Geospatial Information systems for modeling accessibility to service points in Zimbabwe with reference to health provision centers

doi:10.21203/rs.3.rs-3583427/v1

Download PDF

Research Article

Geospatial Information systems for modeling accessibility to service points in Zimbabwe with reference to health provision centers

https://doi.org/10.21203/rs.3.rs-3583427/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Objectives Understanding geographical accessibility to services is important for informed decision making and policy formulation pertaining to infrastructural development and sustainable disaster management. The aim of this study was to develop an interactive travel time tool that can be used to model accessibility to any destination in Zimbabwe.

Setting We modelled physical accessibility to the health facility network in Zimbabwe using time as a metric measure. The research takes into consideration the various factors that influence the speed of traversing the landscape across different landcover types and terrain-conditions. Modelling was done using SAGA GIS and High Speed Computing.

Results This research was applied specifically on health service centres as destinations, and the model output revealed how certain areas with rugged terrain or dense vegetation and impassable streams can cause a set of villages to be “cut off” from apparently close health facilities as the bird flies.

Conclusions The study successfully demonstrates the application of GIS in accessibility modelling and reflects opportunities for integrating accessibility modelling with service provision and infrastructural planning.

Geographic Information Systems

Multi-hazards

GIS

Remote Sensing

Accessibility modelling

Several studies (Delematre 2012; Rekha et al 2017; (Hierink et al., 2020; Verma and Dash, 2020; Iyer et al., 2020) have examined the disparities in access to health care service as a function of several factors. They observe that inequalities in health care accessibility are unavoidable due to several factors and the extent to which they impact is a combination of the spatial arrangement of the health care delivery system, the location and distribution of the population within a region, and the characteristics of the transportation infrastructure (Delamater et al., 2012). Keyes (2019) identified the important factors that determine geographical (i.e., physical) accessibility of a population to the network of functional health facilities. These factors include road networks and natural barriers (such as rivers, water bodies and flooded areas), terrain, vegetation cover (NDVI) among other factors. Despite all the other factors a well-planned transport network system is key in facilitating the accessibility to and from any destination. Modelling the accessibility to service provision is key to service provision and infrastructure development. Fisher (2016) identifies accessibility as a necessity in evidence-based infrastructure planning as well as effective disaster management. Although, the transport network forms the backbone of the health infrastructure it is often destroyed by climate induced floods and landslides. For instance, Cyclone Idai affected Chimanimani, Chipinge and Masvingo districts of Zimbabwe killing hundreds of people and destroying roads, bridges and settlements. In fact, Chimanimani was cut off from the rest of the country as roads and communication networks were destroyed by the cyclone. These damages impact negatively on accessibility, prompting the need for developing and identifying alternative routes within the shortest possible time. This therefore calls for the development of rapid decision support systems that rides on Geographic Information System for use in modelling the shortest routes that can be taken to access services especially during disasters. Basing on these models, different spatial metric measurements such as time and distance can be used to quantify accessibility. These models allow the processing of visual, dynamic and analytic results. GIS can be used to analyse multiple variables and then interpret these variables to simulate or predict potential events.

Easy access to public service provision such as health centres is an objective to governing bodies and local authorities of any country. Sustainable Development Goal 3 (SDG 3) aims to ensure healthy lives and promote well-being for all, at all ages .This calls for the development of physical infrastructure through which quality health care services are available and accessible to individuals and communities. Information on accessibility to service centres is important for informed decision making and policy formulation. Modelling and understanding geographical accessibility is important to support service provision and rehabilitation of the transport network system. Geo-spatial neighbourhood and iterative cost accumulation algorithms exist that can provide quick and comprehensive information about the impact of infrastructure and terrain on accessibility. Using GIS, fast and dynamic process models that integrate existing knowledge about terrain and landcover variables in the real world can be developed into a set of relationships and equations for quantifying accessibility. These algorithms can be used to model geo-spatial accessibility as an accumulated cost of time with the aim to support strategic planning for infrastructural development and sustainable disaster management. Travel time studies have been applied in modelling accessibility to improve infrastructure development, service provision and disaster management and these are described in detail in the LITERATURE REVIEW chapter of this thesis.

