This is a shortening of the guidance document provided by the Global Taskforce of Cholera Control (GTFCC), specifically for medium- or high-cholera burden countries, available here.

An example application of this approach can be found in the literature in Kiama et al (2025).

Audience: Primarily WASH and health clusters located in medium-/high-burden countries that need to quickly identify priority geographies where multisectoral interventions must be deployed.
Deploy time: Depends on the mechanism. Once data is acquired, an Excel tool (available on the GTFCC  website) can aid in data creation.  
Prerequisites: Have available all national cholera program (NCP) operational geographic units (Excel spreadsheet and spreadsheets), population data (COD-PS preferred, although gridded population estimates can be utilized as needed); case and screening data within the last 5-15 years (preferred)

Data & Tools:

• Basic Datasets
  • List of NCP operational geographic units (most recent)
  • Geographic units in geospatial vector data format for geographic information system (e.g., shapefile)
  • Population (most recent annual; preferred to use COD-PS, but can use gridded datasets if COD-PS is not available)
• Priority Index Calculation
  • Number of reported cholera cases (suspected and tested positive) (weekly)
  • Number of reported cholera deaths (suspected and tested positive) (weekly)
  • Number of reported suspected cholera cases tested for cholera (regardless of the testing method) (weekly)
  • Number of reported suspected cholera cases tested positive for cholera (weekly)

Steps:

1. Collate all of the data for the priority index:

   1. Make sure you address the following steps before calculating the priority index:

      1. Address missing data

      2. Determine appropriate cholera test positivity indicator

      3. Calculate individual indicators, before the overall priority index

2. Calculate each of the four component subindicators individually for each NCP operational geographic unit: incidence, mortality, persistence, and cholera test positivity. These indicators are derived from epidemiologic and cholera testing data over the analysis period.

   1. First, calculate the three main epidemiologic indicators: incidence, mortality, and persistence:

      1. Incidence rate in an NCP operational geographic unit = the total number of cholera cases (including suspected cases and cases tested positive) reported in the unit over the analysis period / (the cumulative person-time (i.e., the sum of population of the geographic unit for each year over the analysis period)) * 100,000

         1. This indicator is the number of cholera cases reported per 100,000 person-years over the analysis period.

      2. Mortality rate in an NCP operational geographic unit = the total number of deaths attributed to cholera reported in the unit over the analysis period / (the cumulative person-time (i.e., the sum of the annual population over the period)) *100,000

         1. This indicator is the number of deaths attributed to cholera reported per 100,000 person-years in the unit over the analysis period.

      3. Cholera persistence in an NCP operational geographic unit = the number of weeks with at least one reported suspected cholera case over the analysis period / the total number of weeks over the analysis period

         1. This indicator is the percentage of weeks with at least one reported suspected cholera case in the unit over the period of interest.

3. Determine whether to incorporate testing data in your PAMI index. To do so, you first calculate representativeness for testing, followed by the overall testing prevalence.

   1. Weekly testing coverage (i.e., representativeness in testing) for cholera in an NCP operational geographic unit: the number of weeks with at least one reported suspected cholera case tested for cholera (regardless of the testing method and of the result) over the analysis period / the number of weeks with at least one reported suspected cholera case over the analysis period –

      1. If this is relatively representative, you then can proceed with positivity rate

      2. This indicator is the percentage of weeks with at least one suspected case tested for cholera among weeks with at least one suspected case reported in the unit over the analysis period.

   2. Use this indicator to then determine if you calculate the positivity rate or the number of years of cases, following this flow-chart:
      

   3. The following two indicators may be utilized here:
      1. Positivity rate (preferred indicator if representativeness allows) in an NCP operational geographic unit = the number of reported suspected cholera cases tested positive for cholera over the analysis period (regardless of the testing method) / the number of reported suspected cholera cases tested for cholera over the analysis period (regardless of the testing method).
         1. This indicator is the cholera positivity rate in the unit over the analysis period.
      2. Number of years with case(s) tested positive
         1. This indicator is the number of years with at least one case tested positive for cholera (regardless of the testing method) reported in the NCP operational geographic unit over the analysis period.

