air dispersal models have been widely used in conjuncture with GIS for spatial analysis and visualization of spatial and temporal impacts of air pollution in the project affected area (Agarwal 2005; Matejicek 2005; Wang et al. 2008; Becker et al. 2010; Ilic et al. 2014). Integration of GIS with models substantially reduces the data preparation time (Becker et al. 2010). Data inputs like satellite imageries, ambient air quality and meteorological data from monitoring stations, demographic data, roadway maps and road traffic emission data have been used in GIS to generate air quality maps in roadway project areas (Cook et al. 2008; Tang et al. 2010; Superczynski & Christopher 2011; Li et al. 2014). Spatial interpolation methods like ordinary kriging, universal kriging, inverse distance weighing and spline have been widely used to predict the concentration of air pollutants of the study area (Janssen et al. 2008; Kumar & Foster 2009; Wang & Kockelman 2009; Tang et al. 2010; Syafei et al. 2013; Halek & Kavousi-rahim 2014; Li et al. 2014). Thematic maps and DEM have been used to visualize these studies (Chattopadhyay et al. 2010).

Topographical and spatial features have been found to impact the distribution of air pollutants during temperature inversions in mountainous areas (Wallace et al. 2010). GIS-based multiple regression model of state wide spatial distribution of NO₂ concentration has shown that elevation and distance from main highways are main predictors of vehicle-related air pollution (Gonzales et al. 2005). GIS have been used to measure the atmospheric deposition of air pollutants in the mountain area. Landscape features such as forest edges, elevation, aspect and vegetation types were considered in the prediction modelling. High elevation and forest edge type were major accumulators of air pollutants. The study was based on the atmospheric deposition on the forest floor (Weathers et al. 2000). A GIS-based buffering method was used in a long-term and continuous study, on the spatial distribution of air pollutants near roads. The study have shown that elevation and prevailing wind direction are the major determinants of their spatial distribution (Kimbrough et al. 2013). A multifactor assessment model was developed in GIS framework to divide the study area into environmental function zones in Bor, Serbia. A linear programming model was used for optimal control of SO₂ pollution. Study shows that cutting down emissions and ecological restoration are the optimal means of controlling emissions (Illic et al. 2010).

The significance of vegetation cover, meteorological conditions like rainfall and valley effect on the concentration of air pollutants have not been studied in depth. Also, limited evaluation of the performance of air pollutant dispersal models has been done for complex terrain like mountain and forested area. No spatial interpolation method has been identified as a standard technique for prediction of air quality.

3.2. Spatial analysis of noise quality impact assessment

Noise pollution is a major negative impact of highway traffic. The major models used for noise level calculation are CORTN (United Kingdom’s calculation of road traffic noise) and FHWA TNM 2.5 (United States Federal Highway Administration traffic noise model version 2.5). These models are used in association with GIS software, like TNoiseGIS, to generate automatic noise emission values for any number of receptor points and can generate soundscape of the noise-affected area (Pamanikabud & Tansatcha 2001). The FHWA TNM 2.5 model has been widely used for prediction of noise level in the EIA of highway projects (Agarwal 2005; CISMHE 2008; NHAI 2013). It computes noise level based on a series of adjustments to a reference sound level considering traffic flow, distance and shielding effects (FHWA 1998; Vij & Agarwal 2013). The FHWA TNM 2.5 model has been used to show that traffic noise causes depreciation of property value (Cheng et al. 2011). The FHWA TNM 2.5 model and GIS have been used to construct noise maps around highways. The map was reclassified using state noise impact category to evaluate spatial impact of the noise and its impact on the landuse planning (Agarwal 2005). Apart from GIS-independent models like FHWA and CORTN, GIS integrated models like SPreAD-GIS have been developed. SPreAD-GIS is an ArcGIS toolbox developed for traffic noise modelling made especially for wilderness areas (Reed et al. 2010). SPreAD-GIS and CadnaA, a GIS-based model, have been used for road traffic noise modelling in forested areas. Study shows that road traffic is a major pollutant in protected areas. A significant number of wildlife species are affected by traffic noise. The authors observe that validation, calibration and comparison of these models is a major limitation at present (Barber et al. 2009, 2011). Studies on noise quality assessment in mountainous area reveal that traffic increases during dry seasons. Also, noise level increases with increase in relative humidity and uphill movement of traffic (USDI 2012). Moreover, close proximity of the residences to the roadway in mountain areas expose the residents to higher level of traffic noise (CFC 2009). Effective roadless volume (ERV), a spatial measure, uses two parameters: firstly, the shortest distance between a location and the nearest road; and secondly, spatial distribution of traffic noise, to measure landscape penetration by roads and the related impact due to traffic noise. ERV was found to be highest near urban areas and places with high vehicular traffic noise (Nega et al. 2012). Preparation of GIS-based noise analysis involve spatial data collection viz., roadway geometry; satellite and areal imageries for geolocation of vegetation areas, buildings, houses and other noise barriers, and receivers; global positioning system (GPS) readings; DEM; traffic data-like traffic volume, composition, speed and standard emissions at various locations along the roadway (Matejicek & Janour 2011). Spatial analysis of noise assessment includes spatial interpolation for predicting noise level in the unsampled space of the study area. The spatial interpolation methods include kriging, inverse distance weighting, triangulated irregular network (TIN) and nearest neighbour method (Murphy et al. 2006; Kurakula 2007; Eason 2013; Taghizadeh et al. 2013) (Table 1).

