electricsheepafrica/africa-wash-zambia

---
annotations_creators:
- no-annotation
language_creators:
- found
language:
- en
license: other
multilinguality:
- monolingual
size_categories:
- 1K<n<10K
source_datasets:
- original
task_categories:
- other
task_ids: []
tags:
- africa
- humanitarian
- hdx
- electric-sheep-africa
- geodata
- hxl
- water-sanitation-and-hygiene-wash
- zmb
pretty_name: "Water Point Data Exchange (ZMB)"
dataset_info:
splits:
- name: train
num_examples: 5315
- name: test
num_examples: 1328
---

# Water Point Data Exchange (ZMB)

**Publisher:** Water Point Data Exchange · **Source:** [HDX](https://data.humdata.org/dataset/wpdx_zmb) · **License:** `Creative Commons Attribution Share-Alike` · **Updated:** 2025-12-14

---

## Abstract

WPdx is a platform to compile crowdsourced data focused on rural water points (wells, springs, tapstands) with contributions from governments, NGOs, and researchers. Shared data is cleaned and harmonized using the [WPdx Data Standard](https://www.waterpointdata.org/share-data) to create a robust analysis-ready dataset. There are [two primary datasets](https://www.waterpointdata.org/access-data) available from WPdx. The first is WPdx-Basic, which includes all records shared with the platform. The second, WPdx-Plus is a subset of the Basic dataset which focuses on countries where district and/or national data is available and undergoes additional cleaning. WPdx-Plus is used as an input to the [WPdx Decision Support tools](https://www.waterpointdata.org/use-data/) and is the basis for the datasets posted here. The decision support tools provide insights on rural basic water service access and recommendations regarding prioritized water point repair, highlight areas with apparent service gaps, and identify water points which are at high risk of failure. For each country where WPdx data is available, there are two datasets available for download are described below.

- The wpdx_adm_region_analysis file provides an overview of the population served, unserved and uncharted for each available administrative level. The file also includes a data quality analysis based on water point record age.
- The wpdx_water_points file provides water point level details for each point shared with WPdx including all the WPdx data standard parameters plus results from the WPdx Rehab Priority, Service Gap/New Construction and Status Prediction analyses.
- A data dictionary describing each parameter included in each file is available in the Data and Resources section.
To share data with the WPdx platform, please visit (www.waterpointdata.org/share-data) and/or reach out to [[e-mail]](mailto:[e-mail])

Each row in this dataset represents country-level aggregates. Temporal coverage is indicated by the `report_date`, `created_timestamp` column(s). Geographic scope: **ZMB**.

*Curated into ML-ready Parquet format by [Electric Sheep Africa](https://huggingface.co/electricsheepafrica).*

---

## Dataset Characteristics

| | |
|---|---|
| **Domain** | Demographics and population |
| **Unit of observation** | Country-level aggregates |
| **Rows (total)** | 6,644 |
| **Columns** | 38 (16 numeric, 20 categorical, 2 datetime) |
| **Train split** | 5,315 rows |
| **Test split** | 1,328 rows |
| **Geographic scope** | ZMB |
| **Publisher** | Water Point Data Exchange |
| **HDX last updated** | 2025-12-14 |

---

## Variables

**Geographic** — `lat_deg` (range -17.8885–-8.747), `lon_deg` (range 22.0562–32.8382), `clean_country_id` (ZMB, #country+code+v_iso3), `clean_country_name` (Zambia, #country+name), `activity_id` (range 1.0–835783141.0) and 14 others.

**Temporal** — `report_date`, `created_timestamp`.

**Demographic** — `management_clean`, `usage_cap` (range 50.0–300.0).

**Identifier / Metadata** — `status_id` (Yes, No, Unknown), `source` (SNV-Zambia, Water4, World Vision), `water_source_clean` (Borehole/Tubewell, Piped Water, Protected Well), `dataset_title`, `esa_source` and 1 others.

**Other** — `water_tech_clean` (Hand Pump - India Mark, Public Tapstand, Motorized Pump), `clean_adm1` (Lusaka, Luapula, Central), `clean_adm2` (Chibombo, Chifunabuli, Samfya), `installer` (GRZ, World Vision, WaterAid), `status_clean` and 4 others.

---

## Quick Start

```python
from datasets import load_dataset

ds = load_dataset("electricsheepafrica/africa-wash-zambia")
train = ds["train"].to_pandas()
test = ds["test"].to_pandas()

print(train.shape)
train.head()
```