Recently, attempts have been made to model accessibility during Cyclone Idai where road infrastructure was grossly damaged (with above 90% of road networks in Chimanimani and Chipinge damaged), bridges were swept away and damaged and 584 km of roads being damaged by landslides (IFRC, 2019). Key access roads to cyclone hit areas were destroyed and delays in travel time were significant. In the research of (2020 an accessibility model was developed and run on a small scale (Chimanimani district) and the model development was done in ILWIS, because of its strong and well thought out distance calculation using iteration. The results of the ILWIS model were quite good but however, developing a process model that depends on several or different GIS input parameters, requires the model to be run several times as the input parameters vary. This is necessary to see the impact of the change on the model result. Developing an interactive model with ILWIS requires basic coding, which implied more time hence the local model was implemented using the QGIS model developer tool. GIS based cost accumulation algorithms were used to model post disaster accessibility in the affected districts. The implementation was however made on a local scale hence the need to integrate the model to the national scale in order to reveal gaps in accessibility that are faced by the nation at large.

Model simulations involving complex algorithms like the cost iterative processes generate massive volumes of dynamic geo-spatial data and are time consuming. GIS applications, such as time-critical simulation, forecasting, emergency service decision-making, deal with massive volumes of geo-spatial data hence complexity of the processing algorithms and also time-critical demand. The processes therefore need high performance capabilities for data storage, manipulation, retrieval, and processing. Application of the algorithms on the high performance computer requires parallelization of the GIS software or the algorithms in order to increase the speed of the computer calculations and achievement of desired results.

Study Area

The study was carried out in Zimbabwe, a landlocked country located in south-east Africa, and lies on coordinates (Lat-17.829167, 31.052222) the Zambezi and Limpopo Rivers. The country is bordered by South Africa to the south, Botswana to the south-west, Zambia to the north, and Mozambique to the east. The capital city is Harare, and the second largest city is Bulawayo. The population of Zimbabwe was estimated at 15 178 979 with 48 percent being male while 52 percent were female, giving a sex ratio of 92 males for every 100 females. The landscape can be divided into three geographical regions: the inland plateau, the high-veld and the Escarpment. The 1,200 m high inland plateau takes up the majority of the country while the peneplains extend between the cities of Harare and Bulawayo. The plateau falls away in the north and south towards the low-veld.

The three main transport modes that serve the Zimbabwean economy are roads, railways, and aviation. Air is important for tourism leaving road transport as the dominant means of transport. There are more than 103 000 km of classified roads in Zimbabwe, 10 753 km of which are paved and 17 072 km are urban (Fig. 1). About 10% of the mapped roads are paved and these primary roads play a major role of facilitating movement of the country’s population and they link the main economic centres within the country, enabling internal transportation of people and goods. The Urban areas are serviced by urban roads that may or may not be sealed and are managed by urban councils. A little more than 70 percent of the mapped road-network is made up of tertiary feeder and access roads. These are gravel roads that link rural areas to the paved primary road network. In addition to the managed roads people are using tracks along which to travel in-between homesteads and to reach the closest roads. The total length of the mapped tracks at the time of writing was 207 950 km.

Map of Zimbabwe

Data-preparation

The Fig. 2 shows the workflow according to which the geo-spatial analysis was conducted, illustrating the input data required for determining travel time as well as the output based on the disaster intervention decisions and Infrastructure development decisions.

The first step in the workflow was to pre-process the geo-spatial data used for the geo-spatial analysis. Several geo-spatial circumstances contribute to determine how fast a person can travel across a specific section of the landscape. These are the land-cover present (roads, vegetation cover, or open water, and the slope of the terrain. Accessibility to healthy service was the main focus of the research hence health centre locations were used as destinations points, while each individual pixel of the rest of the country is representing the source locations. Data for creating the traversability grid were therefore collated and pre-processed as outlined below.

