Quantifying shallow groundwater table uncertainty from sensor density and placement in very dense networks: A case study

ABSTRACT

Graphical abstract showing the study sites with the estimated groundwater table, a table of the six study cases, the methodological workflow on the stochastic sampling and the resulting NSE and MAE values.

Effective monitoring of shallow groundwater is essential for climate adaptation, yet clear guidelines for sensor network design remain limited. This study uses two dense monitoring networks (up to 83 sensors/km2) in contrasting settings to evaluate how sensor density and placement influence groundwater surface reconstruction. Six cases were analysed, including long-term averages, event responses, and reduced-network configurations. Ordinary kriging was applied to thousands of random sensor combinations, and performance was assessed using Nash–Sutcliffe efficiency (NSE) and mean absolute error (MAE) relative to a baseline interpolation. Results show that sparse networks can yield high performance. Increasing sensor density primarily improves robustness by reducing error variability, while gains in model efficiency are limited beyond intermediate densities. At the village site, increasing sensor density does not consistently improve accuracy, reflecting the complexity of groundwater dynamics in residential environments. Across the top 20% of configurations, MAE ranges from 0.15 to 1.2 m depending on study site, corresponding to relative uncertainties of 4–17%. This demonstrates that strategic sensor placement is as important as sensor density, and acceptable uncertainty can be achieved with fewer sensors when key gradients and hydraulic extremes are captured. The findings provide practical guidance for designing cost-effective groundwater monitoring networks.