Homepage (Introduction) | ge3216groupix

Introduction: VGI and Disaster Management

For many societies, disaster management frameworks are important strategies to mitigate the impact of natural disasters on communities (Baharin et al. as cited in Horita et al., 2013). Figure 1 shows a four-phased cycle framework surrounding a disaster: mitigation, preparedness, response, recovery (McLoughlin, 1985; Horita et al., 2013). The phases were planned and internally reviewed by government agencies where only government-produced information was recognised as ‘credible’ hazard information (Sutton et al., 2008).

Figure 1: Cycle of disaster management and associated activities (Poser and Dransch, as cited in Horita et al., 2013)

This is where Volunteered Geographic Information (VGI) seeks to enhance disaster management frameworks through publicly-generated information. VGI is a recent phenomenon where geographic information collectively created by citizens through appropriate technologies, allow users to upload geotagged information online which can be extracted for spatial analysis (Goodchild, 2007). The technology includes integrating with the Global Positioning System (GPS) technology in smartphones and social media applications,

Goodchild (2017) states that VGI offers the availability of local-specific spatial information, often neglected by government officials, at at lost cost, due to the low financial cost of accessing such online data. Therefore, researches has identified VGI to be a feasible way to enhance, update or complement existing geospatial datasets (Horita et al., 2013).

The integration of VGI into disaster management frameworks is a recent emerging body of research. These researches centres its investigation to the potential for VGI to enhance information collection and dissemination during the “response” phase of a disaster (Horita et al., 2013). In the 2011 Queensland flood in Australia, geotagged flood related images uploaded on Wikimapia, Flickr and OpenStreetMap were analysed to generate a Digital Elevation Model (DEM) of the flood height and extent (McDougall, 2011). Conversely, few research centres their investigation to the using of VGI to generate and disseminate information during the mitigation and preparedness phases of disaster management (Horita et al., 2013).

This project aims to expand on this literature gap through exploring how VGI can be used to aid vulnerability assessments of hazards, generating information which can enhance the mitigation phase of disaster management. San Diego, United States of America will be used as the case study for this project

RESEARCH AREA AND JUSTIFICATION

Research Area and Justification

In this study of San Diego’s disaster management, the project considered both the tsunami and wildfire hazards.

Tsunamis in San Diego

Most of the eighty tsunamis ever observed or recorded in California, are small and have yet to cause insignificant damages and loss of human lifes (McCaffrey, Toman, Stidham & Shindler, 2013). Yet, given the increase in population and economic developments along the coastal regions in San Diego, California, the dangers of tsunamis cannot be disregarded (McCaffrey, Toman, Stidham & Shindler, 2013). Sudden offshore earthquakes that results in shorter response time for warning has the danger to cause significant economic damages and the lost of lives. With the Port of San Diego, a major economic infrastructure in California, also located within the the coastal areas, damages to the port due to the failure to develop an effective disaster management policy will be detrimental to the economy of California.

Wildfires in San Diego

Wildfires are common in San Diego. The county is characterised by semi-arid coastal plains, occasional Santa Ana winds and high temperatures (Office of Emergency Service [OES] & Unified Disaster Council [UDC], 2010). Due to high temperatures and low humidity, foliage on the forest floor easily dries up and become vegetative fuel for wildfires. Historically, wildfires have brought significant damages to San Diego. Between 1950 - 2007, eight “States of Emergency” were declared due to severe wildfires (OES & UDC, 2010). In 2007, a major wildfire killed 10 people and saw 33 other casualties. Another wildfire in 2003, caused 14 deaths (OES & UDC, 2010). Therefore, the frequency and damages that entails wildfires in San Diego makes it necessary for mitigation and preparation plans and policies to be prioritised to reduce the loss of lives and properties.

This can also be the first steps in using VGI to address the lack of research in mitigation and preparedness policies. While current mitigation plans, heavily relies on government-sourced geographical data for hazard identification and vulnerability analysis, this research aims to encourage greater community involvement in the formulation of disaster mitigation plans.

Research Area and Justificatin

METHODOLOGY

This project uses Flickr data, provided by Dr Yan Yingwei [1]. Data from 2012 to 2017 was extracted, with 100,000 VGI Flickr data points (Figure 2). The temporal limitation ensures that the analysis will be relevant to recent trends of actual population presence and activities to specific areas of interest.

methodology and analysis

Figure 2: All 104013 VGI Flickr data in San Diego from 2012-2017

High quality data has to be retrieved and extracted to ensure that the data remains applicable for contemporary disaster management policies.

For Figure 3, spatial and qualitative tools were applied. Under spatial tools, three parameters were utilised:

Point Density Map: Identifies areas experiencing high amounts of human traffic flow;
Road Service Layer:Identifies areas far away from ambulance staging locations;
Fire and Tsunami Hazard Layer: Directly identifies area prone to wildfire and tsunami risks.

These parameters identify regions in San Diego that can be considered as places with the most vulnerable to high rates of human casualties during wildfire or tsunami occurrences, known as ‘priority areas’. VGI data within the ‘priority areas’ will then be extracted. Keyword Filtering was then applied as a qualitative method to extract high quality VGI data that contains photographs information useful for designating unique disaster management policy.

