BFA_2010-2025_JMR_v01_M Joint Food Security Monitor - Burkina Faso Mathijs Lomme Bo Pieter Johannes Andrée Zacharey Carmichael Steve Penson 2025-10-20 NADA Joint Food Security Monitor - Burkina Faso Burkina Faso, 45 areas, 2010-01-01 - 2025-09-01, version 2025-10-20 Joint Monitoring Report (JMR) data BFA_2010-2025_JMR_v01_M Lomme, Mathijs Andrée, Bo Pieter Johannes

IPC

FEWS NET

ACLED

FAO

OCHA

WFP

UNICEF

WHO

ACAPS

WorldPop

FEWS NET

 For details on the terms and conditions for usage of the data, please refer to the Terms and Conditions when accessing the microdata. MetadataEditor World Bank's Food Systems 2030 World Bank's Food Systems 2030 TF0C0728 TF0C0828 Data Help Desk Ongoing food security assessement in Burkina Faso Joint Monitoring Report (JMR) team Burkina Faso Joint Monitoring Report Integrated Food Security Phase Classification Threshold modeling FEWS NET Food prices Exchange rates Drought rainfall Conflict Burkina Faso Burkina Faso, down to sub-national level, 13 admin 1 areas and 45 admin 2 areas The sub-national levels follow the COD standard (https://knowledge.base.unocha.org/wiki/spaces/imtoolbox/pages/2557378679/Administrative+Boundaries+COD-AB) Sub-national level, admin 2 Alert levels for each indicator that drives food insecurity in Burkina Faso, where the level can be Typical (no alert raised), Heightened or Critical World Bank Group. (2024). Development Data Quality Policy. https://ppfdocuments.azureedge.net/de65051a-a1ee-410e-aba8-9c302f59be2f.pdf time-series monthly ACLED CONFLICT ACLED (Armed Conflict Location and Event Data), is an independent, impartial, international non-profit organization collecting data on violent conflict and protest in all countries and territories in the world.Link: https://acleddata.com/data-export-tool/ WFP RAINFALL WFP publishes this dataset on HDX, where it contains dekadal rainfall indicators computed from Climate Hazards Group InfraRed Precipitation satellite imagery with insitu Station data (CHIRPS) version 2, aggregated by subnational administrative units.Link: https://data.humdata.org/dataset/bfa-rainfall-subnational WORLD BANK RTFX The World Bank's Real Time Exchange Rates, contains monthly exchange rate estimates by product and market, using FAO and WFP data as input.Link: https://microdata.worldbank.org/index.php/catalog/study/BFA_2023_RTFX_V01_M WORLD BANK RTFP The World Bank's Real Time Food Prices, contains monthly food price estimates by product and market, using FAO and WFP data as input.Link: https://microdata.worldbank.org/index.php/catalog/study/BFA_2021_RTFP_V02_M Data is provided by livelihood zone, which is transformed to admin level 2 areas by using the population percentages in the different livelihood zones. The IPC phases are rounded to the nearest integer adjusted for whether or not humanitarian assistance was available in that admin level 2 area. Data is provided in .tif files on a yearly basis. The gridded data is aggregated to the admin level 2 areas by taking the sum by area. Monthly population numbers are kep constant throughout each year. All events contain a date, a lat/lon location and the number of fatalities. We map the lat/lon locations to an admin level 2 area and we aggregated to a monthly level per admin level 2 area by taking the sum of the total fatalities. Additionally, we keep track of neighbouring conflict by taking the sum of all fatalities in directly neighbouring admin level 2 areas per admin level 2 area. The total rainfall, which is given in mm by admin level 2 area by dekad, is aggregated to a monthly level by taking the sum. Admin level 2 areas with missing data are filled in by taking the average total rainfall of all direct neighbouring admin level 2 areas. Exchange rates are provided on a monthly basis for various markets. These markets are mapped onto admin level 2 areas using their lat/lon coordinates. Admin 2 areas without a market take the average of all markets found in the corresponding admin 1 area. If there are still admin 2 areas with missing exchange rates, we take the average of all neighbouring admin 2 areas. Food prices are provided on a monthly basis for various markets. These markets are mapped onto admin level 2 areas using their lat/lon coordinates. Admin 2 areas without a market take the average of all markets found in the corresponding admin 1 area. If there are still admin 2 areas with missing food prices, we take the average of all neighbouring admin 2 areas. TODO TODO World Bank Microdata Library, JMR Collection 4 per country page The values presented in these datasets are all based on publicly-available data. The datasets are published as open data. Data Help Desk BFA_JMR_data.csv The main dataset, showing for each month/admin level 2 area in Burkina Faso, per indicator, the original value, the indicator value and the alert level 2025-10-20 The indicator value follows from applying a certain method on the original value, which can be found in the BFA_JMR_model_details.csv file. Breaching thresholds on this indicator value result in the given alert levels. BFA_JMR_model_details.csv Details about the indicators used in the JMR model for Burkina Faso 2025-10-20 For all used indicators, this dataset shows the alert thresholds determined by the model and