U-Alert
Lead Research Organisation:
University of Sussex
Department Name: Sch of Global Studies
Abstract
Key words:
Africa, floods, heat waves, air pollution, climate change adaptation, mobile data, satellite, remote sensing, urban, risk
Summary:
Africa is one of the most vulnerable continents in the world to climate change with widespread low adaptive capacity and resilience. Climate change is already exacerbating many development challenges such as endemic poverty and inequality, ecosystem degradation, natural hazard related disasters and food security. The frequency and severity of extreme climatic events including droughts and intense rainfall events has increased in recent years. These climatic events are now having a significant impact on people's lives and livelihoods causing deaths and damage from heat waves, flooding and air pollution. Climate change is also making these events more unpredictable and difficult to manage, especially in densely populated urban environments where emergency services and infrastructure is often inadequate and thus unable to respond to rapidly changing situations and multiple simultaneous pressures.
At the same time, urban areas are becoming increasingly populated as people migrate from rural areas in search of better employment opportunities and a standard of living. Due to this sudden influx of people (and expansion of urban areas and slums), many vulnerable people are being forced on settle on marginal land such as flood plains and steep slopes and thus being placed at greater risk. There is a growing demand for new forms of environmental monitoring to better predict and be able to respond extreme climatic events. Unfortunately, many regions in Africa are characterised by poor environmental data and sparse weather stations, there is a lack of observations to validate and downscale climate change projections, accurately predict and assess impacts as well inform short/long term adaptation strategies. Many cities and infrastructure users in Africa are reliant on the services and expertise of international and national experts constrained by data limitations, who are reliant on coarse resolution climate data and projections which can result in misinterpretation and misinformation being issued. There is thus a pressing need for better localised hazard prediction systems to monitor and manage the impacts of extreme events such as flooding, heat stress and air pollution, informed from and delivered directly to at risk populations.
It has been demonstrated that signals sent between mobile phone antennas can provide an alternative means to measure rainfall intensity and duration in situations where in-situ data isn't available. This data can be combined with hazard models to predict the degree and extent of flooding in urban areas. Furthermore, other studies have demonstrated that accurate air temperature information can be derived from internal smartphone sensors and by measuring battery discharge fluctuations. Simultaneously, satellite derived aerosol measurements can provide information on air pollution and public health risks. The combination of the aforementioned mobile derived data and remote sensing techniques can be used to warn urban populations of impending climate risks. The growing number and use of mobile phones in developing countries allows SMS warnings and advice to be delivered to those in need of this information, as well as a means to collect data on essential climate variables through the creation "citizen observatories". This project aims to assess the feasibility and scope of bringing together such technologies to collect information of urban risk and vulnerability, analyse it to produce warnings and deliver it directly to mobile phone users in affected areas in an "easy-to-action" message.
Africa, floods, heat waves, air pollution, climate change adaptation, mobile data, satellite, remote sensing, urban, risk
Summary:
Africa is one of the most vulnerable continents in the world to climate change with widespread low adaptive capacity and resilience. Climate change is already exacerbating many development challenges such as endemic poverty and inequality, ecosystem degradation, natural hazard related disasters and food security. The frequency and severity of extreme climatic events including droughts and intense rainfall events has increased in recent years. These climatic events are now having a significant impact on people's lives and livelihoods causing deaths and damage from heat waves, flooding and air pollution. Climate change is also making these events more unpredictable and difficult to manage, especially in densely populated urban environments where emergency services and infrastructure is often inadequate and thus unable to respond to rapidly changing situations and multiple simultaneous pressures.