Traverse speed grid input parameters:

Roads of Zimbabwe

The geo-spatial accessibility model was developed to cover the whole of Zimbabwe. The main overland transport modes that serve the Zimbabwean economy are roads, and tracks. In this research the roads vector data was extracted from OpenStreetMap. The roads were classified into four road classes’ i.e. paved/sealed roads, urban roads, gravel roads and tracks (Fig. 3. The OpenStreetMap roads of “Trunk”, “Primary”, and “Secondary” are generally paved in Zimbabwe, while the “Tertiary” and “Road” classes delineate gravel roads. The “Tracks” and “Paths” of OpenStreetMap where grouped to represent one class of tracks, which are not constructed roads but are routes used for local travel. A similar classification was also adopted by the Zimbabwe National Road condition Survey of 2016/17. During the 2016/17 survey, roads were classified according to road use and surfacing condition, however they did not include tracks.

Slope

The elevation data was derived from the SRTM30 DEM. “SRTM30 is a near-global digital elevation model (DEM) comprising a combination of data from the Shuttle Radar Topography Mission, flown in February 2000 and the U.S. Geological Survey's GTOPO30 data set”.

The slope (Fig. 4) was calculated from the digital elevation model as slope in degrees, using the Pythagoras theorem, where no change in elevation over distance is 0⁰ and the maximum slope recorded was 78°.

Rivers

The same SRTM30 data was used to delineate the river channel network for Zimbabwe, in using resources at the ZCHPC. The Strahler function of TauDEM Version 57 was used to delineate the river channels and a Strahler order of 5 was specifically used define the smallest streams. The waterways delineated from the hydrological model are classified according to the Strahler ordering system.

Open water

The open water was mapped using a classification algorithm based on Sentinel-2 satellite imagery of 2021 for the Zimbabwe National Wetlands Inventory (Fig. 5). The water raster was geo-referenced to the Strahler order extent and pixel size.

Vegetation density

Vegetation density was derived using the Normalized Difference Vegetation Index (NDVI), which is widely used to map green vegetation biomass. The NDVI was calculated from Sentinel-2 imagery (of 2022) using the Google Earth Engine (GEE) platform (Fig. 6).

Bridges and Impassable Bridges

Where roads cross rivers there are in general bridges and drifts. However, there are also reports for impassable sections of the roads, where the bridge or drift is destroyed or broken and the road is impassable during the rainy season. The bridges were captured as serviceable bridges and those that needed repair, or those that were destroyed, were separated into another point layer and referred to as impassable.

Health-facilities

The destination dataset representing hospitals (Fig. 7) and clinics was obtained from OpenStreetMap data, the OCHA archive, as well as the ministry of health and child welfare (MOHCW).

Traversability

Preceding data pre-processing was the geo-spatial analysis consist of coding, weighting, and aggregating the raster layers into one layer that represents the traversability speed of each pixel in the raster covering the whole of Zimbabwe. First each landcover, in the different data layers, was given a CODE, which was ranked according to the landcover that takes precedence in influencing the traversability (Columns 1 and 2 in Table 1). The higher the data code the higher the precedence. Each landcover was converted into a raster layer where all pixels with that landcover were assigned a specific code (CODE column in Table 1). Then the maximum code was determined for all the landcover code rasters, cell-by-cell basis, such that for example where a road was crossing a river the code of the road would be assigned to that cell (the maximum code).Finally the lookup table (LUT) was used to convert the maximum code layer into a weight layer, representing the number of seconds it would take to traverse each coded pixel in the raster, as in the reclassification table (Table 1, last column). This last column gives travel speed values for each land cover class calculated as the number of seconds to travel across one grid cell (30m). The formula:

[Travel time = (speed/distance) x 3600],

was used to convert the Travel time in km/hr to seconds per 30 meters. Travel time values were calculated as the number of seconds to travel across one 30m grid cell.

In the scripted version of the information system the LUT is integrated into the model workflow in such a way that travel speed values can be altered after each run of the model to produce new output scenarios based on differing input parameters. All the parameters to the landcover grids were defined in the LUT except the slope. This is because the slope pixels were weighted with reference to Tobler’s hiking function. Tobler’s hiking function determines the average walking speeds on different slopes. In this case we assumed an off-road walking speed of 2 km/hr (equal to 54 seconds dwell-time in one pixel) and each increase by one degree in slope increases the dwell-time in one pixel by 6 seconds. However, where slope values of greater than 30 degrees were considered impassable and were assigned a “Barrier” or no-data value.