4. Calculate the 50th and 80th percentiles of each epidemiologic component subindicator’s distribution across all of the NCPs.
   1. Once these two percentiles are calculated for each indicator, each of the NCP operational geographic units’ values for each indicator should be compared to these percentiles to determine the score category:

   Epidemiologic Indicator	Score
   	0 points	1 point	2 points	3 points
   Incidence	No case	> 0 and < median (50th percentile)	≥ median (50th percentile) and < 80th percentile	≥ 80th percentile
   Mortality	No death	> 0 and < median (50th percentile)	≥ median (50th percentile) and < 80th percentile	≥ 80th percentile
   Persistance	No case	> 0 and < median (50th percentile)	≥ median (50th percentile) and < 80th percentile	≥ 80th percentile

5. Calculate the priority index for each NCP operational geographic unit by summing the scores of the component subindicators as follows:

   Priority index = incidence score + mortality score + persistence score + cholera test positivity score (if applicable)

   To aid in implementation, you can use the Excel tool created by the Global Taskforce on Cholera Control (GTFCC)

6. Join the spreadsheet of your priority index to the shapefile(s) of interest, using your GIS tool of choice (e.g., QGIS, ArcGIS). Then, map that shapefile of interest.

7. Conduct a stakeholder validation workshop for each NCP operational geographic unit. The exact setup of this workshop is outside of the scope of this how-to – please refer to the guidance document linked above for more information.

Outputs: Priority index, by each operational NCP geographic unit, and a map of all priority indices by each NCP geographic unit.

Validation & QA: Ensure that you check for missing data (review the linked guidance document for additional information)

Maintenance: Ensure that calculation is done every 5 years, whenever a new NCP dataset is done. As mentioned in the PAMI documentation, “as a general principle, PAMI analysis should be updated when a new version of an NCP is developed (typically every five years). Earlier updates may be considered if there are significant changes in [cholera epidemiology].”

Risks & Considerations: Ensure that missing data is sufficiently dealt with before creating the index.

This is a shortening of the guidance document provided by the Global Taskforce of Cholera Control (GTFCC), specifically for no- or low-burden countries, available PAMIs for cholera elimination for no-/low-burden countries (i.e., vulnerability index documentation)

An example application of this approach can be found in the literature in Kiama et al (2025).

Audience: Primarily WASH and health clusters located in no-/low-burden countries that need to quickly identify priority geographies where multisectoral interventions must be deployed.
Deploy time: Depends on the mechanism. Once data is acquired, an Excel tool can aid in data creation.  
Prerequisites: Have available all national cholera program (NCP) operational geographic units (Excel spreadsheet and spreadsheets), population data (COD-PS preferred, although gridded population estimates can be utilized as needed); vulnerability assessment factors within the last 5-15 years (preferred)

Data & Tools:

• Basic datasets:
  • List of NCP operational geographic units (most recent)
  • Geographic units in geospatial vector data format for geographic information system (e.g., shapefile)
  • Population (most recent annual; preferred to use COD-PS, but can use gridded datasets if COD-PS is not available)
• Vulnerability Assessment (data changes according to context, so a subset of data is used here)
  • Location adjacent to cross-border cholera-affected areas or identified PAMIs
  • Location along major travel routes with transportation hubs
  • Major population gatherings
  • Areas with high population density or overcrowded settings (e.g., urban slums, refugees/IDP camps)
  • Areas with high-risk populations (e.g., seasonal workers/fishermen/miners in informal settlements)
  • Hard-to-access populations
  • Population received oral cholera vaccine more than three years ago (two-doses campaign with a coverage for both round >70%)
  • Areas at high-risk for extreme climate and weather conditions (e.g., heavy rains, floods, droughts)
  • Areas affected by complex humanitarian emergencies
  • Areas with more than 30% of the population using unimproved water facility type
  • Areas with more than 50% of the population using unimproved sanitation facility type

Steps:

1. Collate all of the data for the vulnerability index. Only utilize the indicators that are necessary to your geography of interest.