At the same time, urban areas are becoming increasingly populated as people migrate from rural areas in search of better employment opportunities and a standard of living. Due to this sudden influx of people (and expansion of urban areas and slums), many vulnerable people are being forced on settle on marginal land such as flood plains and steep slopes and thus being placed at greater risk. There is a growing demand for new forms of environmental monitoring to better predict and be able to respond extreme climatic events. Unfortunately, many regions in Africa are characterised by poor environmental data and sparse weather stations, there is a lack of observations to validate and downscale climate change projections, accurately predict and assess impacts as well inform short/long term adaptation strategies. Many cities and infrastructure users in Africa are reliant on the services and expertise of international and national experts constrained by data limitations, who are reliant on coarse resolution climate data and projections which can result in misinterpretation and misinformation being issued. There is thus a pressing need for better localised hazard prediction systems to monitor and manage the impacts of extreme events such as flooding, heat stress and air pollution, informed from and delivered directly to at risk populations.
It has been demonstrated that signals sent between mobile phone antennas can provide an alternative means to measure rainfall intensity and duration in situations where in-situ data isn't available. This data can be combined with hazard models to predict the degree and extent of flooding in urban areas. Furthermore, other studies have demonstrated that accurate air temperature information can be derived from internal smartphone sensors and by measuring battery discharge fluctuations. Simultaneously, satellite derived aerosol measurements can provide information on air pollution and public health risks. The combination of the aforementioned mobile derived data and remote sensing techniques can be used to warn urban populations of impending climate risks. The growing number and use of mobile phones in developing countries allows SMS warnings and advice to be delivered to those in need of this information, as well as a means to collect data on essential climate variables through the creation "citizen observatories". This project aims to assess the feasibility and scope of bringing together such technologies to collect information of urban risk and vulnerability, analyse it to produce warnings and deliver it directly to mobile phone users in affected areas in an "easy-to-action" message.
Planned Impact
The immediate impact of this project will be felt by researchers, data providers and environmental consultancies through the provision of a technically robust case for support for an early warning system of flooding, heat stress and air pollution based risks in urban environments in developing countries. Further the project will impact the satellite applications community by assessing the evidence and feasibility of using commercial cellular communication networks in combination with satellite data to measure and disseminate information on environmental hazards. At international, regional and national levels the research will derive value through the involvement of various stakeholders from the onset of the project by the creation of a community of practice whereby an interactive network of community members, researchers and humanitarian organisations engage to produce a common knowledge system, shared goals and specifications for an urban based early warning system. This research will impact international and national policy via existing connections from the project team into national ministries and planning authorities and international networks such as the IPCC, the Resilience Centre, the Stockholm Environment Institute and the Intergovernmental Authority on Development (IGAD) Climate Prediction and Applications Centre (ICPAC). We will also use the PI's association with the ALERT project; a three-year program funded by the department for International Development (DFID)'s Disasters and Emergencies Preparedness Programme (DEPP) through the START Network. The project is delivered through a multi-agency consortium consisting of Care International, Concern, Handicap International, HelpAge International, Islamic Relief, Oxfam and Coventry University. The ALERT project aims to improve ways by which humanitarian agencies are able to prepare for and respond to disasters through the use of technology. Whilst ICPAC is responsible for receiving and processing climate information to provide climate services seven member countries namely: Djibouti, Eritrea, Ethiopia, Kenya, Somalia, South Sudan, Sudan and Uganda as well as Burundi, Rwanda and Tanzania.
The mid to long-term impact of the provision of accurate early warning information derived from the system being assessed will be to enable institutions and individuals to respond in a timely manner to hazardous situations. In setting the innovation in informal settlements we are choosing to develop an early warning system for often marginal populations with little or no access to formal governmental disaster preparedness support. Furthermore the innovation will provide support for populations that because of their setting have been seen traditionally as difficult to provide early warning information for due to the transient and unplanned nature of their location. Previously no system for the provision of early warnings of floods at this scale, for this population and with this flexibility exists.
The mid to long-term impact of the provision of accurate early warning information derived from the system being assessed will be to enable institutions and individuals to respond in a timely manner to hazardous situations. In setting the innovation in informal settlements we are choosing to develop an early warning system for often marginal populations with little or no access to formal governmental disaster preparedness support. Furthermore the innovation will provide support for populations that because of their setting have been seen traditionally as difficult to provide early warning information for due to the transient and unplanned nature of their location. Previously no system for the provision of early warnings of floods at this scale, for this population and with this flexibility exists.