The following conditional statement was used in a map-calculator to combine the LUT-derived travel-speed-grid with the slope-grid as follows:

h = ifelse(a > 9999,0/0,ifelse(a > 0,a,ifelse(gt(b,30),0/0,ifelse(𝑊𝑒𝑖𝑔h𝑡 b < 0,0/0,(6*b) + 54)))),

where a is the output from the lookup table, b is the slope in degrees. The function 0/0 is used to assigns a no-data value to the resulting pixel.

With this the whole country now has weight-factors, which reflect the relative decrease in speed in less accessible terrains compared to the best accessible terrain. This is the traversability raster.

Table 1

LUT showing land cover classes, travel speed and travel time.
IMPEDANCE FACTOR	CODE	Travel speed (Km/hr)	Travel time (sec/pixel)
Strahler order 5	105	1	108
Strahler order 6	106	1	108
Strahler order 7	107	Barrier	99999
Strahler order 8	108	Barrier	99999
Strahler order 9	109	Barrier	99999
Strahler order 10	110	Barrier	99999
Strahler order 11	111	Barrier	99999
Strahler order 12	112	Barrier	99999
Strahler order 13	113	Barrier	99999
Strahler order 14	114	Barrier	99999
Open Water	201	Barrier	99999
Forest	202	Barrier	99999
Tracks (non-maintained routes)	300	5	21.6
Sealed (wide tarred roads)	301	120	0.9
Urban(roads within urban areas)	302	40	2.7
Gravel(roads other than wide tarred)	303	80	1.35
Intact Bridges	304	100	1.08
Destroyed bridges	305	Barrier	99999

Cost-accumulation

The traversability raster was then used as a weight factor in the cost-accumulation iteration, starting from the destination points that are provided. The cost-accumulation creates a new grid with each cell attributed the time it takes to travel to the nearest destination in seconds.

Travel-time unit of Measure

In the last step the travel-time is converted from seconds to hours or minutes as a unit of measure.

Parallelizing the algorithms

The raster-based geo-spatial cost accumulation is implemented in may GIS software, and we found it implemented in ILWIS, R, GRASS GIS, SAGA GIS, and Google Earth Engine. However, in literature we did not find a reference to any of them being parallelized. In this research we developed a solution where the cost-accumulation iteration can be run with data parallelism. At the time of the research the High-Performance Computer of the ZCHPC has about 280 processing nodes with 12 cores each. The PBS Script was developed and used to submit the travel-time processing algorithm to the High-Performance Computer (HPC). With this script the resources were requested, and the processing requests were queued. The PBS could then schedule the jobs to be executed in parallel and required no further input from the user. In other words, the cost accumulation calculation was divided into several jobs that could be submitted as a batch and distributed to several nodes on the HPC. The output tiles were then merged by running geo-spatial computations and remove the edge effect between tiles. We tested whether the speed of computing the national output was improved using the parallel computing script by running the national algorithm on one node and then running it in parallel. The different computing durations were recorded and are being presented in the results section.

Validating the model output

The model output was validated against real experienced travel times. Rural District Councils and individuals were interviewed so as to collect the real experience of travel-times at sampled sites on the ground. The survey form captured relevant data such as areas known to have gaps in health service provision and travel accessibility to services. The questionnaire also captured individual responses on the estimated travel times that they take to reach to the nearest hospital from their rural and urban areas. The results from the interview survey were compared to those of the accessibility modelling using GIS.

Travel-time to the nearest health facility

The results of the travel time derived from the geo-spatial model illustrate that travel time (hours) to the nearest health facility from any location in the Zimbabwe varies across space. It can be observed that the travel time is generally less in the central parts of the country than towards the borders. The highest travel time is observed in the southern, western, and northern parts of the country. Interestingly these areas coincide with protected areas such as Gonarezhou National Park, Hwange National Park, Matusadonha National Park and Nyanga National Park (Fig. 8).