   1. Make sure you address the following steps before calculating the vulnerability index:

      1. Address missing data, specifically:

         1. Missing information for a measurable vulnerability indicator for most NCP operational geographic units.
         2. Missing information for a vulnerability indicator at the NCP operational geographic unit level.
         3. Incomplete (or of suboptimal quality) information for a vulnerability indicator for some NCP operational geographic units.

      2. Calculate individual indicators, before the overall vulnerability index

2. Score each of your subindicators with a presence index (1/0), for each NCP, according to the following metrics:

   Generic vulnerability factor	Example of measurable indicator for PAMI identification	Scoring Criterion
   		0	1
   Confirmed cholera imported case(s) in the NCP operational geographic unit considered	NCP unit with at least one confirmed cholera outbreak reported over the past five year/analysis period	No	Yes
   Cross-border areas adjacent to frequently cholera-affected areas or identified PAMIs in neighbouring country(ies)	NCP unit with at least one confirmed cholera case imported (from another country or another NCP operational geographic unit reported during the analysis period), or Cross-border NCP unit adjacent to cross-border areas frequently affected by cholera outbreaks or classified as PAMI in neighbouring countries	No	Yes
   Location along major travel routes with transportation hubs	NCP unit located along transportation pathway(s) with transportation hub(s)	No	Yes
   Major population gatherings	NCP unit hosting major population gathering(s)	No	Yes
   High population density locations or overcrowded settings	NCP unit with high population density or overcrowded settings	No	Yes
   High-risk populations	NCP unit with high-risk population	No	Yes
   Hard-to-access populations	NCP unit with hard-to-access population	No	Yes
   Population that received oral cholera vaccine (OCV) more than 3 years ago	NCP unit with a population vaccinated more than three years ago (two-doses OCV campaign with a coverage for both round >70%)	No	Yes
   High risk for extreme climate and weather conditions	NCP unit exposed to extreme climate and weather conditions	No	Yes
   Complex humanitarian emergency	NCP unit located in an area under a complex humanitarian emergency	No	Yes
   Unimproved water	NCP unit with more than 30% of the population using unimproved water facility type (calculated as sum of % of population with unimproved service level and % of population using surface water) OR more than 15% of the population using surface water	Does not meet any of the two criteria	Meets one or more criteria
   Unimproved sanitation	NCP unit with more than 50% of the population using unimproved sanitation facility type (calculated as % of population with unimproved sanitation service level and % of population practicing open defecation) OR more than 30% of the population practicing open defecation	Does not meet any of the two criteria	Meets one or more criteria
   Limited access to hygeine	NCP unit with more than 50% of the population with no handwashing facility on premises	Does not meet the criteria	Meets the criteria

3. Calculate the vulnerability index for each NCP operational geographic unit by summing the scores of the component subindicators. These can be done in an unweighted format (i.e., where all vulnerability factors are considered equally), or in a weighted format (i.e., where a particular vulnerability factor is (or factors are) emphasized compared to other factors, usually by multiplying it by a larger number (e.g., multiplying one indicator by 3 if it is thought to be 3x more 'predictive' of the indicator)).

   Unweighted

   Vulnerability index = ∑(0/1 subindicators)

   Weighted

   Vulnerability index = ∑(0/1 subindicators*weighting factor))

   To aid in implementation, you can use the Excel tool created by the Global Taskforce on Cholera Control (GTFCC)

4. Join the spreadsheet of your vulnerability index to the shapefile(s) of interest, using your GIS tool of choice (e.g., QGIS, ArcGIS). Then, map that shapefile of interest.

5. Conduct a stakeholder validation workshop for each NCP operational geographic unit. The exact setup of this workshop is outside of the scope of this how-to – please refer to your guidance document(s) for information.

Outputs: Vulnerability index, by each operational NCP geographic unit, and a map of all vulnerability indices by each NCP geographic unit.

Validation & QA: Ensure that you check for missing data (review the guidance document as appropriate)

Maintenance: Ensure that calculation is done every 5 years, whenever a new NCP dataset is done. As mentioned in the PAMI documentation, “as a general principle, PAMI analysis should be updated when a new version of an NCP is developed (typically every five years). Earlier updates may be considered if there are significant changes in [vulnerability factors].”