Table 1. Description of MCDM methods.

Method	Description
Agent-based process (ABP)	It is used to assess the effect of agents (individuals, organizations or groups) on the system as a whole
Cellular automata	It is a discrete time model consisting of a regular grid of cells which exist in certain state(s). Each cell has a specific relation with its neighbouring cells. Based on certain fixed rules a new generation of cells are formed with altered state
Evidential reasoning approach	It uses a belief structure considering ignorance and randomness in decision-making for problems having both qualitative and quantitative criteria to reach the solution
Fuzzy logic	It is appropriate for problems with imprecise and uncertain data inputs. It utilizes insufficient information and evolution of available knowledge to identify the best alternative
Genetic algorithm	It is a heuristic method which mimics the process of natural selection to select the best alternative
Mathematical programming	It is a broad collection of optimization methods involving maximization or minimization of objective function subject to given constraints, such as linear programming
Monte Carlo simulation	In this process repeated sampling is used to determine the properties of some phenomenon
Simple additive weighting (SAW)	It is a maximization problem of a linear equation subject to given sets of alternatives and criteria to find the best alternative
Tabu search method	A local search method involving potential solution and a set of similar solutions (in relation to potential solution) to improve the solution of the problem. It prohibits or marks ‘Tabu’ to prevent repetitive search of previously visited solutions
The technique for order preference by similarity to ideal solution (TOPSIS)	It finds the best alternative by selecting the shortest distance from ideal solution and farthest distance from the negative ideal solution in an Euclidian space of alternatives and criteria

Source: Wikipedia (2016a), Setinc et al. (2013), Wilson and Gonzales (1985), Triantaphyllou et al. (1998), and Pohekar and Ramachandran 2004.

Like air quality prediction, spatial noise quality impact assessment is mostly confined to urban areas with simple terrain. Also, almost all the models have been developed to address human-noise exposure, grossly ignoring the need for developing such models for ecologically sensitive areas. Fast economic development, like growth in tourism and strategic development, like rapid increase in army establishments can substantially increase in traffic volume in mountainous highways. Further study is needed on how the rapid increase in traffic-induced noise affects the rural community, wildlife and vegetation in terms of spatial and temporal scales.

3.3. Spatial analysis of socio-economic impact assessment

Highway projects cause socio-economic impacts viz., displacement of residents, displacement of business and community services, impacts on the residents like, disrupted and inconvenient detours, local road closures, impact on the travel time, negative health effects, impact on business and community services (Stevenson 1995). Improvement of roadways also leads to improvement in travel times and emergency response time; however, at a cost of displacement of people, alteration of landscape and encroachment of personal property and recreational facilities (MOT 2009). Construction of new roads or broadening of existing roads may have negative impacts on the informal sectors of business and local socio-economic interactions amongst traders (Ajayi et al. 2013). Geurs et al. (2009) observed that socio-economic impact assessment (SIA) has been underexposed in transport-related impact assessments. There is a need for assessment of transport related temporary impacts, health impacts, social cohesion, the distribution and accumulation of impacts across population groups and social justice.