Organisations
Publications
| Description | We have evaluated the potential design, impacts and implications of an innovative and integrated early warning system for extreme events in developing urban areas based around use of the mobile network. |
| Exploitation Route | The following outlines a follow-on project utilising the results of the initial scoping exercise contained within this report. It is envisioned that a robust mobile EWS could be developed as part of a large programme of support, involving humanitarian response and local stakeholders. The results of this scoping exercise suggest that cross-validation of results and forecasts will be necessary, largely due issues of credibility, error propagation and uncertainty in the underlying evidence provided by unconventional data sources such as mobile networks. The proposed project would seek to address this project by bringing together a complementary team of engaged (and impacted) stakeholders to develop robust strategies to respond to extreme climate events in urban areas. Project proposal This scoping exercise culminated in the development of an outline proposal focussing specially on the application of mobile derived observations to estimate flooding. The results of this scoping report could be similarly used to develop a temperature based proposal, indeed both proposals could be integrated to produce a larger proposal addressing multiple elements of urban risk. This outline proposal builds on past and ongoing exchanges involving climate scientists from national meteorological departments and universities in Senegal and Kenya and the UK with policymakers from a number of international humanitarian organisations in two demonstration studies, one in Kenya and one in Senegal. The aim of these exchanges has been to explore how climate science can more effectively support humanitarian, disaster risk reduction and development planning. The exchange has created a space for the cross-sectoral, cross-departmental dialogue required to support community resilience. Many participating in the exchange have strongly welcomed the creation of space for this dialogue, recognising its current absence at many levels of decision making. In Senegal, exchange workshops have brought together expertise and extension workers from the Government Departments of Water and Agriculture, with the National Meteorological Agency and university and institutional expertise in climate research and environmental mapping. The approach has also provided an unforeseen framework for innovation, creating the linkages and space to develop new approaches which offer the potential to better meet users' information needs. The process includes community-based evaluation, technical reviews of the provision and use of climate information following each season and activities to identify learning from across the two exchanges. The evaluations have enabled communities to identify gaps in the provision of relevant climate information and suggest more effective formats and channels for information dissemination. The technical reviews have afforded scientists the opportunity to consider whether there are additional sources of climate science date or new ways of using emerging technologies which might allow them to better meet user needs. It is within this process that the demand for this innovation has been identified. It is also the intention of this proposal to take advantage of this pre-existing network to set the priorities designing the early warning system (EWS), and develop and test the EWS including assessing dissemination and response. Methodology Risk Knowledge Phase The first part of the development of the EWS could focus on understanding the detailed risks faced by floods in the informal settlements of Dakar and Nairobi to include cascading risks such as communicable diseases linked to poor sanitation. In initiating the project a series of workshops could be held with members of the informal settlements to assess the risks posed by flooding with information gained on flood extents and consequences of flooding to include cascading risks. Community participants could be identified from the vulnerability and capacity assessment carried with attention paid to the representation of gender, age and ethnicity. The workshops could also involve the participation of members of the Government Departments of Water and Agriculture, the National Directorate of Water Resources, the National Meteorological Agency and university and institutional expertise in climate research and environmental mapping. During these workshops Global Position System (GPS) based mapping could be developed of flood affected areas with the participation of the community. Additionally data would be retrieved during this period to help calibrate and validate the remotely sensed (photogrammetry and SPOT stereo satellite data) derived digital elevation model, land cover and drainage patterns (derived from Landsat satellite data). A network of rain gauges could then be installed and assigned to particular residents to safeguard and record measurements. These data could be used to validate the rainfall measurements. Knowledge gained from the workshops could be used to set priorities for mitigation and prevention strategies and designing the early warning system. To personalise risk information there is a range of information which could be included in baseline data which a smart phone user could enter into an app created for the purpose. The position of the household could be automatically