Methods: This study employs a geospatial-based model to predict and map out malaria risk area in Kilombero Valley. Environmental factors related to malaria transmission were considered and assigned valuable weights in the Analytic Hierarchy Process (AHP), an online system using a pairwise comparison technique. The malaria hazard map was generated by a weighted overlay of the altitude, slope, curvature, aspect, rainfall distribution, and distance to streams in Geographic Information Systems (GIS). Finally, the risk map was created by overlaying components of malaria risk including hazards, elements at risk, and vulnerability.

Results: The study demonstrates that the majority of the study area falls under the moderate-risk level (61%), followed by the low-risk level (31%), while the high-malaria risk area covers a small area, which occupies only 8% of the total area.

Conclusion: The findings of this study are crucial for developing spatially targeted interventions against malaria transmission in residual transmission settings. Predicted areas prone to malaria risk provide information that will inform decision-makers and policymakers for proper planning, monitoring, and deployment of interventions.

Data acquisition and description

The study employed both primary and secondary data, which were collected from numerous sources based on the input required for the implementation of the predictive model. Data collected includes the locations of all public and private health centers that were downloaded free from the health portal of the United Republic of Tanzania, Ministry of Health, Community Development, Gender, Elderly, and Children, through the universal resource locator (URL) (http://moh.go.tz/hfrportal/). Human population data was collected from the 2012 population housing census (PHC) for the United Republic of Tanzania report.

Rainfall data were obtained from two local offices; Kilombero Agricultural Training and Research Institute (KATRIN) and Kilombero Valley Teak Company (KVTC). These offices collect meteorological data for agricultural purposes. Monthly data from 2012 to 2017 provided from thirteen (13) weather stations. Road and stream network shapefiles were downloaded free from the MapCruzin website via URL (https://mapcruzin.com/free-tanzania-arcgis-maps-shapefiles.htm).

With respect to the size of the study area, five neighboring scenes of the Landsat 8 OLI/TIRS images (path/row: 167/65, 167/66, 167/67, 168/66 and 168/67) were downloaded freely from the United States Geological Survey (USGS) website via URL: http://earthexplorer.usgs.gov. From July to November 2017, the images were selected and downloaded from the USGS Earth Explorer archive based on the lowest amount of cloud cover coverage as viewed from the archive before downloading. Finally, the digital elevation data with a spatial resolution of three arc-seconds (90m by 90m) using WGS 84 datum and the Geographic Coordinate System were downloaded free from the Shuttle Radar Topography Mission (SRTM) via URL (https://dds.cr.usgs.gov/srtm/version2_1/SRTM3/Africa/). Only six tiles that fall in the study area were downloaded, coded tiles as S08E035, S09E035, S10E035, S08E036, S09E036, S10E036, S08E037, S09E037 and S10E037. 

Preparation and Creation of Model Factor Parameters

Creation of Elevation Factor

All six coded tiles were imported into the GIS environment for further analysis. Data management tools, with raster/raster data set/mosaic to new raster feature, were used to join the tiles and form an elevation map layer. Using the spatial analyst tool/reclassify feature, the generated elevation map was then classified into five classes as 109–358, 359–530, 531–747, 748–1017 and >1018 m.a.s.l. and new values were assigned for each class as 1, 2, 3, 4 and 5, respectively, with regards to the relationship with mosquito distribution and malaria risk. Finally, the elevation map based on malaria risk level is levelled as very high, high, moderate, low and very low respectively. 

Creation of Slope Factor

A slope map was created from the generated elevation map layer, using a spatial analysis tool/surface/slope feature. Also, the slope raster layer was further reclassified into five subgroups based on predefined slope classes using standard classification schemes, namely quantiles as 0–0.58, 0.59–2.90, 2.91–6.40, 6.41–14.54 and >14.54. This classification scheme divides the range of attribute values into equal-sized sub-ranges, which allow specifying the number of the intervals while the system determines where the breaks should be. The reclassified slope raster layer subgroups were ranked 1, 2, 3, 4 and 5 according to the degree of suitability for malaria incidence in the locality. To elaborate, the steeper slope values are related to lesser malaria hazards, and the gentler slopes are highly susceptible to malaria incidences. Finally, the slope map based on malaria risk level is leveled as very high, high, moderate, low and very low respectively.

Creation of Curvature Factor

Curvature is another topographical factor that was created from the generated elevation map using the spatial analysis tool/surface/curvature feature. The curvature raster layer was further reclassified into five subgroups based on predefined curvature class. The reclassified curvature raster layer subgroups were ranked to 1, 2, 3, 4 and 5 according to their degree of suitability for malaria occurrence. To explain, this affects the acceleration and deceleration of flow across the surface. A negative value indicates that the surface is upwardly convex, and flow will be decelerated, which is related to being highly susceptible to malaria incidences. A positive profile indicates that the surface is upwardly concave and the flow will be accelerated which is related to a lesser malaria hazard, while a value of zero indicates that the surface is linear and related to a moderate malaria hazard. Lastly, the curvature map based on malaria risk level is leveled as very high, high, moderate, low, and very low respectively.  