Questionnaire method and direct interview have been used to collect data on employment opportunities, educational facilities, health facilities, demographic profile, per capita income, industrial activities and tourism as attributes in SIA of highway projects. These socio-economic attributes data have been used in GIS for spatial analysis of social impacts (Agarwal 2005). SIA in the mountainous areas of India have been conducted to evaluate the impacts of rural development programmes on the people (APSERI 2000). GPS tagged socio-economic data were integrated in remotely sensed images and GIS-based thematic maps of biophysical attributes in Nepal to prepare information database. The database was used in multiple linear regression in predicting the farm income potential from cost distance from the market and land quality parameters (Krishna Bahadur 2011). Spatial interpolation and cost distance analysis have been used to observe the impact of road network and accessibility to road on the socio-economic condition in mountainous areas. Socio-economic conditions greatly vary with mountain altitude and the relation is inverse in nature. Moreover, improvement in the road condition and accessibility substantially improve socio-economic condition of remote areas (Rudiarto & Handayani 2011). Spatial analysis of gender-related socio-economic attributes, viz. size of landholding, use of agro-chemicals, gender-based workload, were studied in Nepal. The study shows that road is a major determinant in the spatial intensity and distribution of these attributes (Brown 2003). Biophysical and socio-economic parameters have been used to predict the forest cover type using GIS framework. Biophysical parameters like elevation and road network are better predictors of forest cover type than socio-economic parameters. (Gyawali et al. 2008).

SIA of transport-related projects are few and superficial in depth. Also, highway development-linked changes in population density, composition and their consequent impact on the socio-economic attributes like employment opportunities, income, ethnic composition, crime rate, etc. remain unexplored. Moreover, due to the innate topographical challenges, dense vegetation cover and limited arable land of mountain areas create new scenarios for investigation of the socio-economic impact of highway projects.

3.4. Spatial analysis of BIA

Habitat loss and fragmentation are major threats to biodiversity (Gontier et al. 2006). Spatial analysis of BIA for road projects helps in evaluating the expected losses of ecosystem (Geneletti 2003). Spatial indicators like core area of the habitat, isolation and level of disturbance along with expert-based value function have been used in GIS framework to assess the effects of roadway projects on habitat fragmentation (Geneletti 2004a). Landscape criteria viz., rarity, isolation, dimension of forest ecosystem and exposure to disturbance were used as criteria in MCDM method to construct value functions for ranking of roadway project alternatives (Geneletti 2004b). Ecosystem attributes such as naturalness of vegetation, habitat suitability of representative species, land production, landscape and geomorphology along with MCDM were used to value various roadway alternatives based on ranking criteria including environmental perspective, socio-economic perspective and neutral perspective. Sensitivity analysis was done to assess the stability of the obtained ranking (Geneletti 2005). Spatial analysis for zoning of a protected area was based on ecological indicators viz., relevance and rarity of habitat, potential and actual presence of species and outstanding natural features; cultural and heritage indicators like cultural heritage sites, compatible landuses and landscape assets; and recreational and tourism indicators for instance, transportation infrastructure, recreational activities and tourism facilities (Geneletti & Duren 2008). GIS-based BIA for road alternatives have shown that in some cases the proposed road leads to greater ecosystem loss and fragmentation than the existing road (Monavari & Momen Bellah Fard 2010). Buffering method has been used to calculate the impact of roads on habitat fragmentation using two calculation methods viz., cross-boundary connections and cutting-out procedure. GIS was used to calculate the effective mesh size of fragmented area (Girvetz et al. 2008). Questionnaire method has been used to collect data on public perception of impacts on biological attributes due to the highway project. It was used in GIS framework to assess spatial distribution and impact intensity of the project on the biological attributes (Agarwal 2005).

Mountains are fragile ecosystems with a large number of endangered and endemic species of flora and fauna. Construction and broadening of highways is a major threat to these ecosystems. Spatial analysis of such ecological impacts is invaluable for evaluation of the viability of such development projects.

3.5. Spatial analysis of landslide susceptibility impact assessment

Landslide leads to loss of life and property, blocking of roads and rivers, disruption of communication and triggering of flood. Construction of roads increases potential slope instability, it interrupts surface drainage, ditches and culverts, and it alters the subsurface water movement, changes the distribution of mass on a slope surface and facilitates erosion by slides due to road construction-related deforestation. All of these factors facilitate landslide during and after road construction (Swanson & Dyrness 1975). The majority of landslides occur along the roads and on faults. Shallow landslides are more frequent along roads as compared to faults. Landslide areas are more frequent at a distance between 80 and 100 m away from the road edge (Hosseini et al. 2011). Road construction facilitates human settlements in landslide-prone areas. The decision for settlement depends on risk perception, livelihood strategies and road-related development interventions (Lennartz 2013).