generated from the location of the smart phone. Through integration with mapping software, this would provide an indication of the physical risk of the household, considering proximity to water bodies, and topography and soil types to calculate surface run off and flood routes. A more accurate picture could be provided if households are able to share details on the physical condition of the home (composition of walls and ceiling and connection to utilities). Then it would be necessary for a household representative to enter the names, ages and physical needs and medical conditions for each member of the household. All of this data could be shared with first responders, government and humanitarian agencies. Potentially this information could be integrated with, and corroborated by local records or even the national census, combining other indicators of urban risk and vulnerability contained in this report. Monitoring and Predicting Phase Development of the technical component of the early warning system will comprise the retrieval of Received Signal Level (RSL) measurements; the conversion of these data into rainfall measurements; identification of convective rain cells; advection of cells; the input of these data into the flood model and overlay of model outputs on population information. Currently RSL data is collected in the area from a network of commercial microwave communication links of lengths from 0.3 to 20km at 15 minute intervals at frequencies from 8-38GHz. RSL data in general are subject to variability caused by variability of the drop size distribution along the link (Messer et al 2006), wet antenna attenuation (Leijnse et al 2007) and uncertainty in determination of clear air attenuation due to water vapour induced attenuation and scintillation effects (Holt et al 2003; Rahimi et al 2003). Moreover, at tropical latitudes networks operate at lower frequencies which can increase the error due to an increased non-linear relationship between rainfall and attenuation at lower frequencies (Overeem et al., 2011). The methodology followed to convert the RSL data to gridded rainfall data follows the work of Goldshtein et al (2009) with a four stage approach of preprocessing; point rain rate estimation; line to points data conversion and generation of rain maps. In the preprocessing stage the data are used to first identify rainy periods and separate them from dry periods using the correlation of attenuation of neighbouring links. The next process in this preprocessing stage is the extraction of the rain induced attenuation from the RSL data by subtracting dry period measurements from the wet RSL measurements. The third process of preprocessing involves using the simplified power-law empirical relation between rain rate and attenuation: A(dB)=aR^b L (7 1) where R (mm/h) is the rain rate along the link L (km) and the coefficients a and b are mainly functions of frequency, polarization but also of drop size distribution (DSD) and rain temperature (Olsen et al 1978; ITU-R 1999). The coefficients a and b will be computed assuming generic DSD and an analytical expression proposed by Zhao (2001). Errors involved in this process are caused by variations in the DSD along the link, although these are claimed by Goldshtein et al (2009) to be relatively small. The final processes in pre-processing involve estimating measurement error (quantization noise) dependent on the link length and the attenuation value; and the division of the microwave link into equal smaller intervals determined by the typical size of rain cells in the region (Goldshtein et al 2009).The second, third and final stages in the conversion of the RSL data to rainfall fields involve various geo-statistical methods to combine the error and rainfall measurements to estimate point and gridded rainfall data. Previous studies for similarly distributed networks of links have produced gridded rainfall products at resolutions of 0.775 x 0.755 km (Zinevich et al 2008). Point rain rate estimation by this process allows both the validation by and if deemed appropriate integration of the rain gauge data. Gridded rainfall data can then be input into the hydrological model to determine the ongoing flood response. Additionally a forecast product could be developed that follows on from the work of Joyce et al (2004) that uses propagation vector matrices computed from spatial lag correlations of successive gridded rainfall fields to estimate the advection of convective rain cells. These advected fields could then be input into the flood model to provide more lead time in forecasting any flood responses. A number of off-the-shelf models currently exist that use rainfall estimates for flood modelling. Working closely with organisations such as the Environment Agency, Small and Medium Enterprise (SME), Ambiental who have previously developed software (Flowroute-i) and component methodologies for integrated catchment-scale flood modelling which reduce the time, cost and data input requirements traditionally associated with the catchment scale hydrological / hydraulic modelling. In essence, in this case hydrological and hydraulic processes are modelled in a coupled, end-to-end fashion, using direct rainfall and grid-scale parameterisation of natural losses (e.g. infiltration) and man-made sinks at the flood routing stage to obviate the need for an independent hydrological modelling phase. Results based on independent benchmarking tests show that this approach considerably reduces the time / cost associated with traditional catchment