Hello, I'm doing research on the degree of optimization of a groundwater monitoring network for a study area in The Netherlands. I'm conducting an elimination approach; monitoring wells that do not have informational relevance to the network are eliminated from the network. Now, I would like to analyze the sensitivity of the input data (groundwater level data of all monitoring wells) to understand how variations in the input data can affect the output of the model (optimal number of monitoring wells, monitoring wells that area eliminated). The sensitivity analysis would give insight into the reliability of the model, thinking of the eliminated monitoring wells and how I can minimize the RMSE of the model and MAE of the eliminated monitoring wells. What would be a suitable method for a SA in Python? I heard things about Monte Carlo and Sobol, but I'm not sure if those will fit in my Python script.
Many thanks in advance!
Hannah Willemijn, Sensitivity analysis is a powerful tool to understand how changes in input data impact the output of your model. In your case, you're looking to assess the reliability of your groundwater monitoring network optimization model by exploring how variations in groundwater level data affect the results. So far concerned with MC Simulation, imagine this like rolling a dice multiple times. Instead of fixed groundwater levels, you randomly vary them within a certain range for each monitoring well. Then, run your model with these different sets of input data. By observing how the outputs vary, you gain insights into the uncertainty and variability of your model. This method is relatively straightforward to implement but may require many model evaluations. While for SA, Sobol analysis provides a more systematic way to understand which input parameters have the most influence on your model. It decomposes the variance in your output into contributions from individual parameters or their interactions. While more complex than Monte Carlo, it offers detailed insights. You use Sobol indices to quantify the sensitivity of your model to each input parameter. I recommend starting with Monte Carlo. In Python, you can use libraries like NumPy and pandas to handle random sampling. Modify your existing script to run the model iteratively with different sets of randomly sampled groundwater level data for each monitoring well. Observe how the outputs vary and consider metrics like RMSE and MAE to evaluate the model's reliability. If you find that you need more detailed insights into parameter sensitivities, you can then explore Sobol analysis using libraries like SALib. This involves a more structured approach and may require more computational resources. Remember, the choice between these methods depends on your specific research goals and available resources. Start simple, and as you become more comfortable, you can explore more advanced techniques. For more connect with me on https://wa.me/+923440907874. Good luck with your research!
Dear Engr. Tufail , many thanks for your detailed answer and suggestions to my query!
If I understand it correctly, it is the best to provide different scenarios (high and low GWL data e.g.), compare the results of each scenario, and evaluate the results of the scenario by analyzing the RMSE of the reduction percentage and the MAE of the eliminated monitoring wells.
Do you also have a suggestion on an approach to determine the ideal reduction percentage for elimination? This is in correlation with the RMSE. A result of the elimination will be: a figure which describes: x-axis: n monitoring wells, y-axis: RMSE, according to the reduction percentage).
Sensivity analysis is very important in modeling, by considering the limitations of observation wells. you try to have low difrence between observation wells and observation wells in the modeling for intance
Hello. Currently I am studying on groundwater quality index modelling
I explain about sensivity analysis in there:
Assessing the earth dams’ effect on the groundwater of its location case study: Kord-Oliya dam
- Badges
- Science topic
- Similar topics
- Computer Science
- Research Topics