Poster Presentation NCGRT/IAH Australasian Groundwater Conference 2019

Design of a simple tool for automated groundwater level forecasting (189)

Andrew L. Dark 1 , Tim Kerr 1 , Helen K. Rutter 1 , Julian J. Weir 1
  1. Aqualinc Research Ltd., Christchurch, New Zealand


Knowledge of the likely variation in groundwater levels over an upcoming season is desirable for water resource managers and users to plan more efficiently for the near future. The objective of this project was to test the technical feasibility of a simple automated groundwater level forecast product.

Design and Methodology

Groundwater level eigenmodels (Bidwell and Burbery, 2011) are a logical approach for rapid implementation of a forecasting system (e.g. Bidwell 2005). The input to an eigenmodel may be simply generated using a soil-moisture balance model (e.g. Aqualinc’s IrriCalc model), driven by weather, land-use, irrigation and soil properties. Assuming the irrigated area and land-use remains constant, only weather variability is required to generate forecasted land surface recharge and groundwater levels.

To overcome the limitations of seasonal weather forecasting, the groundwater forecast model is driven with an ensemble of past weather observed for the forecast season. This assumes that the future weather is likely to fall within the range of historical weather. Using past weather greatly simplifies the overall forecast system as the data are already available. The forecasting method then simplifies down to five steps:

  1. Obtain current weather variables;
  2. Update the modelled soil moisture state;
  3. Update the modelled groundwater state;
  4. Apply historic weather to the forecast period to generate land-surface recharge and groundwater level forecasts and
  5. Present the range of forecast groundwater levels. 

This approach requires judgement from the user on the likely outcome within the forecast range.  

The system can be implemented in R on a standard personal computer with the resulting forecast uploaded to a web site.

Data and Results

The following figure shows outputs from the forecast tool, including recent observed water levels, the historic range of water levels, and the forecast range based on the ensemble of historic climate data. 


A useful by-product of the forecast process is a semi-realtime update of modelled soil moisture and land-surface recharge, as shown below:


This project has enabled discussions on the utility of groundwater forecasting and the direction of further development. Possible next steps include developing the ability to run scenarios (e.g. changing irrigation demand), and assimilation of soil moisture and water-use measurements.   


Simple automatic groundwater level forecasts for New Zealand (or elsewhere) are technologically possible using currently-available tools.  Key advantages are the low cost and rapid implementation, compared to numerical groundwater models.  Feedback from potential users will determine the path of future development.




  1. Bidwell, V., Burbery, L (2011). Groundwater Data Analysis - quantifying aquifer dynamics. Prepared for Envirolink Project 420-NRLC50 (No. 4110/1). Lincoln Ventures Ltd.
  2. Bidwell, VJ (2005): Realistic forecasting of groundwater level, based on the eigenstructure of aquifer dynamics. Mathematics and Computers in Simulation 69:12-20.