Risks & Considerations: Ensure that missing data is sufficiently dealt with before creating the index.

Audience: GIS analysts with access to risk surfaces and geospatial raster and vector datasets (e.g., climatological, elevation, hydrological)
Deploy time: Download of data should take ~1 day, deployment should take 1 day
Prerequisites: Have basic familiarity with usage and implementation of rasters and vector datasets; have basic familiarity with GIS-related programs

Data & Tools:

• GIS Platform: ArcGIS and/or QGIS
• Access to modeled risk surface of a disease (e.g., Malaria Atlas Project for malaria (which covers prevalence rate (P. falciparum rate)) or risk index/hotspot (e.g., PAMI, for cholera)
• Access to geospatial datasets of interest. These can include climatological datasets, elevation datasets, hydrological datasets - review the below for more information.)

Steps:

1. Acquire modeled prevalence surface(s) of interest.
   1. For the malaria prevalence surfaces from Malaria Atlas Project:
      1. Navigate to https://data.malariaatlas.org/, then on the left-hand side, click on “Trends”. On the resulting panel, click on the “Pixel” option. You should be able to see the Incidence Rate (e.g., Incidence Rate Pf).
      2. From this window, click on the “Incidence Rate Rasters” button located underneath the map. This will help you download a prevalence raster.
   2. For cholera, we suggest using the PAMI existing hotspot identification (i.e., the “Priority Areas for Multisectoral Interventions (PAMI) Identification” use cases mentioned above) to highlight areas of prevalence/cholera presence. As a result, prevalence here is not going to be mapped using a gridded surface, but using prevalence by NCP areas.
      1. Instead of redefining risk classifications for cholera, satellite data and tools should be used to complement PAMI’s existing hotspot identification. While some examples can include identifying areas at high risk for extreme weather events as part of PAMI’s optional cholera vulnerability assessment, updating risk scores between PAMI updates, or monitoring dynamic environmental conditions in real time, we are incorporating the identification of areas at high risk for weather events.
   3. For other types of disease risk, please either:
      1. Search for terms such as “[disease] risk map surface” in your preferred option of choice (for raster datasets), or
      2. Search for any type of epidemiologic index (for vector datasets).
2. Acquire geospatial datasets of interest, depending on the disease case being analyzed.
   1. Depending on the disease that you are interested in focusing upon, please search for the following types of datasets:

1. Once all datasets are downloaded to your system*, or acquired on Google Earth Engine*, clip all available datasets down to the shapefile within your GIS program.
   1. For low-connectivity conditions, it is suggested to use the following workflow:
      1. Download raster files (e.g., TIFs) of each dataset of interest for your geography of interest, for 7-14 days before any outbreak is identified.
      2. Ensure that all of the datasets of interest are in the same coordinate reference system (CRS)/projection. This will depend on your GIS program of interest.
      3. Suggested: resample all of your datasets so that they fit the highest resolution dataset.
      4. On your GIS program of interest, find the bounding box (i.e., the extent of the shapefile boundaries). This bounding box should be used to clip each of your datasets. Again, the program used will affect how you find this bounding box, or you can use a standalone platform (e.g., bboxfinder).
      5.  Clip each of the layers to the bounding box extent.
2. Add all clipped layers to your GIS program’s interface, along with your previously-created prevalence surface and/or hotspots.

Outputs: Map of overlaid risk estimates on top of relevant geospatial datasets

Validation & QA: Ensure that all datasets are available for the right spatial extents and temporal periods of interest, and that all raster datasets are at the same resolution.

Maintenance: All datasets must be temporally and spatially relevant. Make sure that all of your climate and hydrology datasets are of the time period(s) of interest. DEMs and LULCs are less likely to be updated in a timely manner, so find the most relevant dataset for your purposes.

Risks & Considerations: As the emergency situation progresses, ensure that all climate and hydrology datasets are updated. Downloading datasets on a daily basis and substituting in datasets in your maps is suggested.