the corresponding scoring metrics. Information about which method was used on the original value to determine the indicator value is listed in the description. BFA_JMR_pcodes.csv The sub-national, admin 2, areas for Burkina Faso 2025-10-20 BFA_JMR_data.csv is pcoded. That dataset links to this dataset through the admin2_pcode variable. This dataset holds all additional information regarding the admin level 2 areas, like name and admin level 1 information. BFA_JMR_country_level.csv Country level wide number of alerts and population living in areas at risk for Burkina Faso 2025-10-20 This dataset is the aggregated version of BFA_JMR_data.csv, where all alerts are summed by indicator and date. Furthermore, the population living in areas at risk is aggregated in order to give a quick country level overview. Country code ISO3C codes, also known as ISO 3166-1 alpha-3 codes, are three-letter country or territory codes that are part of the ISO 3166 international standard. These codes are used to uniquely represent and identify countries and dependent territories in a standardized manner. Each ISO3C code corresponds to a specific country or territory and is often used in various applications, such as international trade, banking, internet domain names, and statistical analysis, to simplify and standardize country and territory references. Country code in ISO 3166-1 alpha-3 format IPC phase cutoff The FEWS NET IPC phases are integers from 1 to 5, where the higher the number, the higher the level of food insecurity. This cutoff value means that the model was trained on values below the high risk cutoff vs values of the cutoff and above. A value of three thus indicates a model that was trained on IPC3+ risk. The FEWS NET IPC phase cutoff for high risk Admin 2 pcode P-codes (Place codes) are unique geographic identification codes, usually represented by combinations of letters and numbers to identify a specific place, point, positional location or feature on a map or within a database. There is only one P-code per administrative unit. The COD-AB standard is used (Common Operational Datasets). Administrative boundary level 2 pcode Year The year variable represents the year associated with each data point, provided in numerical format (e.g., 2023). Month The month variable indicates the month number (1-12) corresponding to each data point, presented in a numerical format. Administrative boundary level 2 pcode Indicator Different indicators are measured against historical known escalations in food security. Examples of possible indicators are: Food prices, Fuel prices, Exchange rate, Conflict, Drought rainfall, Drought NDVI and Drought ASI. The indicator name used in the JMR model Grouping For each indicator, this column groups into: Original value: The raw value of the indicator, before any modeling has occuered. For instance, for the indicator Drought rainfall, the Original value will show the rainfall in mm Indicator value: The value of the indicator after the modeling/transformation has occurred. For instance for the indicator Drought rainfall, the Indicator value could be the Standard Precipitation Index of the total rainfall in mm. The methods that were used for each indicators can be found in the model details datafile. Alert level: The alert level corresponding to the given Indicator value. A 0 corresponds to Typical, a 1 corresponds to Heightened and a 2 corresponds to Critical. The thresholds for each indicator can be found in the model details datafile. For the indicator GLM, which is the combined model indicator, an additional option for grouping is added; Population living in areas at risk, which is an indication of the people living in that specific admin 2 area at the certain date that are at risk of experiencing an escalation in food insecurity. For the indicator Target, which is the outcome all indicators are measured against, four different groupings are presented: Original IPC phase: Categorical integer between 1 and 5, where a higher number indicates a level level of food insecurity, following the IPC phase format. IPC phase excluding HA: The IPC phase adjusted for Humanitarian Assistance (HA). When an area received HA, the idea is that this lowered the IPC phase by 1. For example, an area in IPC phase 3, which received HA, will have an IPC phase excluding HA of 4. Binarized target indicator: The model uses a binary target indicator, where a 1 corresponds to admin 2/date combinations where the IPC phase is above a set threshold (IPC level 3+ or IPC level 4+ depending on the country). A 0 corresponds to all IPC phase clasifications below this cutoff. Possible escalations: This study focusses on escalations in food insecurity, which means an increase in IPC phase from below the cutoff to above the cutoff. To train the model on these escalations, we define admin 2/date combinations where the previous recorded IPC phase was below the IPC phase cutoff as a possible escalations. Categorical description for what the value column is showing Value A numeric value which captures the value corresponding to the selected indicator and grouping. The value corresponding to the indicator/grouping combination Country code ISO3C codes, also known as ISO 3166-1 alpha-3 codes, are three-letter country