GIS Modelling with Rapidly Changing Data sets an Application of Model Builder to Assess Public Accessibility in Colombo City

: Geographic Information System modelling with rapidly changing datasets requires systematic development of analysis sequences incorporating sufficient details and establishing flow of each process. Absence of a structured approach would consume a significant time for recalculations to accommodate datasets that require consistent and frequent updating. A typical case is the modelling of the spatial variations of the accessibility in a specific land extent based on zoning changes, unwelcome incidents, national or local ceremonies etc. Apart from such dynamic data the relatively static data such as land use, road network admin boundaries become a part of a GIS model's base data set. A GIS modeller needs the vision to structure a particular application enabling the generation of output maps on each occasion of changes affected either to one, a few or all data layers. Hence it is of great importance to identify the method, strengths and weaknesses of such an application. ModelBuilder is a component of ArcGIS software which enables creating, editing and management of GIS models. There are two basic application methods. One is the capability enabling exploratory project works, and the other includes the development of generic tools that would be reused and shared. The ModelBuilder creations enable the visualization and exploration of results in ArcMap/ArcCatalogue. The ModelBuilder also facilitates the changes of parameter values, rerun selected processes, add or delete processes and intermediate data. The present work is an application of ModelBuilder to assess the spatial variability of accessibility in the city of Colombo. Data sets of 1:50,000 scales were used with ArcGIS software. The systematic development of ModelBuilder assembly and the potential of GIS modelling results generation with ease for varying data layers. Situations are demonstrated.


Introduction
Geographic Information Systems are cutting edge information technology tools that facilitate the modelling of geographic information to arrive at rational decision making. In case of geographic data modelling or commonly known spatially distributed modelling, there are two kinds of data. They are relatively static data and dynamic data. The relatively static data are such data that can be taken as static within a short time span of about six months or one year. Data corresponding to places of public safety mobilization, similar to locations of meetings, incidents, movement of working population in and out of cities etc., falls into the category of dynamic. Sometimes even zoning demarcations fall in to the category of dynamic data. In modelling efforts, it is of great importance that model inputs are suitably adjusted whenever changes have taken place.
GIS model inputs can change in many ways. The geometric feature and their attributes may vary in one layer or in many layers, thus causing a modeller to perform many operations repeatedly to achieve the appropriate output. In case of real life scenario modelling in GIS, a modeller has to carryout many operations involving both single layer and multilayer computations to arrive at acceptable results, which require great care. Such GIS work consumes significant time. Once a particular GIS model is developed scientifically, then changes in the geographic environment would cause only a change in the base data while the process would remain constant. A decision maker therefore, would expect the analysis of a particular system at a very short time if only a few changes for the base data are to be incorporated. Under these circumstances a modeller needs to find ways to customize his/ her model in such a way that once changes are made to the base data, the execution of processes would take place with minimum time.
The model builder of ArcGIS software is a tool that enables model components to be first defined as input output modules consisting of process operations and then facilitating the combining of components to a single system. In this assembly, the system flow direction and sequences are clearly defined and therefore, once a system is designed to operate with ModelBuilder tool, it would function efficiently when changes to base data are effected. There are many advantages of using model builder listed in literature. Along with significant easiness to work, use of model builder would also ensure that the GIS databases adhere to the rules of operations, the process sequence would be made fixed and static, would be specially declared, would enable making changes, to each and every component to a section of the model without causing much labour to others, and would enable easy parameter or layer changes with reliability, (http://www.esri.com, http://www.nysgis.state.ny.us [1]) In ad hoc model computations on ArcGIS platform the incorporation of a unique process with a set of base data is not possible. Therefore, repetitive model computations in an ad hoc system would consume a significant time thus making calibration and verification of complex model a cumbersome task. Though it seems obvious that the advantage of model builder is in the saving of time and its reliability, there lacks a comparison of time advantage against the conventional, ad hoc, step by step modelling while attempting to incorporate a real life case study. The present work is a case study of Colombo city accessibility in an environment of changing unwelcome incidents.
Colombo city often experiences unwelcome incidents such as public meetings, political rallies or protests etc. They are often treated as unwelcome incidents by many, because of the traffic congestions created by such incidents which are unacceptable due to loss of time and increase of fuel expenditure. If a decision-maker could identify the effects of changes causing public accessibility in a geographically distributed manner, then there are opportunities to provide many rational resource mobilization solutions. The case study using information of land use, road network, administration boundaries etc as static data and unwelcome incidents as dynamic data, applies a conceptual GIS model to assess the accessibility. The case study application is carried out with and without the model builder in order to critically evaluate advantages in the use of ModelBuilder for GIS modelling.

Objective
Objective of the study is to develop a GIS model for the assessment of public accessibility in Colombo city with and without incorporation of ModelBuilder tool and to make a critical evaluation.