Parallelizing the algorithms

Results of this study demonstrated that running the accessibility model for the whole country using a personal computer took 8 days. However, using the parallelized script on HPC this reduced the same process to 15 minutes. This is more than a thousand times faster than running the process on any other computer thereby making the model suitable for use in disaster situation.

Validating the model output

Figure 10 shows the modelled travel-time (y-axis) versus the experienced travel time at locations across Zimbabwe. It can be observed that there is a very close relationship between the travel time estimated by the model and what was observed on the ground with an R squared value of 0.89.

This study analyzed physical accessibility to an existing health facility network considering several factors that influence the speed of traversing the landscape across different landcover types and terrain-conditions resulting in travel times to health facilities across Zimbabwe. The research on the Zimbabwe geo-spatial accessibility model to health centres took into consideration the presence of geographic barriers to accessibility such as vegetation, terrain, river network channel. The Zimbabwe Geo-spatial Accessibility Model is flexible and allows integration of variable input parameters to assess different accessibility scenarios. This makes the model ready for implementation for infrastructure planning, management and in the disaster management cycle at both local and national scales. Dynamic models are desirable over static models because they allow simulation of scenarios, an important aspect when planning for the what-if scenarios. The results of the Zimbabwe geo-spatial accessibility model showed areas with significant gaps in accessibility, with locals travelling distances of nearly 6 hours to access the nearest health centre. Areas to the western border, the

North-western border and the southern border recorded significant gaps in accessibility and service provision.

Throughout the national scope, the combined influence of different input parameters contributes significantly to the increase in travel hours in most areas. The areas in white represent no-data values and these areas are completely impassable. These no-data areas present barriers to accessibility which could either be high NDVI values representing thick forest, large inland water bodies or stream classes of Strahler order greater than 5, as well as very steep slopes, as explained in the lookup table. As evidenced from the modelling, we can generally conclude that health service in Zimbabwe is not equally spatially accessible to all populations because some individuals must travel more than 3 hours to get access to the nearest health centre.

The most time-consuming step in the travel-time algorithm is the cost-accumulation iteration. It is a geo-spatial iteration that continues until the travel-time for each pixel in the raster is calculated starting off from the destination points, in this case the health facilities. In this algorithm, each next cost accumulation iteration is dependent on the result of the previous iteration. This makes parallelization of computations not trivial. Several cost-accumulation implementations were analyzed, and these include SAGA GIS, R, GRASS and ILWIS. However, all of this software except for R had no parallelized solution for the cost-accumulation iteration. R could run effectively parallel on the HPC but however the R movecost algorithm only calculates a slope dependent cost accumulation. The result obtained when using slope as the only barrier to accessibility is however biased because it ignores other barriers to accessibility. Our solution to parallelize the cost accumulation reduced the computation time from several days to only 15 minutes for the whole of Zimbabwe.

The edge effect error occurred due to the fact that the study area was defined by borders. Distortions occurred because areas close to the boundary could have been classified as having poor geographic access to health service provision even though they may in fact be proximate to resources across the boundary. In solving this error, we executed a second run of the algorithm with the second run of the algorithm being offset to half the distance of the neighbouring tiles so as to make sure all destinations close to the boundary would be included in the cost accumulation. The minimum travel distance between neighbouring tiles was then computed using results from the two model runs. The minimum calculation removed the edge effect impact from the data.

The overall response rate from the questionnaires survey was above 90%. Questionnaire reported travel times were compared to GIS estimated travel times as a way of evaluating the model. For the observed travel time, the assumption made was patients would go directly to health centres making the method comparable to the GIS estimated travel times which did not include any delays in travelling to the next health centre. The statistical analysis presented a moderately strong and statistically significant correspondence between the reported ravel times presented from the survey and estimated travel times derived from the GIS modelling. While we found a high degree of correlation between travel times estimated by the GIS model and as reported from the survey questionnaire, GIS estimates were significantly lower in value compared to observed data. This is possible because models include a lot of generalizations and assumptions hence GIS models may perform better or worse in different conditions based on infrastructure conditions, geographic topology, and various other factors. The Association of Rural District Councils (ARDC) was also interviewed, and they counted a notable number of wards that were lacking behind in health service provision. The association counted 16 notable wards (Nyaminyami, Gokwe, Binga, Makonde, Beitbridge, Gwanda, Matobo, Nkayi, Mbire, Muzarabani, Chimanimani, UMP, Mudzi, Tongogara, Shurugwi and Vungu, Mwenezi, Chiredzi). The result from the Geo-spatial accessibility model shows strong correspondence with data provided by the association. Travelling to the nearest health centre in almost all the singled-out areas takes approximately more than 5 hours as evidenced from the map.