---

## Schema

| Column | Type | Null % | Range / Sample Values |
|---|---|---|---|
| `lat_deg` | float64 | 0.0% | -17.8885 – -8.747 (mean -13.6636) |
| `lon_deg` | float64 | 0.0% | 22.0562 – 32.8382 (mean 28.3276) |
| `status_id` | object | 0.0% | Yes, No, Unknown |
| `report_date` | datetime64[ns] | 0.0% | |
| `source` | object | 0.0% | SNV-Zambia, Water4, World Vision |
| `water_source_clean` | object | 0.2% | Borehole/Tubewell, Piped Water, Protected Well |
| `water_tech_clean` | object | 10.5% | Hand Pump - India Mark, Public Tapstand, Motorized Pump |
| `clean_country_id` | object | 0.0% | ZMB, #country+code+v_iso3 |
| `clean_country_name` | object | 0.0% | Zambia, #country+name |
| `clean_adm1` | object | 0.0% | Lusaka, Luapula, Central |
| `clean_adm2` | object | 0.0% | Chibombo, Chifunabuli, Samfya |
| `activity_id` | float64 | 64.7% | 1.0 – 835783141.0 (mean 199349881.7678) |
| `wpdx_id` | object | 0.0% | #loc+wpdx+id, 6G3H2H7P+6X2, 6G3H3R44+2JV |
| `install_year` | float64 | 4.1% | 1902.0 – 2025.0 (mean 2012.9584) |
| `installer` | object | 51.2% | GRZ, World Vision, WaterAid |
| `management_clean` | object | 77.3% | |
| `status_clean` | object | 0.0% | |
| `subjective_quality` | object | 64.4% | |
| `local_population` | float64 | 21.9% | 0.0 – 41041.0 (mean 1027.0206) |
| `assigned_population` | float64 | 21.9% | 0.0 – 38511.0 (mean 414.7093) |
| `facility_type` | object | 0.0% | |
| `water_source_category` | object | 0.2% | |
| `water_tech_category` | object | 13.2% | |
| `created_timestamp` | datetime64[ns] | 0.0% | |
| `dataset_title` | object | 0.0% | |
| `usage_cap` | float64 | 0.0% | 50.0 – 300.0 (mean 285.9777) |
| `criticality` | float64 | 23.5% | 0.0232 – 1.0 (mean 0.602) |
| `pressure` | float64 | 23.5% | 0.0067 – 128.37 (mean 1.4771) |
| `distance_to_primary` | float64 | 0.0% | 4.3026 – 239413.2135 (mean 49399.4096) |
| `distance_to_secondary` | float64 | 0.0% | 1.4752 – 150711.1541 (mean 27075.433) |
| `distance_to_tertiary` | float64 | 0.0% | 0.0274 – 73647.6255 (mean 8618.7245) |
| `distance_to_city` | float64 | 7.3% | 566.2263 – 350229.8767 (mean 136674.2026) |
| `distance_to_town` | float64 | 0.0% | 194.175 – 137327.1401 (mean 46497.9163) |
| `is_urban` | object | 0.0% | |
| `days_since_report` | float64 | 0.0% | 1.0 – 5089.0 (mean 2820.8437) |
| `staleness` | float64 | 0.0% | 20.1066 – 99.9685 (mean 46.2766) |
| `esa_source` | object | 0.0% | |
| `esa_processed` | object | 0.0% | |

---

## Numeric Summary

| Column | Min | Max | Mean | Median |
|---|---|---|---|---|
| `lat_deg` | -17.8885 | -8.747 | -13.6636 | -14.7774 |
| `lon_deg` | 22.0562 | 32.8382 | 28.3276 | 28.6927 |
| `activity_id` | 1.0 | 835783141.0 | 199349881.7678 | 148869781.0 |
| `install_year` | 1902.0 | 2025.0 | 2012.9584 | 2014.0 |
| `local_population` | 0.0 | 41041.0 | 1027.0206 | 280.0 |
| `assigned_population` | 0.0 | 38511.0 | 414.7093 | 135.0 |
| `usage_cap` | 50.0 | 300.0 | 285.9777 | 300.0 |
| `criticality` | 0.0232 | 1.0 | 0.602 | 0.5577 |
| `pressure` | 0.0067 | 128.37 | 1.4771 | 0.4733 |
| `distance_to_primary` | 4.3026 | 239413.2135 | 49399.4096 | 36805.3022 |
| `distance_to_secondary` | 1.4752 | 150711.1541 | 27075.433 | 21121.8565 |
| `distance_to_tertiary` | 0.0274 | 73647.6255 | 8618.7245 | 5093.4509 |
| `distance_to_city` | 566.2263 | 350229.8767 | 136674.2026 | 135307.4841 |
| `distance_to_town` | 194.175 | 137327.1401 | 46497.9163 | 41454.336 |
| `days_since_report` | 1.0 | 5089.0 | 2820.8437 | 3398.0 |

---

## Curation

Raw data was downloaded from HDX via the CKAN API and converted to Parquet. Column names were lowercased and standardised to snake_case. Common missing-value markers (`N/A`, `null`, `none`, `-`, `unknown`, `no data`, `#N/A`) were unified to `NaN`. 18 column(s) with >80% missing values were removed: `clean_adm3`, `clean_adm4`, `scheme_id`, `rehab_year`, `rehabilitator`, `pay_clean`.... 18 column(s) were cast from string to numeric or datetime based on parse-success rate (>85% threshold). The dataset was split 80/20 into train and test partitions using a fixed random seed (42) and saved as Snappy-compressed Parquet.

---

## Limitations

- Data originates from Water Point Data Exchange and has not been independently validated by ESA.
- Automated cleaning cannot correct for misreported values, definitional inconsistencies, or sampling bias in the original collection.
- The following columns have >20% missing values and should be treated with caution in modelling: `activity_id`, `installer`, `management_clean`, `subjective_quality`, `local_population`, `assigned_population`, `criticality`, `pressure`.
- Refer to the [original HDX dataset page](https://data.humdata.org/dataset/wpdx_zmb) for the publisher's own methodology notes and caveats.

---

## Citation

```bibtex
@dataset{hdx_africa_wash_zambia,
title = {Water Point Data Exchange (ZMB)},
author = {Water Point Data Exchange},
year = {2025},
url = {https://data.humdata.org/dataset/wpdx_zmb},
note = {Repackaged for machine learning by Electric Sheep Africa (https://huggingface.co/electricsheepafrica)}
}
```

---

*[Electric Sheep Africa](https://huggingface.co/electricsheepafrica) — Africa's ML dataset infrastructure. Lagos, Nigeria.*