Announcing the Groundwater Spatial Modeling Challenge

Ugh! - Haven’t we already fully explored the potential of spatial modeling of groundwater parameters? Haven’t we reached the limits there? In response to questions about whether we have fully explored the potential in these areas, we are launching this challenge to harness the swarm intelligence of the groundwater modeling community. Our aim is to dispel doubts and encourage creativity in feature development, model selection and training strategies. And why? Because regionalization, interpolation from point data into space, remains a crucial step in generating spatially continuous information and maps. And maps remain a crucial basis of information in sustainable water and resource management.

For this purpose we provide a dataset of Nitrate-concentrations in approximately 1800 wells taken within a single month in southwestern Germany. The measured concentrations in the wells are representing concentrations in the most shallow aquifers. Besides Nitrate concentration as target variable, the dataset contains various features, describing the environmental and geological context of each sample site. Different types of models can be applied to model Nitrate concentrations, ranging from deterministic and geostatistical models to statistical/data-driven approaches such as machine learning models. We invite all interested researchers and data science enthusiasts to participate in this challenge as a team or single person. A ranking will be carried out while the challenge is open using a predefined set of model performance metrics on a secret test split. Further information on how to participate and the required data is available at https://groundwater-spatial-modeling-challenge.github.io/challenge2024/.

The well defined rules of the challenge regarding the feature set, data splitting and metric choices, will allow the groundwater community to learn from different approaches and conduct a systematic comparison. 

The results of the challenge will be presented at the General Assembly of the EGU in 2025 and documented in a peer-reviewed paper with model contributors as co-authors on request. Through  this challenge, we hope to increase the awareness in the groundwater community on the range of approaches available for (spatial) modeling of groundwater variables and their advantages and disadvantages.