A landslide inventory map is prepared by using soil parameters like soil depth, stoniness, hydraulic conductivity, soil drainage behaviour, soil slope, soil type, soil erosion, surface texture, inner texture, lithology; drainage network, road network and landuse pattern (Sharma et al. 2009); as well as DEM and geology of the area (Dhakal et al. 2000). NDVI, slope roughness, tangential curvature, relative and total slope height and wetness index are also used in landslide inventory mapping (Lee 2008). Some of the triggering factors of landslides are heavy rainfall, geology and human activities (Lee 2008; Sharma et al. 2009; Sengupta et al. 2010). Geological conditions viz., monoclinic structure and clayey – silty lithology, and morpho-structure play a significant role in spatial distribution of landslide. Also, landslide area variables such as altitude, length, width, height, slope angle, aspect play important role in landslide prediction.

GIS is a useful tool for landslide inventory mapping. Information theory and regression analysis have been used in GIS framework to prepare landslide hazard zones. Rock type, multiple joint sets, active tectonism and very high annual precipitation act as the conditioning factors while rapid pore pressure build-up, seismic activity and anthropogenic activities are the triggering factors in landslide occurrence (Ramakrishnan et al. 2005). Mapping and monitoring of landslides show that intense and exceptional rainfall is the major triggering factor. Also, steep slopes and silty-clayey sediment cover acts as the conditioning factor (Bardi et al. 2014). GIS-based logistic regression and bivariate statistical analysis have been used to predict the presence or absence of landslide occurrence as a dependent variable against a set of independent variables viz., bed rock–slope relationship, lineaments, slope gradient, aspect, elevation and road network. Road network plays a major role in determining the occurrence and distribution of landslides (Ayalew & Yamagishi 2005). In other studies, artificial neural networking, weight of evidence model, decision tree, support vector machine, adaptive neuro-fuzzy inference system, bivariate statistical analysis, multivariate adaptive regression spline model and multi-variate logistic regression model have been used in association with GIS to prepare landslide susceptibility maps due to highway projects (Wang et al. 2008; Wang et al. 2009; Pradhan, 2010; Pradhan et al. 2012; Krishna & Kumar 2013; Pradhan 2013; Guri et al. 2015). GIS-based analysis of multi-source datasets for universal soil loss equation (USLE) model has shown significant correlation of soil erosion and landslide occurrence (Pradhan et al. 2012). Macroscopic landslide hazard zonation has been done by measuring landslide hazard evaluation factor using DEM. The spatial data used in the mapping are lithology, structure, slope morphology, relative relief, landuse and landcover and hydrogeological conditions (Surendranath et al. 2008). Overlay analysis in GIS has been used in a deterministic model to predict landslide occurrence (Gupta et al. 2009). Selecting the right set of conditioning and triggering factors and MCDM method is crucial in the spatial analysis of highway project-induced landslide susceptibility mapping.

3.6. Spatial analysis of soil quality impact assessment

Soil erosion significantly increases in the post-construction stage of roads. Soil erosion is at its peak in the months of highest rainfall (Beschta 1978). Construction and broadening of roads lead to denudation of soil, exposing it to soil erosion. In fact, forest roads are prone to soil erosion. Lower or no traffic reduces erosion of the road whereas erosion is more pronounced in wetter climates (Elliot et al. 2009). In unpaved forest road conditions, cutslopes show highest soil loss as compared to sidecast fill and roadbed. Gradient, plant cover density and stone cover play significant role in preventing soil erosion. Initially, after construction or disturbance, the sediment load is high, but it tapers off due to exhaustion of loose surface materials (Arnáez et al. 2004). Road properties, viz. road geometry, slope, length, width, surface and maintenance, along with soil properties and vegetation cover, are major determinants of sediment yield. Road surface, cutbanks, fillslopes, bridges and culvert sites and ditches are the main sources of sediment. Increased water flow and exposed soil surface causes highest erosion and sediment yield (Forman & Alexander 1998). Road erosion is intense during the initial years of its construction or disturbance. The sediment yield eventually tapers off. Sediment production is greatest in slopes which are greater in proportion to the square of the slope of the road segment. Increase in vegetation cover prevents sediment loss (Luce & Black 2001).