or territory codes that are part of the ISO 3166 international standard. These codes are used to uniquely represent and identify countries and dependent territories in a standardized manner. Each ISO3C code corresponds to a specific country or territory and is often used in various applications, such as international trade, banking, internet domain names, and statistical analysis, to simplify and standardize country and territory references. Country code in ISO 3166-1 alpha-3 format IPC phase cutoff The FEWS NET IPC phases are integers from 1 to 5, where the higher the number, the higher the level of food insecurity. This cutoff value means that the model was trained on values below the high risk cutoff vs values of the cutoff and above. A value of three thus indicates a model that was trained on IPC3+ risk. The FEWS NET IPC phase cutoff for high risk Indicator Different indicators are measured against historical known escalations in food security. Examples of possible indicators are: Food prices, Fuel prices, Exchange rate, Conflict, Drought rainfall, Drought NDVI and Drought ASI. The indicator name used in the JMR model Alert level For each indicator, two thresholds are set; a heightened and a critical. These thresholds are the outcome of the univariate analysis of each indicator against the binarized target indicator. This modeling aims to minimize a loss function, which is a weighted average between the False Positive Rate and the False Negative Rate. For the heightened risk thresholds a heavier weight is put on minimizing the False Negative Rate and for the critical risk thresholds a heavier weight is put on minimizing the False Positive Rate. The alert level, eiher heightened or critical Threshold A numeric value. In the description field it is made clear that an alert is raised whenever the indicator value goes either above or below the threshold. For instance for the indicator Drought Rainfall, an alert is raised when the indicator value is BELOW the threshold, whereas for most other indicators, alerts are raised when the indicator value goes ABOVE the threshold. The threshold that, if breached and depending on the selected alert level, raises either a heightened risk alert or a critical risk alert Description Holds information on which method with which time-window was used on the indicators original value in order to transform it to the indicator value. For instance for the indicator Food prices, the method used could be Percentag of moving average over the last 6 months. For the indicator Drought Rainfall, the method used could be the Standard Precipitation Index, which time-window 12 months. Description of the indicator Loss metric for escalations in training dataset A value between 0 and 1. The loss function is a weighted average between the False Positive Rate and the False Negative Rate, where a Heightened risk alert puts a heavier weight on minimizing the False Negative Rate and a Critical risk alert puts a heavier weight on minimizing the False Positive Rate. One of the two metrics the model is optimized over, the lower the value, the better the performance. Looks only at the escalations in the target indicator in the training dataset False Positive Rate escalations A value between 0 and 1. A lower rate indicates the model is not raising too many false alerts. The False Positive Rate of the food insecurity escalations in the training dataset False Negative Rate escalations A value between 0 and 1. A lower rate indicates the model is not missing to many actual alerts. The False Negative Rate of the food insecurity escalations in the training dataset Loss metric for the whole training dataset A value between 0 and 1. The loss function is a weighted average between the False Positive Rate and the False Negative Rate, where a Heightened risk alert puts a heavier weight on minimizing the False Negative Rate and a Critical risk alert puts a heavier weight on minimizing the False Positive Rate. One of the two metrics the model is optimized over, the lower the value, the better the performance. Looks at the whole training dataset False Positive Rate all A value between 0 and 1. A lower rate indicates the model is not raising too many false alerts. The False Positive Rate of the food insecurity predictions in the training dataset False Negative Rate all A value between 0 and 1. A lower rate indicates the model is not missing to many actual alerts. The False Negative Rate of the food insecurity predictions in the training dataset Admin 0 pcode P-codes (Place codes) are unique geographic identification codes, usually represented by combinations of letters and numbers to identify a specific place, point, positional location or feature on a map or within a database. There is only one P-code per administrative unit. The COD-AB standard is used (Common Operational Datasets). Administrative boundary level 0 pcode Admin 1 name Following OCHA and United Nations best matching area names Administrative boundary level 1 name Admin 1 pcode P-codes (Place codes) are unique geographic identification codes, usually represented by combinations of letters and numbers to identify a specific place, point, positional location or feature on a map or within a database. There is only one P-code per