No-Code Analysis: Threshold Calculation

Audience: Surveillance Epidemiologists
Deploy time: <1 day for data acquisition, ~1-3 hours for calculations (tops)
Prerequisites: Knowledge of Microsoft Excel or another spreadsheet application; ability to get access to historical case datasets.

Data & Tools:

• Microsoft Excel (for threshold calculations)
• Shapefiles and/or GIS tool (if you want to map the thresholds)

Steps:

1. Acquire all historical case datasets as baseline datasets
   1. It is advised to use at least 1-3 years’ worth of datasets, focusing on the same time period from previous years (e.g., epidemiological week).
2. To calculate the simple historical average:
   1. Compute the mean (=AVERAGE(range_of_interest)) and standard deviation (=STDEV.P(range_of_interest)) of your distribution
   2. Using these means and averages, you can calculate the typical surveillace results:
      1. Alert threshold  = mean + 1*SD
      2. Outbreak threshold = mean + 2*SD
   3. If your number of current cases for the time period, rounded to whole numbers, exceeds the alert or outbreak threshold, you can treat this as an alert or outbreak
3. If using an early warning style threshold (e.g., WHO’s EWARN), you can work with the following definitions:
   1. A single case threshold indicates that a disease case happening in a non-native area is an outbreak alert.
   2. Calculating the doubling threshold (where the number of cases has doubled compared to the previous week), and assessing the doubling time (i.e., how long it takes for the number of cases to double) can help show outbreak speed.
      1. The faster the doubling time has occurred, the more quickly an intervention should be presented.
4. Regardless of the threshold used, you can categorize for each area of interest if the number of cases exceeds the alert or outbreak threshold.
5. If you want to map this information in a GIS platform:
   1. Join your calculated spreadsheet to your shapefiles of interest.
   2. Use a symbology that can indicate a yes/no binarization

Outputs: Alert and/or outbreak thresholds

Validation & QA: Ensure that you have the correct datasets for the correct time periods and spatial areas of interest.

Maintenance: Update these on a weekly basis, or as data is released out into the public. Make sure to keep track of both past and future occurrences of the thresholds, to see if a location has started to exceed a known threshold.

Risks & Considerations: Ensure that the same time period (across multiple years) and the same spatial areas are represented in your baseline datasets.

Audience: Surveillance epidemiologists, data/decision makers, staff in WASH or humanitarian clusters that need to collapse environmental factors with epidemiological datasets.
Deploy time: <1 day for data collation, 1-3 hours for risk score calculation and mapping
Prerequisites: Knowledge of Excel or a spreadsheet program; knowledge of a GIS program (if you need to map the risk score)

Data & Tools:

• Case dataset for your indicator of interest (summarized to a certain administrative boundary)
• Geospatial clipped datasets of interest, as shown above (raster or vector)
• Excel spreadsheet or spreadsheet tool of interest
• GIS software
• Shapefiles for your areas of interest (i.e., same admin boundaries)

Steps:

1. In contrast to the PAMI risk scores shown above, you can use these scores to quantify a combination of healthcare capacity, disease burden, and exposure to the pathogen of interest, instead of calculating these individually.
2. Determine your variables to include in the composite risk score. Some variables can include:

1. Ensure that all datasets, if geospatial in nature, are in the same projection as your admin boundary of interest.
2. If you are aggregating a raster dataset down into an administrative boundary of interest, we suggest performing a zonal statistics procedure. This can find an average or a sum (depending on the conditions given to the procedure) of your raster dataset as it overlays with your administrative boundary. (Note that the exact syntax for this procedure depends on which GIS platform is used.)
   1. The intended output for a zonal statistic is a tabular output, where you have your administrative area and the average of your dataset of interest within your administrative boundary of interest. This procedure can be repeated for every single variable of interest, such as:

1. If you want to create a risk score in an Excel spreadsheet format:
   1. Export your dataset as a tabular output (i.e., TXT or CSV file) once you have the zonal statistics conducted for all variables of interest
   2. In an Excel spreadsheet, you then have to normalize each of the variables to encompass a range of [0,1].
      1. Unlike the PAMI scores, you are not working directly with binary variables (i.e., 0/1), but continuous, bounded variables that are between 0 and 1.
   3. You can normalize the datasets using either:
      1. Min-max scaling: x_scaled = (x-x_min) / (x_max - x_min)
      2. Z-score standardization: z = (x - \mu)/(sigma)

For the above example, the two scaling approaches are shown for the first variable:

1. 1. Determine the weights of each of the normalized indicators. As shown in the PAMI example, the indicator can either be equally weighted (i.e., weight is 1 for each indicator), driven by expert knowledge (e.g., if one factor is two times as likely to contribute to something, multiply that factor by 2), or
   2. Generally, a composite risk score (combining disease burden, exposure, vulnerability, transmission potential, and healthcare capacity) is calculated using a weighted average, along the following formula:

EPI_i=∑k=1nwkZikEPI_i = \sum_{k=1}^{n} w_k Z_{ik}EPIi​=k=1∑n​wk​Zik​

where:

iii = spatial unit (county, grid cell, district)

kkk = risk factor

wkw_kwk​ = weight

ZikZ_{ik}Zik​ = standardized variable

For our hypothetical example here:

* For this worked example, (0.2 weight is given to var1, 0.8 weight given to var2)

5. 5. To map the disease risk score created in an Excel spreadsheet, join the spreadsheet of your composite index to your shapefile of interest in your GIS program of interest
6. To directly create the index in your GIS program of interest:
   5. In your shapefile of interest (where you have already performed the zonal statistics procedure(s) mentioned above):
   6. Turn on the editing feature in your GIS program (e.g., ArcGIS Pro directions; QGIS directions) to allow yourself to add additional columns outside of the algorithm.
   7. Using your GIS program's Field Calculator, perform each of the normalization, weighting, and summarization steps shown above in the Excel spreadsheet directly in your shapefile.
      5. The tabular backend will look very similar to the Excel spreadsheet of interest, although you will be editing the attributes of the shapefile directly (i.e., not affecting the geometry).
   8. At the end, you will have a shapefile with the index of interest, that can directly be mapped in the GIS program of interest.
   9. If using this approach, make sure to export your shapefiles to keep your information.

Outputs: A disease risk score for every geographic area of interest (either in a spreadsheet format only (if in Excel, not joined to a GIS program), or a shapefile with indices attached..

Validation & QA: Ensure that you handle areas with missing data with either an imputation of the average or eliminate those areas if most variables are not collected in that area.

Maintenance: Update these risk scores on a daily basis, or as you have specific factors change (e.g., climate, humidity).

Risks & Considerations: Ensure that you consult with your stakeholders to better understand what variables are most relevant and least relevant for your use case.

Audience: Surveillance epidemiologists (if GIS-savvy), data managers, or individuals with heavy data skills
Deploy time: 1 day to acquire datasets, 1 day to produce results
Prerequisites: GIS tool of interest, case-specific data with dates incorporated

Data & Tools:

• GIS platform of interest (ArcGIS Pro, QGIS)
• Case data of interest, including date
  • If capturing number of cases in a particular time period, the date that should be recorded is the date of data collection.

Steps:

1. Data & Tools:

   • Case data, aggregated to shapefile(s) of interest.
     • Points are preferred for this particular case, but if aggregated data exists, try to use the smallest level of aggregation.
   • GIS program of interest or coding program of interest.

   Steps:

   1. This mechanism creates a prevalence raster, that spatially models on x, y, and date. However, it does NOT capture the relative effect of other covariates.

