wellswatch 0.1.0

Access and visualize groundwater-status predictions for monitoring wells in Greece.

WellsWatch

WellsWatch is a lightweight Python library for retrieving published groundwater-status predictions from Zenodo and visualizing red and non-red monitoring wells across Greece.

It is designed for researchers, water-resource managers, hydrogeologists, and other users who need a simple and reproducible way to access and map the national groundwater prediction dataset.

WellsWatch does not train a machine-learning model or generate new predictions. It retrieves and visualizes predictions that have already been published on Zenodo.

---

Why WellsWatch?

Groundwater observations and model outputs are often distributed across multiple files and monitoring locations. WellsWatch provides a simple Python interface that:

• downloads the requested prediction file directly from Zenodo;
• validates the expected data structure;
• extracts the prediction date from the filename;
• distinguishes red and non_red monitoring wells;
• creates a national map of Greece;
• saves the generated map as a high-resolution image when requested.

The main workflow requires only two lines of Python:

from wellswatch import show_map

show_map(2025)

---

Example maps and hydroclimatic context

The two maps below illustrate groundwater-status predictions under contrasting hydroclimatic conditions.

1 August 2015 — wet year

The year 2015 was a comparatively wet year within the 2013–2025 series. The map shows a larger proportion of non-red wells than in the recent dry years, although red-state conditions were still predicted at many monitoring stations.

from wellswatch import show_map

show_map(2015)

1 August 2025 — dry year

The year 2025 was a comparatively dry year, following the very dry conditions observed in 2024. The map is dominated by red wells, indicating widespread deeper-than-normal groundwater conditions relative to the historical reference used by the classifier.

from wellswatch import show_map

show_map(2025)

In both maps:

• Red wells represent monitoring stations classified as being in a red groundwater state.
• Non-red wells represent monitoring stations not classified as red.
• WellsWatch uses the published class label directly and does not recalculate the classification from the probability column.

Predictions from 2013 onward and annual precipitation

The figure below summarizes the annual percentages of wells predicted as red and non-red from 2013 onward, together with mean annual precipitation and its standard error of the mean (SEM).

Across the series, wetter years generally show a larger percentage of non-red wells, whereas drier years tend to show a larger percentage of red wells. The contrast is particularly clear between wet 2015 and dry 2025. This figure provides a descriptive comparison of groundwater-status predictions and precipitation; it should not, by itself, be interpreted as evidence of a causal relationship.

---

Installation

After publication on PyPI:

pip install wellswatch

For local development from the GitHub repository:

git clone https://github.com/Mil-afk/wellswatch.git
cd wellswatch
python -m venv .venv

Activate the virtual environment on Windows:

.\.venv\Scripts\Activate.ps1

Install the package in editable mode:

python -m pip install -e ".[dev]"

---

Quick start

Display a map

from wellswatch import show_map

show_map(2025)

The current Zenodo record contains prediction files for 1 August of each year from 2013 to 2025. WellsWatch is designed to provide maps for all available years from 2013 onward, as new annual prediction files are added to the Zenodo record. Therefore:

show_map(2025)

is equivalent to:

show_map(2025, month=8, day=1)

Save a map

from wellswatch import show_map

show_map(
    2025,
    save_path="wellswatch_2025.png",
)

Change marker size

from wellswatch import show_map

show_map(
    2025,
    marker_size=20,
)

Load prediction data without plotting

from wellswatch import load_prediction_from_zenodo

data = load_prediction_from_zenodo(2025)

print(data.head())
print(data["realtime_pred_class_name"].value_counts())

Inspect the available Zenodo files

from wellswatch import list_zenodo_files

for filename in list_zenodo_files():
    print(filename)

Read a local prediction file

from wellswatch import load_prediction_file

data = load_prediction_file(
    "13_pred_2025_08_01_with_predictions_01.xlsx"
)

---

Returned data

load_prediction_from_zenodo() and load_prediction_file() return a pandas DataFrame.

The prediction files contain the following columns:

Column	Description
row_id	Row identifier
station_code	Groundwater monitoring-station code
lon	Longitude in decimal degrees
lat	Latitude in decimal degrees
realtime_pred_class_name	Published class label: red or non_red
realtime_pred_red_probability	Predicted probability of red-state classification

WellsWatch adds:

Column	Description
prediction_date	Date extracted automatically from the filename

The returned DataFrame also includes source metadata in data.attrs:

print(data.attrs)

Example:

{
    "source": "Zenodo",
    "zenodo_record_id": "20427364",
    "source_filename": "13_pred_2025_08_01_with_predictions_01.xlsx"
}

---

Main API

Function	Purpose
show_map()	Download predictions from Zenodo and display or save the map
load_prediction_from_zenodo()	Retrieve a prediction file directly from Zenodo
load_prediction_file()	Read a local Excel or CSV prediction file
list_zenodo_files()	List files available in the Zenodo record
plot_wells()	Plot an existing prediction DataFrame
predictions_to_geodataframe()	Convert prediction data to a GeoDataFrame
extract_prediction_date()	Extract the prediction date from a filename
validate_predictions()	Validate required columns and values

---

Data source

The groundwater-status predictions are distributed through the following open Zenodo dataset:

Tziritis, E., Iatrou, M., Drougas, C., Tasoglou, S., & Arampatzis, G. (2026). Hydrogeological controls on seasonal groundwater level dynamics across Greece: a national-scale assessment integrating monitoring data, earth observation, and machine learning (2013–2022) [Data set]. Zenodo.

DOI: 10.5281/zenodo.20427364

The dataset is licensed under the Creative Commons Attribution 4.0 International License (CC BY 4.0).

---

Project structure

wellswatch/
├── docs/
│   ├── red_nonred_precipitation_2013_2025.png
│   ├── wellswatch_2015.png
│   └── wellswatch_2025.png
├── src/
│   └── wellswatch/
│       ├── __init__.py
│       ├── mapping.py
│       └── zenodo.py
├── tests/
├── LICENSE
├── README.md
└── pyproject.toml

---

Authors

• Evangelos Tziritis
• Miltiadis Iatrou
• Christos Drougas
• Spyridon Tasoglou
• George Arampatzis

Maintainer: Miltiadis Iatrou

---

Citation

When using the groundwater prediction data, cite the Zenodo dataset:

Tziritis, E., Iatrou, M., Drougas, C., Tasoglou, S., & Arampatzis, G.
(2026). Hydrogeological controls on seasonal groundwater level dynamics
across Greece: a national-scale assessment integrating monitoring data,
earth observation, and machine learning (2013–2022) [Data set]. Zenodo.
https://doi.org/10.5281/zenodo.20427364

When referring specifically to the software, also cite the WellsWatch GitHub repository:

WellsWatch, version 0.1.0.
https://github.com/Mil-afk/wellswatch

---

Licenses

• The WellsWatch source code is distributed under the MIT License.
• The Zenodo prediction dataset is distributed under the CC BY 4.0 License.

---

Acknowledgement

WellsWatch uses publicly archived prediction files from Zenodo and country-boundary data from Natural Earth for map visualization.