administrative unit. The COD-AB standard is used (Common Operational Datasets). Administrative boundary level 1 pcode Admin 2 name Following OCHA and United Nations best matching area names Administrative boundary level 2 name Admin 2 pcode P-codes (Place codes) are unique geographic identification codes, usually represented by combinations of letters and numbers to identify a specific place, point, positional location or feature on a map or within a database. There is only one P-code per administrative unit. The COD-AB standard is used (Common Operational Datasets). Administrative boundary level 2 pcode Country Following OCHA and United Nations best matching country names Administrative boundary level 0 name Country code ISO3C codes, also known as ISO 3166-1 alpha-3 codes, are three-letter country or territory codes that are part of the ISO 3166 international standard. These codes are used to uniquely represent and identify countries and dependent territories in a standardized manner. Each ISO3C code corresponds to a specific country or territory and is often used in various applications, such as international trade, banking, internet domain names, and statistical analysis, to simplify and standardize country and territory references. Country code in ISO 3166-1 alpha-3 format Country code ISO3C codes, also known as ISO 3166-1 alpha-3 codes, are three-letter country or territory codes that are part of the ISO 3166 international standard. These codes are used to uniquely represent and identify countries and dependent territories in a standardized manner. Each ISO3C code corresponds to a specific country or territory and is often used in various applications, such as international trade, banking, internet domain names, and statistical analysis, to simplify and standardize country and territory references. Country code in ISO 3166-1 alpha-3 format IPC phase cutoff The FEWS NET IPC phases are integers from 1 to 5, where the higher the number, the higher the level of food insecurity. This cutoff value means that the model was trained on values below the high risk cutoff vs values of the cutoff and above. A value of three thus indicates a model that was trained on IPC3+ risk. The FEWS NET IPC phase cutoff for high risk Year The year variable represents the year associated with each data point, provided in numerical format (e.g., 2023). Month The month variable indicates the month number (1-12) corresponding to each data point, presented in a numerical format. Administrative boundary level 2 pcode Indicator Different indicators are measured against historical known escalations in food security. Examples of possible indicators are: Food prices, Fuel prices, Exchange rate, Conflict, Drought rainfall, Drought NDVI and Drought ASI. The indicator name used in the JMR model Grouping For each indicator, this column groups into: Original value: The raw value of the indicator, before any modeling has occuered. For instance, for the indicator Drought rainfall, the Original value will show the rainfall in mm Indicator value: The value of the indicator after the modeling/transformation has occurred. For instance for the indicator Drought rainfall, the Indicator value could be the Standard Precipitation Index of the total rainfall in mm. The methods that were used for each indicators can be found in the model details datafile. Alert level: The alert level corresponding to the given Indicator value. A 0 corresponds to Typical, a 1 corresponds to Heightened and a 2 corresponds to Critical. The thresholds for each indicator can be found in the model details datafile. For the indicator GLM, which is the combined model indicator, an additional option for grouping is added; Population living in areas at risk, which is an indication of the people living in that specific admin 2 area at the certain date that are at risk of experiencing an escalation in food insecurity. For the indicator Target, which is the outcome all indicators are measured against, four different groupings are presented: Original IPC phase: Categorical integer between 1 and 5, where a higher number indicates a level level of food insecurity, following the IPC phase format. IPC phase excluding HA: The IPC phase adjusted for Humanitarian Assistance (HA). When an area received HA, the idea is that this lowered the IPC phase by 1. For example, an area in IPC phase 3, which received HA, will have an IPC phase excluding HA of 4. Binarized target indicator: The model uses a binary target indicator, where a 1 corresponds to admin 2/date combinations where the IPC phase is above a set threshold (IPC level 3+ or IPC level 4+ depending on the country). A 0 corresponds to all IPC phase clasifications below this cutoff. Possible escalations: This study focusses on escalations in food insecurity, which means an increase in IPC phase from below the cutoff to above the cutoff. To train the model on these escalations, we define admin 2/date combinations where the previous recorded IPC phase was below the IPC phase cutoff as a possible escalations. Categorical description for what the value column is showing Value A numeric value which captures the value corresponding to the selected indicator and grouping. The value corresponding to the indicator/grouping combination