      1. If you want to have the relative effect of other covariates, ensure that you're using the Map Calculator function shown above.
      2. There are a multitude of functions that can be used to create these surfaces.
         1. Available programs:
            1. R-INLA Project
            2. PyMC
            3. Stan
      3. This is showing the kriging procedure (i.e., Gaussian process model), but other modeling approaches (e.g., Bayesian spatiotemporal model) can also be utilized.
         1. Kriging/Gaussian process models are relatively 'fast' to create, compared to the Bayesian spatiotemporal model, which is why it is being shown here.
   2. Acquire your case data of interest, making sure that it is organized in the following format:

1. Bring your spreadsheet of coordinates into your GIS program of interest. Make sure to plot the spreadsheet as points, with a basemap underneath, to ensure that points are showing up in the correct locations.
   1. Save the coordinates as a shapefile in WGS84.
   2. Afterwards, make sure to reproject your coordinates into your coordinate reference system of interest.

2. Regardless of the coding platform used, the following steps can be conducted for a basic kriging interpolation for a single time period of interest:

   1. Create a spatial grid of some known resolution (e.g., Create a Grid in ArcGIS Pro; Create a Grid in QGIS).

   2. Assess your points of interest, and create a grouping variable (e.g., month, year, or epi week) to 'aggregate' the points to a particular time period of interest.
   3. Aggregate the points to that particular time period, so that all cases within a particular time period are 'separated' out.
      1. In QGIS, this is the Aggregate tool, while in ArcGIS Pro, you can filter or subset out points based on your categorical variable before using their Aggregate tool.
      2. Generally, it may be safer to filter out, and export, points by each time period of interest.
   4. After the aggregation/filtering is conducted, you can then create a basic ordinary kriging (Gaussian Process) surface using your GIS program of interest.
      1. In QGIS, this video is a worked example.
      2. In ArcGIS Pro, follow this documentation to learn more.
   5. The output from this procedure will be a spatial interpolation raster. Therefore, this procedure will have to be conducted for each time period of interest, to track progress through time.
3. Alternatively, you can use the space-time cube functionalities available in newer versions of your GIS program.
   1. In ArcGIS Pro, use the Create Space Time Cube By Aggregating Points (Space Time Pattern Mining Tools) tool to create your space time cube.
   2. In QGIS, you could download the Space Time Cube plugin to create your raster of interest.
   3. The output for this particular procedure will be a netCDF "space-time" cube. Review this to learn more.

Outputs: If using ordinary kriging, you will have a series of rasters that show examples by week. If using the Space-Time cube functionalities, A stack of rasters ("space time cube")

Validation & QA: Ensure that your data is clean and that you have aggregated data to the format that you are interested in using it.

Maintenance: Conduct this every time a dataset is updated.

Risks & Considerations: This may be harder to interpret. Depending on the procedure used, there may be statistical parameters that need to be utilized with which the data analyst may not be immediately familiar.

Audience: Surveillance epidemiologists (if GIS-savvy), data managers, or individuals with heavy data skills
Deploy time: 1 day to acquire datasets, 1 day to produce results
Prerequisites: Use of a GIS program, including Raster Calculator Functionalities

Data & Tools:

• All geospatial datasets of interest
• Case data
  • Points are preferred for this particular case, but if aggregated data exists, try to use the smallest level of aggregation.
• GIS program of interest

Steps:

1. The creation of a spatial risk surface is very similar in the algebra compared to the risk score mentioned above. However, instead of the risk score being aggregated down to different administrative units, you are using data purely in the raster format, which requires Map Algebra and Raster Calculator functionalities.

