Stochastic groundwater models can be a key contribution to hydrogeological studies, especially when the limitations of traditional workflows seem evident. Although efforts have been made in order to promote their use, they continue to be rather uncommon within the industry. There are several reasons for this to happen, including:
- The lack of knowledge about the type of results that can be generated, and about their potential contribution to commercial projects.
- The idea that stochastic groundwater models are too complicated to develop, requiring expertise that consultants might not have, along with excessive implementation times.
Stochastic models are indeed sophisticated tools, however the key to their success relies on (1) understanding how they can make a significant difference throughout a particular project, (2) getting the most value out of the stochastic realizations by using appropriate post-processing methodologies, and (3) speeding up their implementation time through the use of specialized software. Following these ideas, new features have been added to GSMOD, including:
- Improved processing and post-processing times of the stochastic simulations through the use of parallel computing.
- Capacity to generate histograms of the stochastic results (Figure 1), in addition to the stochastic hydrographs previously presented in GSMOD part I.
- Capacity to generate contour maps based on frequency analysis of the stochastic realizations (Figure 2), in addition to the probability maps previously presented in GSMOD part I.
- Capacity to identify the most influential parameters on the calibration and predictive simulations (Figure 3 and 4), allowing to evaluate potential areas of improvement for the model and hydrogeological characterization. Which are the most important parameters for the predictions of interest of your study? Is your calibration highly dependent on these parameters? Are these parameters well characterized?
- Capacity to export contour map animations to GIF and MPG files (Figure 5).
Figure 1: Histogram of the maximum concentration at a given location for a total of 1000 realizations (423 of 1000 with NRMS transport < 10%)
Figure 2: Drawdown maps for the best calibrated model vs percentile 95% of 357 realizations (357 of 500 with NRMS flow < 5%)
Figure 4: Parameter influence on the flow calibration (555 of 1000 with NRMS flow < 5%)
Figure 5: Probability of exceeding chloride concentration limit (423 of 1000 realizations with NRMS transport < 10%)