[1] Dr Yan is a research fellow at the National University of Singapore, Department of Chinese Studies, whose research interest includes Geographic Information System, VGI, social media data and disaster and crisis management.

Methodology and Analysis

METHODOLOGY

human Density and flow

To observe human traffic density in San Diego, the point density tool was used to obtain a raster map which differentiates each cell to 5 classes, according to the number of VGI data within close proximity of each cell, as shown in Figure 3.

Figure 3: Point density map in raster. Classes 3-5 (yellow, orange and red raster cells) were determined as areas with high point density, while Classes 1 and 2 (light and dark brown raster cells) are areas with 0 or only 1 data in close proximity .

Figure 4: Point density map with VGI data filtered.

An analysis of this point density map reveals locations with little or no human presence. VGI data that falls in Classes 1 and 2 were considered areas of low human presence and the lack of activity to justify policy implementations in the area. Data in Classes 3 to 5 will be prioritised instead (Figure 4).

METHODOLOGY

This research also accounted for the ambulance response time in during a medical emergency since our study aims to improve medical response across San Diego, including the urban and rural areas. San Diego’s contract with the ambulance service company, Rural/Metro Ambulance, requires medical emergencies to be tended to within 12 minutes of receiving the emergency call (San Diego County Grand Jury, 2014). Areas with significant human presence outside of this 12 minutes response time coverage may hence face greater risks during a disaster due to untimely response arrival. Hence, we adopted this as one of the three factors in our study to determine areas of concern.

service layer

Figure 5: Road Service Layer map

Figure 6: Road Service Layer map with VGI data filtered

The areas where ambulances cannot arrive within 12 minutes are considered when devising the priority areas for mitigation policies, through the form of a Road Service Layer map (Figure 5). Casualties in these areas are more susceptible to delayed medical response which lowers their survival chances. In emergency relief operations, the immediate needs of the situation must be met with the correct resources in a timely manner (Alexander, 2015). It is useless to plan for the delivery and use of resources that will not reach those in need within a crucial time frame (Alexander, 2015). Therefore, the focus on emergency and response relief should be shifted to mitigation measures instead. VGI data within these areas was therefore further extracted to be studied for improving disaster mitigation measures (Figure 6).

METHODOLOGY

hazard layers (wildfire severity and tsunami inundation maps)

Figure 7: Wildfire severity and tsunami risk zones. Yellow, orange and red areas represent regions most at risk to wildfire events. Blue area represents maximum tsunami runup.

Both the wildfire severity and tsunami risk zones in San Diego are represented in Figure 7. To identify zones of higher susceptibility to fire hazards, the Fire Hazard Severity Zone map provided by The San Diego Association of Governments (SANDAG) was utilised. Factors such as fuels, weather and terrain were considered in the modelling process (SANDAG, 2015).

For zones susceptible to tsunami events, SANDAG’s Tsunami Inundation Map was used. Constructed by the University of Southern California (USC) Tsunami Research Center, this map displays areas likely to be affected by tsunami based on maximum tsunami runup obtained from tsunami modelling (SANDAG, 2015). The model factors in various local and regional tsunami source events, including local and regional earthquakes, submarine landslides and near-shore landslides.

Figure 8: Wildfire severity and tsunami risk zones with VGI data filtered

The inclusion of this map would allow us to identify specific areas in San Diego that are most susceptible to wildfire and tsunami hazards. This allows better identification of regions for prioritising wildfire and tsunami mitigation strategies (Figure 8). Thus, relevant VGI data within these regions can be extracted for analysis of local factors to further enhance wildfire and tsunami mitigation measures.

METHODOLOGY

output map

Figure 9: All filtered VGI data

Hence, based on the three parameters: Point Density Map, Road Service layer and Hazard Layer, we have identified specific VGI data that is deemed as having the greatest relevance and priority which can then be applied to enhance mitigation measures specific to wildfire and tsunami hazards, as shown from Figure 9.

METHODOLOGY

keyword filtering

Figure 10: VGI data depicting a flower, which is irrelevant for disaster mitigation policy

Figure 11: VGI data depicting a gathering of people, which is irrelevant for disaster mitigation policy

After extracting the relevant VGI data, some of these data may still be irrelevant to disaster mitigation, as shown in Figure 10 and 11. Data quality has to be assessed, and irrelevant data should then be filtered out. Goodchild and Li (2012) discusses the evaluation of VGI quality by suggesting an automated quality assessment tool to examine the quality of each data through a designated criteria. Criscuolo et al (2013) also suggested that quality criteria can be determined through a reputation system, which rates VGI data based on the contributor’s age, education, career etc. However, such information may not be publicly accessible. It is subjected to the contributor’s intentions and willingness. Information tagged to photographs can be directly extracted and assessed to verify data relevance to disaster management.