Soil erosion risk mapping provides a visual aid to identify problematic areas for soil erosion. Fuzzy logic along with USLE model has been used in GIS environment to prepare a soil loss risk map for New Mexico State (Bulut et al. 2012). USLE and revised USLE (RUSLE) models along with GIS have been used to prepare soil erosion risk maps (Agarwal 2005; Owusu 2012). GIS and remote sensing methods have been used in USLE model to estimate spatial and temporal pattern of soil erosion in different road types, viz., truck, county, town, village and unpaved road. Soil erosion decreases with increase in distance from road, except for truck road. Soil erosion is highest in lower roads. High rainfall promotes soil erosion (Zhang et al., 2009). ArcGIS model builder has been used to calculate various inputs for USLE model. The model shows that slope is the most significant factor in soil loss in forested watershed (Csafordi et al. 2012). RUSLE-3D model has been used in GIS environment using high resolution remote sensing data (IKONOS & IRS LISS-IV [Indian Remote Sensing Satellite Linear Imaging Selfscanning Sensor]) to prepare landuse and soil maps. These maps were used to derive soil erodibility factor and vegetation cover. DEM was made for spatial topological factor. Study shows that soil erosion is lowest in very dense forests and highest in open forest in hilly landforms (Kumar & Kushwaha 2013).

The Morgan model has also been used instead of the USLE model for the preparation of soil erosion risk mapping (Ustun 2008). GIS-based study on soil erosion in Himalayan region shows that the Morgan model performed better than the USLE model (Jain et al. 2001). Sediment delivery model, a GIS-based road erosion model and AHP have been used to classify erosion risk. Study shows that re-gravelling of roads, reducing traffic volume, addition of drains and increasing the proportion of outslope surface of the road segment decreases the road sediment delivery (Parsakhoo et al. 2014). Other models like forest road erosion calculation tool, an ArcGIS-based tool, uses the water erosion prediction project model. It uses inputs, viz. DEM, road network map vector data and a series of road definition selection tools, to simulate erosion for cutslopes, road surface and road drainage ditches (Cochrane et al. 2007). Sediment yield from a forest road was modelled using GIS. For this, spatial interpolation using DEM and meteorological data were used for determining precipitation factor. Hillslope gradient layer was prepared from DEM. Also, slope, stream, road, soil and local geology layers served as model inputs. Study shows that model output is sensitive to the quality of input data (Akay et al. 2008). Raster calculator in the ArcGIS spatial analyst toolbox has been used in estimating gully head with and without the road scenarios in the pre-analysis stage of road-induced soil erosion (Addisu et al. 2013)

3.7. Spatial analysis of water quality impact assessment

Highways are essentially a non-point source of water pollution. Construction and after-construction conditions of highways generate pollutants which degrade the water quality and affect the habitat of micro-fauna of the nearby streams (Wu et al. 1998; Chen et al. 2009). Road run-off contains high levels of oil and grease, suspended solids, chlorides and heavy metals. Urban road run-off is more polluted than rural road run-off. Road run-off contains relatively higher concentration of pollutants in comparison to adjacent rivers (USEPA [United States Environmental Protection Agency] 1996; Gan et al. 2008). The major heavy metal pollutants present in road run-off are lead, zinc and copper (Bingham et al. 2002). Factors which affect road run-off are traffic volume, rainfall characteristics (viz. number of dry days preceding rain event, rain intensity and run-off volume), highway pavement type, properties of pollutants and season (USEPA 1996; Cloirec 2000; Aldheimer & Bennerstedt 2003; Yannopoulos et al. 2004; Forsyth et al. 2006; Yannopoulos et al. 2013). Minimizing soil erosion and dust accumulation in the roadways, and consideration of geographic and surface characteristics of the road are best management practices for mitigating road runoff induced water pollution (Abewickrema et al. 2013).