Study Area
Colombo the Capital of Sri Lanka is located on the west coast of the country. Colombo city lies in between 07° 12' and 07° 20' of North latitude, and between 80° 11' and 80° 14' of East longitude. Colombo is considered heavily urbanized when compared with the rest of the country. With the recent security concerns, the law enforcement officials have incorporated a modified boundary for Colombo zoning. Since accessibility studies too need to be closely linked to security, the present study considers that the boundary of the law enforcement units as the study area. In the North, study area is bounded by the Kelani River, West by the sea, South by Dehiwala, Esat by Sri Jayawardanapura and Kolonnawa DS divisions.
The study area (Figure 1 (Table 2) serves a total resident population of about 710,000 and an approximate migrant population about 2 million.
The land cover distribution and the road distribution in -each of the administrative divisions within study area are shown in Table  1 and Table 2.
The traffic congestion experienced in most roads of the area is considered as extremely high and this is specially during office and school hours. The city's commercial centre Fort and Pettah are considered locations that should be avoided at any time of the day except late night in case if one desires to access a location in these areas. Locations such as Maradana, Borella, Town Hall, Maligawatta also fall in to such category.
There are many cricket stadiums, theatres and public meeting places such as Torrington Square and Hyde Park Corner, which attracts a significant number of public and private vehicles creating accessibility problems.
The present study covers the accessibility of the road network in relation to the unwelcome incidents that occur at or adjacent to the road network.

Methodology
The methodology flow that of the study area is shown in Figure 2. Model concept development included identification of the objective function for the assessment of road accessibility. Accessibility is concerned with the opportunity that an individual at a given location possesses to participate in a particular activity or a set of activities. Accessibility is usually measured in  ) In a spatial assessment of accessibility, if one would identify an accessibility indicator that could be assigned to land parcels which have to be crossed between the source and destination, then the sum of individual land parcel accessibility indicator over the spatial units crossed by the travel route would be indicating the accessibility. In this study, a GIS model computes spatial accessibility indicator for land parcels in the study area through simple overlay computations. In the GIS model the spatial accessibility indicator was taken to be directly proportionate to the population density, road network distribution, land use, Entry Exit points and spatial distribution of the unwelcome incidents. A questionnaire survey was carried out to identify the existence of other parameters which governs accessibility indicators. Users were requested to rank the parameters according to the influence on accessibility. A sample of 51 persons who move in and out of Colombo city and who frequently access Colombo were involved for the survey. Users indicated parameters and their importance are as shown in the Table 3.
Ratio Estimation Procedure (Jacek [4]) was used to compute the normalized weights for each parameter and these values were used in the model to incorporate relative difference in the influence during overlay operations (Table 4). Each influencing parameter characteristic was extracted from base data layers (TableS) to develop individual data layers for each parameter. As threshold influence values corresponding to spatial zoning of the selected parameters were not available, and especially because such variations depend on the locality and users.     The determination of spatial variation pertaining to each layer characteristic was identified through the user survey. Spatial zoning of each layer that was determined through an analysis of responses, is shown in Table 5. Each layer was zoned into a several quantitative classes. In the Table user percentage indicated the frequency of user responses corresponding to the selection of a particular spatial zoning. In the direct overlay method, Each layer was reclassified and GIS overlay using georeferencing tools of ArcGIS was carried out to arrive at the Accessibility Indicator layer.
Since the objective of the work was to identify the difference between the direct overlay method of ArcGIS, and the use of ModelBuilder, a time count was taken at each operation to facilitate comparison. In both methods the GIS modelling used weighted averaging method to quantitatively assess a combined effect of selected data layers. GUI of each element. This enabled an easy assessment and an easy assembly of model flow chart which had a drastically reduced process flow diagram when compared with that of the direct operation method. The ModelBuilder assembly ensured clarity and easy manoeuvrability thus ease of operation. Each assembly was checked for logical functionality, process execution error, and accuracy of results, in a stepwise manner. This methodology was proved as the best option for easy project completion. ModelBuilder standard colours were used to ensure clarity of model process identification. The process model on the ModelBuilder is shown in the Figure 3. During each computational step, ModelBuilder computation results were compared with the direct overlay method and the operational errors and process errors were verified for accuracy and differences. Result at each operation was compared through a comparison of feature numbers at the end of each land mark process. The landmark process comparison revealed that the results had no differences. The landmark process of selected for comparison are listed in the Table 6. 6.1 Result of user survey with respect to data layer preferences and the computation of weights for the GIS overlay model is shown in Table 3. The computed weights indicated a 25% and 22% value for incidents and roads respectively. Since the influence of waterways on accessibility of a particular land parcel is relatively small, the user rank was the lowest for the waterways dataset.

Model Builder
6.2 Spatial zoning of various parameters are dependant upon the stakeholder opinion through a judgemental assessment could be made based on common knowledge, available literature and experiences with respect to other spatial references. Therefore suitable stakeholder surveys should be carried out using well designed questionnaire (Tan,[5]). The present study incorporated a sample of 51 persons for the values used in the computations. Appropriate and suitably identified samples satisfying the objectives must be used for similar studies. The present study used a judgemental approach together with a frequency analysis of user responses to identify spatial zoning parameters for the model. The user frequency values obtained for each parameter are shown in Table 5. In case of population, the natural breaks of the spatial data frequency of population in Grama Niladari Divisions (GND) was utilized. The details such as land cover and population can use the parameter values of data occurrence frequency for spatial zoning. These capabilities are available in off the shelf GIS software.  Figure 3 and   Figure 4. Figure 6 shows that for the selected incidents, the accessibility indicators on a three class qualitative grouping has 23% for Low accessibility regions within the study area. The accessibility indicator status averaged for each Grama Niladhari Division shows that out of 74 GN divisions, 23 very low accessibility.
6.6 The ModelBuilder output for a different set of incidents is shown in Figure 5.

Conclusions
7.1 GIS model development to identify stakeholder requirements, should utilize appropriate methodologies to identify parameter prioritization and spatial zoning requirements with a suitable incorporation of stakeholder input assessments and state of the art methodology.