Figure 12: VisualBasic Script used to filter Wildfire-related VGI data, through keywords such as: fire*, tree*, camp*, shrub*, bush*

Figure 13: VisualBasic Script used to filter Tsunami-related VGI data, through keywords such as: tsunami*, water*, beach*, wave*

According to Feng and Sester (2018), keyword filtering methods can be used to retrieve disaster related data from VGI. A study in 2016 attempted to incorporate Twitter data to formulate effective evacuation routes during rapid mass flooding events through the case study of the 2015 South Carolina floods. In order to extract flood-related tweets from all geo-referenced tweets during the event, “message or hashtags contain either of the keywords ‘flood*’ or ‘Joaquin’” were filtered (Li et al., 2017). This method allows efficient retrieval of relevant VGI data for disaster management without having to scrutinise every photograph extracted. For our study, keywords such as ‘fire*’, ‘tree*’ ‘shrub*’, ‘bush*’ and ‘camp*’ were used to filter out data relevant to wildfire mitigation policies, while keywords such as ‘tsunami*’, ‘water*’, ‘beach*’, ‘wave*’ were used for to filter out data relevant to tsunami mitigation policies (Figure 12 and 13). By examining VGI filtered with these keywords, it is possible to identify data of greater relevance to the two natural disasters that are included in our analysis.

METHODOLOGY

https://drive.google.com/open?id=1Xg12XmIvLboS8bOkzK5fhtjpXhrH2je1&usp=sharing

Google map overlay

Figure 14: A photo of a VGI data in Paso Picacho Campground after spatial and qualitative filtering, which is useful for designating wildfire mitigation policies

In order to ensure widespread adoption of our project, filtered VGI data was overlaid onto Google Maps for interested users to study[2]. As displayed, the map allows users to freely examine extracted photographic data relevant to wildfire and tsunami vulnerability in San Diego. One example can be seen from Figure 14, in which relevant and high-quality VGI data from the Paso Picacho Campground is extracted. Specific applications of such usage for wildfire mitigation policies will be discussed below.

[2] Map available for public viewing:

Results and Discussion

RESULTS AND DISCUSSION

results and discussion

By incorporating and analysing VGI data, relevant authorities can enhance existing disaster management policies with local-specific information. Our discussion focuses on applying VGI through examining geotagged photographic images of specific locations in mitigating a potential wildfire. In this section, we examine current mitigation policies and explore how VGI can be integrated to further prevention and preparedness measures.

Although our methodologies is applicable for the analysis of different hazards, our discussion focuses specifically on wildfires due to its greater significance in San Diego. To date, only 7 tsunami events have caused damage to San Diego since historical records from 1806. Conversely, wildfires occur frequently: six major wildfires occurred between 2007 and 2012 (The San Diego Regional Fire Foundation, 2012). Furthermore, the aforementioned impact of wildfires highlight the seriousness and urgency of improving wildfire resilience in San Diego.

Fuel reduction treatment and participatory mapping

With reference to the official CIty of San Diego brush management guidelines (The City of San Diego Fire Safety and Brush Management Guide, 2016), current management of reduction treatments recommends thinning that involves cutting down 50% of vegetation over 2 feet in height, to a height of 6 inches in a staggered manner to reduce fuel load and creating zones of fuel breaks. However, such strategies are often based on general guidelines for authorities and members of public to follow and are not specific to localities. To supplement the formulation of fuel reduction treatments for specific localities, VGI data and extracted geotagged photographs can be integrated into the planning process. By looking at the photographic information obtained through our methods of VGI processing, planners can study vegetation types, topography, presence of people and human activities, and presence of community assets such as houses, to assist them in the planning of fuel reduction for specific areas that are deemed as high risk. To make the incorporation of VGI into fuel reduction policies more relevant to local contexts, a dedicated platform (instead of using conventional social media websites such as Flickr) can be created for local communities (for example at the county level) to engage in participatory mapping through VGI (Haworth, Whittaker and Bruce, 2016). This platform will allow community members to take pictures and record information about any potential fuel hazards such as very dense bushlands or overcrowding of camp activities. Hence, by analysing geo-tagged photographs obtained, it could assist authorities in devising and altering thinning plans and designing and positioning of fuel breaks specific to the vicinity of the identified area to facilitate more effective policy making.

Wildfire patrol programmes

Although there are currently no concrete patrol programmes in San Diego, we suggest that VGI-integrated patrol programmes can be looked into by the authorities to complement existing disaster management policies. Taking reference from Carbondale & Rural Fire Protection District’s wildfire patrol programme, wildfire patrol crews are deployed to make site visits to WUIs, perform risk assessments and offer advice to homeowners. (“Wildfire Patrol Programs to Begin in Basalt & Carbondale Fire Districts”, 2015). Such programmes not only reinforce monitoring but also serve as a way to provide wildfire education to the community and to create greater rapport between authorities and the public. However, it is impossible for such patrol programs to cover the entirety of San Diego; remote areas that are well frequented and at risk of wildfires may be neglected by patrols. To address this shortfall, georeferenced data obtained from VGI can be implemented into patrol routes to highlight areas with wildfire risk and significant human presence. Furthermore, priority areas can be constantly reviewed by regularly extracting VGI from online sources, allowing patrol routes to adapt to any changes in distribution of human presence. With the inclusion of VGI data analysis, personnel such as forest rangers can optimise their patrols by prioritising high-risk areas.