Various models have been implemented for predicting water pollution due to highways. Some of these are the urban watershed model (Schueler 1987), FHWA statistical pollutant loading and impact model approach (Driscoll et al. 1989), simulation models like the USEPA storm water management model SWMM (Huber & Dickinson 1988) and the FHWA urban highway storm drainage model (Dever et al. 1983), regression models like multiple linear regression equation by the California Department of Transport (Caltrans) and the United States Geological Survey (USGS) regression equation for estimating pollution load (Winkler 2005). BASINs (Better Assessment Science Integrating point and Non-point sources), a physical-process-based model developed by USEPA, has been used along with GIS to predict the hydrologic effects of land use in regional and local levels in the state of Ohio. The model performed well in predicting the major sources of water pollution (Tong & Chen 2002). Water quality index method has been used to assess water pollution due to highway traffic. LISS III satellite imageries and GIS were used to develop landcover map and assess the water pollution impacts and its effect on landuse planning (Agarwal 2005). Artificial neural networking and GIS were used to predict water pollution in river basin. Water quality index method and inverse distance weight interpolation were used for prediction of water pollution (Gajendra 2011). Except for sediment load, the role of other water pollutants generated from mountain highways remains largely uninvestigated.

3.8. MCDM process in spatial analysis of EIA

Weighing or ranking of environmental impacts of a development project is a MCDM process. MCDM is a set of procedures which facilitate a decision-maker in taking a decision based on a set of possibilities in the form of multiple criteria. The outcome of MCDM is in the form of a ranked order of the chosen alternatives (Malczewski 1999). Some of the GIS-MCDM-based EIA studies include genetic algorithm, mathematical programming, statistical modelling, Monte Carlo simulation, cellular automata simulation, agent-based simulation, process simulation, technique for order preference by similarity to ideal point, simple additive weighting, linear programming, fuzzy logic, Tabu search, evidential reasoning and AHP. The choice of MCDM method depends on the nature of research problem (Wang et al. 2006; Zhu 2011; Zolfani et al. 2011). Moreover, MCDM should be considered as an indicator towards correct weight and not as the final answer to the decision process (Table 2).

Table 2. Description of spatial analysis methods.

Method	Description
Buffering	A process of creating an area around a map feature delineated by a specific distance from the selected feature
Corridor analysis	Method of finding the optimal path between source and surrounding cells in a raster map subject to certain criteria
Cost distance analysis	A process of putting cost to each cell in a raster map (based on factors like elevation, time or fuel spent, etc.) followed by calculating the total cost incurred in moving from one location to another
Cumulative effects analysis	A process of assessing cumulative change in a set of variable of an area during a time period using overlay analysis
Data reclassification	A process of reclassifying cell values in a raster map based on certain criteria to generate a new raster map
Digital elevation model (DEM)	Data model of uniformly distributed elevation data of a terrain
Inverse distance weighing (IDW)	A deterministic spatial interpolation method based on the principle that, estimate of value at a location is more influenced by its immediate surrounding points than the ones away from it
Kriging	A spatial interpolation method based on probabilistic assumption that a spatial change of an attribute has a spatially correlated component. It has several variates like ordinary kriging and universal kriging
Map algebra	A process of using algebraic operators on raster maps to generate a new raster map
Nearest neighbour method	A statistical method that compares the distance between each point with its nearest neighbour point to an average distance
Overlay analysis	A process of combining spatial and attribute data of two maps to generate a new map
Proximity analysis	A distance-based selection process of features or cells in a map
Sensitivity analysis	A process to analyse the effect of change in the parameters on the system
Spatial interpolation	A collection of methods for predicting the value of unsampled points of an area using the values of sampled points
Spline (also known as thin plate spline)	A deterministic spatial interpolation method that generates a surface which passes over the known points with least possible change in slope of the surface
Triangulated irregular network (TIN)	A data model which approximates a terrain using non-overlapping triangles
Uncertainty analysis	A process of measuring the uncertainty or risk in decision-making
Viewshed analysis	A collection of methods used to analyse the portion of a terrain visible from one or more viewpoints

Source: Wikipedia (2016b), Chang (2008), Longley et al. (2005), Demers (2009), and Lloyd (2010).

