Description

Technical info

Dataset

The dataset consists of files with administrative units geometries and one file with HFID variables. The geometries are found (1) in the GADM geometries with distinct administrative levels 0,1 and 2 (GADM.zip), (2) in the geopackage file with combined geometries at relevant administrative level with exact geometries (hfid_all_geom.gpkg) or simplified geometries (simplified_hfid_geom.gpkg). The GADM.zip file contains the shapefiles with administrative level 0 (gadm_410_L0), level 1 (gadm_410_L1) and administrative level 2 (gadm_410_L2) reference geometries.

The data table with the HFID variables (HFID_hv1.csv) is composed of several columns:

• year_month: indicates the year and month of the record in the format "%Y-%m".
• ADMIN_0: name of the administrative level 0 (country) as defined in the GADM.
• ADMIN_1: name of the administrative level 1 (region) as defined in the GADM.
• ADMIN_2: name of the administrative level 2 (sub-region) as defined in the GADM.
• ipc_phase_fews: FEWS NET-IPC compatible phase classification (phase-FEWS variable) if available, else NaN. Possible values: 1,2,3,4,5.
• ha_fews: 1 if humanitarian aid has been distributed (owing to FEWS NET source), else NaN.
• ipc_phase_ipcch: IPC/CH Phase classification (phase-IPC/CH variable) if available, else NaN. Possible values: 1,2,3,4,5,6.
• ha_ipcch: 1 if humanitarian aid has been distributed (owing to IPC API source), else NaN.
• set_ipcch: 1 if HouseHold Group or Internally Displaced People or Urban settlements are found in the area (owing to
IPC API source), else NaN.
• rfg_ipcch: 1 if refugees are found in the area (owing to IPC API source), else NaN.
• fcs_lit: monthly average of population prevalence of insufficient food consumption score from the WFP-LIT source (FCS-LIT variable) if available, else NaN. Possible values: between 0 and 1.
• rcsi_lit monthly average of population prevalence of crisis or above reduced Coping Strategy Index from the WFP-LIT source (FCS-LIT variable) if available, else NaN. Possible values: between 0 and 1.
• fcs_rt mean: monthly average of population prevalence of insufficient food consumption score from the WFP-RT source (FCS-RT variable) if available, else NaN. Possible values: between 0 and 1.
• fcs_rt max: monthly maximum of population prevalence of insufficient food consumption score from the WFP-RT sourceif available, else NaN. Possible values: between 0 and 1.
• fcs_rt min: monthly minimum of population prevalence of insufficient food consumption score from the WFP-RT source if available, else NaN. Possible values: between 0 and 1.
• rcsi_rt mean: monthly average of population prevalence of crisis or above reduced Coping Strategy Index from the WFP-RT source (FCS-RT variable) if available, else NaN. Possible values: between 0 and 1.
• rcsi_rt max: monthly maximum of population prevalence of crisis or above reduced Coping Strategy Index from the WFP-RT source if available, else NaN. Possible values: between 0 and 1.
• rcsi_rt min: monthly minimum of population prevalence of crisis or above reduced Coping Strategy Index from the WFP-RT source if available, else NaN. Possible values: between 0 and 1.
• iso2: ISO 3166-1 code for country in two letters.
• iso3: ISO 3166-1 code for country in three letters.
• region: region name from the United Nations Geoscheme.

Code

The Python codebase to derive this datsaset is found in hfid-code-v1.1.1.zip.

The code is reproducible and modulable in order to add future records, on a monthly or yearly basis, from existing sources and potential other sources of interest, like the population prevalence of answers to surveys enabling single food insecurity indicator computation from the FAO Data in Emergencies. The HFID, structured in a tabular format, facilitates its integration into data-driven and machine learning analyses. Researchers and analysts can conveniently aggregate selected covariates at various levels of granularity, ranging from admin 2 to country level, suiting the specific needs of their studies. This adaptability makes the HFID an instrumental dataset for advanced analytical approaches in food insecurity.

We also provide with some functions automatically reading and processing the HFID dataset, in module targets.utils.utils.py:
• the function read_hfid() reads the HFID tabular dataset (HFID_hv1.csv) enabling the consideration that the "iso2" column with the "NA" value represents "Namibia" country and not a NaN.
• the function read_admins() reads all the shapefiles with specified administrative level reference geometries (GADM.zip). The processing includes the suppression of specific sub regions where inter-national conflicts of identity exist (e.g. China, India, Pakistan, Kosovo).
• the function sanitize_high_phases() allows to decide on the treatment to apply to IPC/CH Phase 6 (keep as such, remove it, or cap it to Phase 5).
• the function add_admins2_geometry_to_hfid() which combines the HFID tabular dataset and the administrative level geometries.

Finally, we propose some vizualisation functions, enabling to plot maps of the variables of the HFID dataset, in module targets.utils.plot_utils.py:
• the function plot_phases_map() plots a map with phases from FEWS NET or IPC/CH (variables phase-FEWS, phase-IPC/CH) at a certain date (year_month), respecting the IPC color code.
• the function plot_outcomes_indc_map() plots a map with values from population prevalence of insufficient food consumption indicators from WFP-LIT and WFP-RT (variables FCS-LIT, rCSI-LIT, FCS-RT, rCSI-RT) at a certain date (year_month).