In addition we recommend improving the model by including more-and-more detailed attributes that influence travel-time. For example, the road-condition can be made more variable in the model.

Relevance of the model to national development

This information system can be customized to address any other specific service provision and monitoring problem, for example the destination points could be the schools, to identify areas where new schools need to be built. Figure 11 shows areas lacking in service provision whereby schools that are more than 4km away from health points were modelled. Pupils at Ruchanyu primary and secondary schools in Shurugwi (Fig. 12) are situated far away from health points. Another example could be to plan for agricultural produce collection and processing centres, such for example in the fresh milk processing chain.

By updating the input geo-spatial data, such as the mapped transport infrastructure (including in near real time when connections are suddenly interrupted), as well as the destination points and their attributes, the information system will also become more precise.

This research was applied specifically on health service centres as destinations, and the model output revealed how certain areas with rugged terrain or dense vegetation and impassable streams can cause a set of villages to be “cut off” from apparently close health facilities as the bird flies.

The model utilizes and stitches together several pre-existing GIS algorithms in-order to interactively derive the final product of travel time or travel cost in hours. The study successfully demonstrates the application of GIS in accessibility modelling and reflects opportunities for integrating accessibility modelling with service provision and infrastructural planning. These research findings are important for tailoring informed policy formulation in infrastructural development sustainable disaster management. However the implementation of the travel algorithm further requires the linking the model with a geo-database and an android based application model. This is a necessary step for the effective and meaningful use of the model in disaster management and infrastructure development. Environmental parameters are not static and are continuously changing, a geo-database will allow the collection, storage and running of differing travel scenarios on the accessibility model.

Almost all research projects that generate modelled data at global scales rely on assumptions, generalizations, and the use of best-available datasets which might even be sub-optimal (Weiss et al., 2018). Even though the accessibility model we developed assumes that travel time is affected by factors none other than infrastructural development and the type of terrain, it can still be expanded to include many attributes to the type of data used in the modelling. One example could be to also consider the road attributes such as road curvature, stop signs, etc., which also have the effect of reducing speed and hence impact negatively on the total accumulated travel time. Wealth status is also a likely determinant of whether someone travels on foot rather than taking a vehicle and thus substantially affects accessibility on the level of the individual. The choice of health centre is not only a function of distance but can also be a choice of quality, a factor not considered in the travel algorithm. Edge effect was a notable problem in the first algorithm run. The effect of edge censoring occurs when the study boundary affects the estimation of a measure and can induce biases which will affect inferences made on the measures (Van Meter et al., 2010). The nearest health centre may lie outside the study area for locations near the study boundary and thus introducing bias into the spatial measure of distance to the nearest health centre. However this was dealt with accordingly, by running the model twice with the second model run having an offset of 0.5 from the tile boundaries.

Next to data uncertainties, travel scenarios present a source of uncertainty as assumptions on travel speeds and modes are were uniformly generalized in the development of the model. In the study travel speeds were generalized across all ages and sexes etc., but however travelling speeds vary with age, sex, type of travelling mode, time of the day among many other factors. Notably travelling speeds vary differently according to the time of travel. With four wheeled vehicles, travelling speeds may increase during the night due to de-congested roads while an individual travelling on foot during the night may have his or her speed generally decreased due to visibility. Age and sex also have direct impacts on travel time with travel speeds of the feminine sex comparatively lower than of those of the opposite sex. In future and as more detailed data becomes available this can be included in the geo-spatial interactive travel-time model.

The most immediate beneficiaries to this model are local authorities, which can use it to identify areas that need more transport infrastructure and health faculties. The other stakeholders that will benefit from this model are the health sector and the disaster response sector.