scale hydrological and flood inundation modelling, whilst maintaining a high level of realism and accuracy in terms of flood inundation mapping and downstream flow rate estimation. Validation of the flood model estimates could be performed with hindcast RSL data and using flood delineations from Synthetic Aperture Radar imagery on ERS-1/2 and ENVISAT (Horritt et al 2001) and the early community delineated maps. The derivation of alerts from the model runs could follow a participatory workshop with stakeholders including the community, and those responsible for the humanitarian response. Disseminating Information Phase Communication of early warning information could be carried out using a variety of forms to be derived and agreed during a participatory workshop with stakeholders including the local community (as described above). If and when the system projects that a high rainfall event will occur the system automatically and instantly communicates with affected households. The app sends differentiated warnings to at risk households. For example, for households in a weak structure near a river, the warning might be to evacuate as soon as possible. For households further away, it might only be sufficient to sandbag their homes. The warnings are also personalised according to the household members, with different warnings for households with small children or elderly relatives, telling them to boil some water or to stock up with milk or medication. The efficacy and efficiency of the early warning system would be maximised if each household had a smart phone, but there could also be integration with Short Message Service (SMS) which use lower frequencies than voice calls and so are not subject to attenuation. Further risk communication could be through local radio and via the network of Red Cross volunteers. While mobile phone ownership has grown rapidly in recent years in Senegal a sizeable proportion of the population still has little or no access to this technology. Various forms of the dissemination of the alerts within the community could be explored using such simple methods as coloured flags and audio alerts. Information could also be sent to humanitarian agencies and first responders. This would enable them to provide a more effective response through a virtual triage to decide on priorities and also a personalised response by bringing vaccines, rehydration powders and water to the households affected by contaminated water, shelter for those whose homes had been damaged and medicine for the young, elderly and infirm. After the event, householders would be prompted to send further information about the floods and their impacts, in order to feedback to the meteorologists for service development and model and mapping improvements. Information would also be sent to government and humanitarian agencies to build knowledge on disaster impacts, communication and response and public health. Response Phase Good governance, appropriate action plans, public awareness and education are recognised as key points in effective early warning. The organisation of early warning early action workshops could be held to ensure these important issues are addressed. In the past, these workshops featured a facilitated dialogue between two communities of practice. The first community consisted of the providers of climate services, experts directly involved in producing climate and meteorological forecast information, including forecasters from national hydro-meteorological services (NHMSs), climate modellers from university climate research centres, hydrologists, remote sensing experts and agro-meteorologists. The other community consisted of potential end-users of climate forecasts, including representatives of communities affected by hydro-meteorological disasters, as well as national and sub-national government disaster managers, planners and boundary organizations able to serve as relays of available climate services down to the community level. Participants in workshops would be tasked with defining and jointly agreeing on a plan of action to communicate timely, accurate and actionable early warnings to populations in communities at risk from predictable climate hazards, and to enable climate information access by these populations, beginning with the pilot disaster-prone community whose representatives were in attendance at the workshop. An early warning>early action workshop will be held using the information derived in the previous stages. Deliverables/outputs Specific outputs of this proposal are as follows: Workshop 1 and workshop report. Hosted during the Risk knowledge Phase the report will outline risks, responses and early warning needs. Technical report. Produced from the Monitoring and Predicting Phase it will detail the EWS software and products. An article will be prepared for submission on the technical aspects of this proposal in an internationally refereed journal. Workshop 2 and workshop report. Hosted during the dissemination phase the report will outline the dissemination procedure of early warning information Workshop 3 and final report. Hosted during the response phase the report will outline the outcomes of the monitoring, evaluation and learning. An article will be prepared for submission on the disaster preparedness aspects of this proposal in an internationally refereed journal. Monitoring, evaluation and learning It is anticipated that the monitoring, evaluation and learning system would take place on two levels, looking at the comparative accuracy of