Creation of Aspect Factor

As a topographic factor associated with mosquito larval habitat formation, aspect determines the amount of sunlight an area receives. The more sunlight received the stronger the influence on temperature, which may affect mosquito larval survival. The aspect of the study area also was generated from the elevation map using spatial analyst tools/ raster /surface /aspect feature. The aspect raster layer was further reclassified into five subgroups based on predefined aspect class. The reclassified aspect raster layer subgroups were ranked as 1, 2, 3, 4 and 5 according to the degree of suitability for malaria incidence, and new values were re-assigned in order of malaria hazard rating. Finally, the aspect map based on malaria risk level is leveled as very high, high, moderate, low, and very low, respectively. 

Creation of Human Population Distribution Factor

Human population data was used to generate a population distribution map related to malaria occurrence. Kilombero Valley has a total of 42 wards, the data was organized in Ms excel 2016 and imported into the GIS environment for the analysis, Inverse Distance Weighted (IDW) interpolation in the spatial analyst tool was applied to interpolate the population distribution map. The population distribution map was further reclassified into five subgroups based on potential to malaria risk. The reclassified map layer subgroups were ranked according to the vulnerability to malaria incidence in the locality such as areas having high population having the highest vulnerability and the less population having less vulnerable, and the new value was assigned as 1, 2, 3, 4 and 5, and then leveled as very high, high, moderate, low and very low malaria risk level, respectively.

Creation of Proximity to Health Facilities Factor

The distribution of health facilities has a significant impact on the malaria vulnerability of the population dwellings in the Kilombero Valley. The health facility layer was created by computing distance analysis using proximity multiple ring buffer features in spatial analyst tool/multiple ring buffer. Then the map layer was reclassified into five sub-layers such as within (0–5) km, (5.1–10) km, (10.1–20) km, (20.1–50) km and >50km. According to a WHO report, it is indicated that the human population who live nearby or easily accessible to health facilities is less vulnerable to malaria incidence than the ones who are very far from the health facilities due to the distance limitation for the health services. Later on, the new values were assigned as 1, 2, 3, 4 and 5, and then reclassified as very high, high, moderate, low and very low malaria risk levels, respectively. 

Creation of Proximity to Road Network Factor

The distance to the road network is also a significant factor, as it can be used as an estimation of the access to present healthcare facilities in the area. Buffer zones were calculated on the path of the road to determine the effect of the road on malaria prevalence. The road shapefile of the study area was inputted into GIS environment and spatial analyst tools / multiple ring buffer feature were used to generate five buffer zones with the interval of (0–5) km, (5.1–10) km, (10.1–20) km, (20.1–50) km and > 50km regards to the effect of road on malaria occurrence. New values were assigned as 1, 2, 3, 4 and 5, and reclassified as very high, high, moderate, low and very low malaria risk level, respectively.

Creation of Distance from Stream Network Factor

Streams provide water to flood plains and stagnant water that increases the breeding sites for vector diseases that lead to malaria transmission and risk at a locality. Proximity to stream map was created using GIS environment system, spatial analyst tools with the application of multiple ring buffer feature, the feature was set to generate the output layer at different distance intervals such as (0–1) km, (1.1–3) km, (3.1–5) km, (5.1–10) km and > 10km. The reclassification was done based on mosquito flight range and new values were assigned as 1, 2, 3, 4 and 5, and then reclassified as very high, high, moderate, low and very low malaria risk level, respectively.  

Creation of Rainfall Distribution Factor

Rainfall leads to the presence of groundwater as a potential for breeding sites, which increases mosquito and malaria transmission. The data provided in the Excel sheet was analyzed by finding the average rainfall quantity of about 5 years ago. The data was imported into the GIS environment and IDW spatial analyst interpolation tool was applied to interpolate the distribution of rainfall quantity received within the study area to generate a rainfall distribution map. The rainfall distribution map was further reclassified into five subgroups based on the predefined quantity of rainfall received. The reclassified map layer subgroups were ranked accordingly to the degree of suitability for malaria incidence in the locality. For example, areas that receive more quantity of rainfall are related to higher malaria hazards and the less quantity of rainfall have less susceptible to malaria incidences, and the new value was assigned as 1, 2, 3, 4 and 5, and then reclassified values were leveled as very high, high, moderate, low and very low malaria risk levels, respectively.  

Creation of Land Use Land Cover (LULC) Classes Factor

LULC class map layer was generated from Landsat 8 images while undertaking this study. To fit the world reference system of the specific area, Erdas imagine 2014 (Hexagon Geospatial, USA) was used to geo-reference and project the satellite image to the WGS84 UTM zone 37S because the Landsat image is always north projected. Seven bands (Band 1, 2, 3, 4, 5, 6 and 7) in all five scenes were stacked differently using the raster-spectral layer/stack feature into a single multispectral image which contains a higher degree of spectral resolution and five scenes were then mosaicked using the mosaicpro feature. Raster subset image feature was applied in which the study area shapefile and mosaicked image were manipulated and produced an image that fit the study area extent. Raster supervised classification feature was used to perform a supervised classification method for classifying the image into basic classes such as water bodies (stream network), forestry, grassland and mixed feature. The mixed feature class is a class in which both settlements, cultivated land, bushes, bare land and scrubs were merged as one class feature. The features were combined due to the low resolution of the Landsat image that makes these features look the same during the classification process. The classified image was exported into the GIS environment for further ranking and reclassification. Also, the LULC class map was reclassified in four subgroups for their susceptibility and Suitability for malaria risk.