Therefore this model should be made accessible to the stakeholders by installing ICT services that link a geo-spatial database of facilities and transport infrastructure to the high-performance computer for rerunning the model at any time.

Further to this study this model can also be applied and validated for any other distributed service accessibility, such as for example schools, police station, churches, and even banking services.

Funding

We, the authors, declare that no funds, grants, or other support were received during the preparation of this manuscript

Competing Interests

All the authors for this work have no relevant financial or non-financial interests to disclose.

Author Contributions

All the three authors contributed to the study conception and design, methodology. Material preparation, data collection and analysis were performed by Theresa Mazhawidza. The first draft of the manuscript was written by Theresa Mazhawidza and all authors Theresa Mazhawidza, Murwira and Munyaradzi Davis Shekede revised the first version of the manuscript to come up with the final manuscript. Accordingly, all the authors read and approved the final manuscript submitted.

Delamater, P.L. et al. (2012) ‘Measuring geographic access to health care: raster and network-based methods’, International Journal of Health Geographics, 11, pp. 1–18. doi:10.1186/1476-072X-11-15.
Hierink, F. et al. (2020) ‘Modelling geographical accessibility to support disaster response and rehabilitation of a healthcare system: An impact analysis of Cyclones Idai and Kenneth in Mozambique’, BMJ Open, 10(11). doi:10.1136/bmjopen-2020-039138.
Huang, F. et al. (2011) ‘Preliminary study of a cluster-based open-source parallel GID based on the GRASS GIS’, International Journal of Digital Earth, 4(5), pp. 402–420. doi:10.1080/17538947.2010.543954.
Iyer, H.S. et al. (2020) ‘Geospatial evaluation of trade-offs between equity in physical access to healthcare and health systems efficiency’, BMJ Global Health, 5(10). doi:10.1136/bmjgh-2020-003493.
Li, Z. et al. (2020) ‘Introduction to big data computing for geospatial applications’, ISPRS International Journal of Geo-Information, 9(8). doi:10.3390/ijgi9080487.
Van Meter, E.M. et al. (2010) ‘An evaluation of edge effects in nutritional accessibility and availability measures: a simulation study’, International Journal of Health Geographics, 9, pp. 1–12. doi:10.1186/1476-072X-9-40.
QGIS (2022) ‘QGIS Desktop 3.16 User Guide QGIS Project’, p. 1427.
Rudolfson, N. et al. (2020) ‘Validating the Global Surgery Geographical Accessibility Indicator: Differences in Modeled Versus Patient-Reported Travel Times’, World Journal of Surgery, 44(7), pp. 2123–2130. doi:10.1007/s00268-020-05480-8.
Stojanovic, N. and Stojanovic, D. (2013) ‘High-performance computing in GIS: Techniques and applications’, International Journal of Reasoning-based Intelligent Systems, 5(1), pp. 42–49. doi:10.1504/IJRIS.2013.055126.
Verma, V.R. and Dash, U. (2020) ‘Geographical accessibility and spatial coverage modelling of public health care network in rural and remote India’, PLoS ONE, 15(10 October), pp. 1–23. doi:10.1371/journal.pone.0239326.
Weiss, D.J. et al. (2018) ‘A global map of travel time to cities to assess inequalities in accessibility in 2015’, Nature, 553(7688), pp. 333–336. doi:10.1038/nature25181.

Download PDF

Version 1

posted

You are reading this latest preprint version

Geospatial Information systems for modeling accessibility to service points in Zimbabwe with reference to health provision centers

Status:

Version 1

Abstract

Figures

INTRODUCTION

MATERIALS AND METHODS

Study Area

Map of Zimbabwe

Data-preparation

Traverse speed grid input parameters:

Roads of Zimbabwe

Slope

Rivers

Open water

Vegetation density

Bridges and Impassable Bridges

Health-facilities

Traversability

Cost-accumulation

Travel-time unit of Measure

Parallelizing the algorithms

Validating the model output

RESULTS

Travel-time to the nearest health facility

Parallelizing the algorithms

Validating the model output

DISCUSSION

Relevance of the model to national development

Conclusion

Declarations

Funding

Competing Interests

Author Contributions

References

Status:

Version 1