rainfall data gathered using the mobile phone system method, and looking at the utility of early warning information provided to communities. As the wider intent of this project is to pilot something that may have benefit in many other contexts and countries, there will also be a significant emphasis on learning; capturing information both qualitatively and quantitatively that will be useful in replicating the system. In terms of monitoring and evaluating the quality of the scientific data, rainfall data derived from the commercial cellular communication network will be compared with that from the raingauges ('sentinel sites') installed within the community during the project, while validation of the flood model estimates will be performed with hindcast RSL data and using flood delineations from Synthetic Aperture Radar imagery on ERS-1/2 and ENVISAT (Horritt et al 2001) and the early community delineated maps. More broadly, the project would aim to intimately involve communities in piloting the early warning system, as the true test of its efficacy will be whether information is timely, and used by communities to react. The project would: Establish a baseline, and a cohort of key informants: Working with local humanitarian organisations potential users of alerts (e.g. Red Cross volunteers in at risk areas, responsible officials and community activists) would be asked to volunteer for the new trial system. These community volunteers will be asked to complete a simple questionnaire on how the current early warning/ early action system works and to trial the app. Monitor the implementation of the new alert system: Using the same cohort enrolled during the baseline exercise, the project will periodically illicit feedback on the usefulness of the new system in terms of speed, accuracy and accessibility, allowing for real time correction of the system. At the same time there will be feedback from householders who use the app to document the impact of flood events. Evaluate whether the alert system made any real difference: either on completion of the project, or following an incident of flooding during the lifetime of the project, the enlisted cohort would be asked to evaluate whether the new alert system worked, and if it did whether it made any difference in terms of response. Where possible secondary data will also be used (for instance in the event of a real flood to determine death and injury, loss and damage and compare to similar incidents). As set out above, the project would also engage stakeholders - primarily communities but also local and national government, NGOs and climate scientists - in a series of workshops to co-design the EWS. This will focus on designing a system that will be used, rather than a perfect technical system that no-body is aware of. The proposed project is primarily concerned with generating evidence in two areas: That more accurate and timely data about rainfall patterns, and particularly about extreme weather events, can be generated from using signal noise in mobile phone networks. That such timely and accurate data can be delivered to communities, local government and local civil society organisations in a way that allows them to prepare and mitigate the potential worst effects of flooding and other weather events. The evidence will be generated through the monitoring and evaluation system set out above. Technical data will be gathered and compared against existing data sets and a handful of project generated sentinel sites. Early warning utility could be collected through the establishment of a cohort of key informants, who will be tracked through the lifetime of the project. A baseline survey could be undertaken to understand current early warning, and then the cohort (representative of users and key stakeholders) could be periodically involved to determine the best ways to deliver EWS information, the format and the frequency in which it can be delivered. Evaluation at the end of the project could help understand whether it has changed the way people are able to plan for flooding, and especially whether there has been any reduction in damage and loss as a result of the project. The combination of quantitative (primarily technical data) and qualitative (primarily user engagement) should provide a powerful insight into the practicability of developing such systems, and the methods by which the delivery of timely and accurate EWS can be improved at low cost. Lessons from the project could be disseminated in a variety of ways, depending on success and the resources available. Firstly, the project could be written as an academic paper, or series of academic studies. Secondly, if the project proves to be successful, then it is anticipated that other companies, donors, organisations, cities and nations will be interested in its replication. At this stage, further donor funding will be sought to develop a series of 'business cases' describing options for further roll out - from open source to commercial. Thirdly, it is intended that regardless of the ultimate success of the project materials will be developed outlining how it was designed, implemented and measured. Again, depending on resources available at this stage it is envisaged that such may be written up as a series of short case studies that can be widely disseminated through the Red Cross and other, similar networks. |
| Sectors | Communities and Social Services/Policy Digital/Communication/Information Technologies (including Software) Environment Government Democracy and Justice |