   1. This mechanism should be used if you want to incorporate additional rasters/factors into your dataset (e.g., climate data). If you simply want to create a prevalence raster and not consider the relative effect of other datasets, utilize the other  prevalence raster mapping procedure highlighted below.
2. Acquire all of your geospatial datasets of interest.
   1. Make sure to standardize the spatial extents of all of your geospatial datasets.
   2. Place all geospatial datasets into the same coordinate reference system, focusing on equidistant projections.
      1. It is suggested to look at a comparison site (such as this one)  to determine which equidistant projection is best for your use case, then use an EPSG database (such as this one) to understand which projection fits your geography of interest.
         1. These EPSG strings can be used in the GIS program!
   3. For raster datasets, ensure that the spatial resolution is consistent between each of the datasets. You can resample up (i.e., to the resolution of the file with the highest resolution) or resample down (i.e., to the resolution of the file with the lowest resolution).
3. Transform non-spatial datasets into spatial datasets.
   1.  For point datasets (e.g., locations of cases):
      1. First plot your locations (x/y coordinates) in your GIS program of choice, in the WGS84 datum.
      2. Reproject your coordinates into the preferred coordinate reference system.
      3. Rasterize your point dataset. This can be done either using the Point-to-Raster functionality (ArcGIS Pro) or Rasterize (QGIS/GDAL).
         1. Make sure that the raster resolution matches the resolution of your other raster datasets.
   2. For linear datasets (e.g., Euclidean location of facility to nearest road)
      1. Plot out the point vector data (e.g., location of facility) and the line vector data (e.g., road networks).
      2. Using either the Near tool (ArcGIS Pro) or Shortest line between features tool (QGIS), calculate the Euclidean distance from each point to the nearest line segment of interest. Your input features will be the point dataset.
      3. Once the tool is done running, the Euclidean distance will be attached to each point.
      4. Rasterize your point dataset. This can be done either using the Point-to-Raster functionality (ArcGIS Pro) or Rasterize (QGIS/GDAL).
         1. Make sure that the raster resolution matches the resolution of your other raster datasets!

4. Using each of your raster datasets, you then will create a normalized version of each raster input.

   1. The easiest version to create is the min-max scaled dataset, but you can also use z-scaling using the same approach.

   2. First, identify your metric of interest, and right click on the raster of interest to get the appropriate values:

      1. Min-Max scaling will need the minimum and maximum values in the raster.
      2. Z-score scaling will need the mean (average) value and the standard deviation.
   3. In the Raster Calculator tool (QGIS, ArcGIS Pro directions), you would set up the scaling for that particular dataset using one of the following formulas:
      1. Min-Max scaling = "YourRasterName" - MinimumValue ) / ( MaximumValue - MinimumValue )
      2. Z-score scaling = ("YourRasterName"-Mean) / (Stdev)
         1. Mean and Stdev can also be calculated directy in Raster Calculator, but it's usually easier to use the values that come directly from the summary statistics of your raster (i.e., right clicking on your raster of interest).
      3. Make sure that you do not create "on-the-fly procedures" - it's easier to work with files that have been directly created and saved in your hard drive.

5. Perform this normalization procedure for each raster dataset of interest.

6. Determine the weighting of each dataset, as shown in the risk score examples shown above.

7. Once all of the raster datasets have been normalized, you will then use a new instance of Raster Calculator in your GIS program of interest. However, instead of performing calculations within a raster, you will be performing calculations between the rasters.

   1. Make sure that you have all of your rasters in the same spatial resolution (grid cell size) and same spatial extent.

   2. As seen in the formula from the risk score, you'll be creating a raster using the formula:composite_score = (weight1*"MyNormalizedRaster1") + (weight2*"MyNormalizedRaster2") + ... + (weightn*"MyNormalizedRastern")This is also known as "map algebra".
8. Let the Raster Calculator do its processing. The smaller the cell size or the larger the area, the longer the Raster Calculator will take to create an export raster.
   
9. Map the raster surface on your GIS program of interest.
   1. Double-check the created raster spatial extent and the spatial resolution when it comes out!
10. Export the raster surface, if desired, into a TIF file.

Outputs: A disease risk raster surface

Validation & QA: Ensure that all datasets are in the right spatial resolution and extents. Make sure that you are typing all formulas correctly into the Raster Calculator feature in your GIS program. Ideally, you also want to check your predictions afterwards using an AUC curve, but this would be more necessary for more extensive procedures (e.g., XGBoost, random forest).

Maintenance: To be performed as new datasets are incorporated in your datasets of interest. A best practice is to create all of the static normalized rasters at one time, and only update the dynamic normalized rasters as the data is updated.

Risks & Considerations: Ensure that you use the correct rasters. Note that this map algebra procedure is a basic risk raster, but is the most explainable type of risk raster that can be calculated. Other procedures, such as random forest or XGBoost, can be utilized to create more extensive rasters, but they are less predictable and potentially involve the user knowing their risks.