AHP is a MCDM method which decomposes the decision process into several levels of hierarchy. Based on pairwise comparison of criteria for alternative options, a comparison matrix is made for evaluation of criteria weights (Saaty 1980, 1990, 2000). The data requirement for AHP is less rigorous than classic statistical methods which are based on historical data (Arriaza & Nekhay 2008). It has the flexibility to combine qualitative and quantitative aspects of opinions given by experts of environmental impacts, and capture the perception of stakeholders in EIA (Ramanathan 2001). The AHP model is very effective for large development projects (Dey 2010). Thereby, it is a convenient MCDM method for EIA. However, caution should be taken in the final ranking of alternatives, if they show close values (Triantaphyllou & Mann 1995).

AHP has been widely used in various disciplines (Vaidya & Kumar 2006). AHP and GIS have been used in land resource planning (Laskar 2003), land suitability assessment for wildlife restoration (Arriaza & Nekhay 2008) and EIA of highway projects (Agarwal 2005). AHP has been used in two ways in GIS viz., combining weights with the attribute map layers to give a weighted composite map; and secondly, for preparation of hierarchical structure of aggregation of priorities (Malczewski 2006). AHP has been used in transport-related EIA (Brozova & Ruzicka 2010). GIS–AHP was used in finding an optimal alternative forest road based on criteria viz. technical, economic, environmental and social factors (Zolfani et al. 2011). It has been used to derive priority maps for alternative road construction planning (Johnson et al. 2011) and for designing forest road networks (Caliskan 2013).

Uncertainty and sensitivity analysis in spatial analysis is progressively becoming important. Such analysis has a major role in understanding of the effect of inputs on the model output, and has a significant role in risk assessment and optimization of resource inputs in development projects (Crosetto & Tarantola 2000, 2001; Lilburne & Tarantola 2009; Feizizadeh et al. 2014). There is an urgent need to apply uncertainty and sensitivity analysis in spatial analysis of mountain highways.

4. Conclusion

Spatial analysis of mountain highways involves use of GIS integrated submodels. Such models use mathematical, statistical and algorithmic methods to predict the spatial and temporal distribution of impacts on environmental attributes. GIS-based techniques viz., overlay analysis, buffering, proximity analysis, spatial interpolation, least cost path, geostatistical methods, uncertainty analysis and sensitivity analysis, are used in the spatial and temporal simulation of impacts. MCDM in conjuncture with GIS plays an important role in weighing the environmental attributes and ranking of alternative scenarios. A wide range of MCDM methods are available in this context; however the nature of study is the predominant criteria for selection of MCDM method. Data visualization, in the form of 3D visualization, TIN, thematic maps, symbolic maps, tables and charts are used in the communication and decision-making of EIA reports. However, application of GIS in transport EIA in mountainous areas is constrained by availability of an appropriate digital database on environmental attributes, such as maps on forest type, landuse, inventory of past landslides to name a few. Secondly, lack of appropriate temporal and spatial resolution of remotely sensed imageries can be a major limitation on spatial modelling. Thirdly, limited distribution of sampling points in the study area will affect the reliability of impact model. Fourthly, location of study area, such as presence of military establishments, affects the accessibility of ground data, which will again affect model validation. Finally, availability of GIS professionals and affordability of GIS software and related hardware is a major constraint on EIA budgeting, especially for developing countries. On the other hand, with the rapid fall in price of PCs, internet service, openware GIS, online GIS, GPS integrated mobiles, and freely downloadable mobile apps there is an immense scope in geovisualization of spatio-temporal impacts of such EIA studies in mountainous areas. Stakeholders and public at large can easily understand the outcomes of EIA in the form of geoapp-based storytelling of impacts, web and mobile-based PGIS in the form of geoforms, and coupling of such EIA spatial models with web-based virtual Earths, like Google Earth and ArcGIS Earth for further analysis and interpretation.

Limited studies have been done in the field of spatial analysis of highway projects in mountainous areas. The majority of GIS-based EIA of highway projects are confined to urban areas and relatively simple terrain. This is largely due to the relative ease of data gathering and limited prediction capacity of the models in complex terrain. Satellite imageries are a major help to understand and evaluate impacts in mountainous areas. However, difficulty in ground truthing, availability of clear weather for image capture, and remote locations of habitations add to the challenge. Moreover, most studies have focused only on one or a limited set of impacts of the project on the physical, ecological or social environment. Thereby, such studies have not provided a holistic picture of the impact scenario. Thus, spatial analysis of mountain highways offers a tempting avenue for further studies.

Disclosure statement

No potential conflict of interest was reported by the authors.