Improved decision making supported by data analytics

Improved decision making supported by data analytics

We are living in the age of big data. But also small data and just data in general. While more information can help to make more informed decisions, navigating through the sheer volume of data can be a challenge. Storing, processing, and analysing data using traditional methods is often time consuming and costly. Before you even sit down to review the information, compiling different formats from multiple sources, can make the scenario even more complex to manage!

Data analytics provides a powerful systematic and reproducible scientific framework for interrogating large and complex datasets. To break it down further, data analytics is about taking raw datasets and analysing them using mathematical and statistical algorithms to test hypotheses and determine insights on system behaviour in settings with complex multi-faceted interactions. The reproducible and transparent scientific workflows have great synergy with adaptive management techniques. As more data is generated, the workflows can easily be updated and re-deployed, whereby new insights can be derived which then management measures can be adjusted.

As hydrogeologists we are not foreign to the concept of big data. Commonly our research and industry projects acquire millions of data points for groundwater levels, pressures, water quality parameters and pumping rates. With increased focus on sustainable management of our water resources, the demand for data acquisition grows and there is a commensurate increase in the complexity of the datasets available for a given project. The scale and complexity of these datasets sometimes mean that the efficiency of traditional hydrogeological methods of analysis are being strained due to budget and time constraints. There are real risks that subtle trends and impacts are potentially being overlooked, or detected too late as they get lost in the masses of background noise.

At EMM we have been using data analytics across a broad range of hydrogeology and geochemistry projects that typically have dozens of variables and millions of spatial, vertical and temporal data points. As we come across challenging problems, we design bespoke data analysis workflows to interrogate the data and derive insights for our clients. At the core of these services are statistical and machine learning algorithms that are peer reviewed, tested, and industry-accepted. They include techniques like regression and classification algorithms, dimensionality reduction techniques (factor analysis and Principal Component Analysis), clustering techniques, time series analysis and chained bespoke algorithms. We have found that coupling these powerful analytical tools with domain expertise (hydrogeology experience) always produces the most powerful outcomes. The subject matter expert brings their experience and knowledge to the process and guides the outcomes.

Data analytics is most effective when clear and targeted questions are posed for a given system or process. Those questions set the context for the scale of the problem, data appropriateness, the suitability of specific algorithms and statistical techniques, and what assumptions are valid.  

Some recent example applications where we have added data analytics to our hydrogeological workflows include:

  • Testing hydraulic connections between shallow aquifers and surface water systems using multiple lines of evidence and testing.
  • Establishing trends and degradation of water quality at scales from a project site through to a state-wide database with thousands of wells.
  • Differentiating between climate and anthropogenic (landholder pumping and/or mine dewatering) stresses and evaluating which stress may be contributing to changes in groundwater levels.
  • Rationalising the adequacy of monitoring networks through data worth evaluations.

Data analytics doesn’t replace traditional hydrogeology practices, but rather complements them and expands the range and breadth of knowledge gained from a given dataset in shorter periods of time. Using these advanced techniques is a demonstrated way of driving and influencing effective, science-informed decision making.

Tim Ezzy
Tim Ezzy
Associate Director