S.S. Papadopulos & Associates, Inc. and Watermark Numerical Computing (the developers of PEST) are pleased to provide consultancy services in model calibration and predictive uncertainty analysis.
Please contact us if you need help with any of the following:
Use of PEST in the calibration of complex environmental models, including groundwater flow and transport models, surface water quality and quantity models, and any other kind of model
Using the latest technology ( such as SVD-Assist ), we can show you how to apply regularized inversion to the calibration of complex models. With regularized inversion model-to-measurement fits are better, and model predictive error variance is smaller than using traditional zone-based techniques.
No graphical user interface or simulation program provides solutions for all modeling applications. We have developed programs to support the modeling process in a wide variety of settings. We can write custom software to enhance your modeling in specific application areas; assist you in processing environmental data; and allow you to most efficiently and effectively incorporate data into the model calibration and predictive analysis processes.
Quantification of model predictive uncertainty
Where the cost of being wrong is high, model predictive uncertainty analysis is an essential component of model deployment. Using PEST and ancillary software we can help you understand potential errors associated with critical model predictions through comprehensive analysis of uncertainty. This includes innate parameter variability, constraints on this variability imposed by model calibration, and the possible contribution from complexity that is beyond the reach of the calibration process.
Parallelization of the model calibration and predictive uncertainty analysis process
Once an inversion problem has been properly posed and a PEST dataset created for your model, we can help you calibrate that model faster using Parallel PEST on a network of fast computers.
Optimization of data acquisition
Data acquisition is expensive yet sound environmental management necessitates it. What is the best data to gather? It is that which reduces the uncertainty of a key model prediction the most. The extent to which the acquisition of data can reduce model prediction uncertainty can be quantified using PEST in regularization mode together with ancillary software. Hence, different data-gathering strategies can be evaluated using sound scientific bases.
Models or mediation and negotiation
PEST enables the quantification of the predictive limits of available data and modeling software, leading to understanding of the limits of the model as deployed. This quantification of uncertainty allows models to be deployed to support negations between stakeholder groups, and as a basis for dispute resolution.
We can provide these services tailored to your modeling requirements. One option is a 40-hour “PEST Immersion” which includes:
8 hours for us to review your modeling needs and prepare specific materials.
A 12-hour in-house PEST course introducing basic aspects of non-linear parameter estimation, and the advanced aspects that are relevant to your modeling context.
A 12-hour period of collaboration to set up your model for non-linear parameter estimation, regularized inversion, or predictive analysis.
8 hours of follow-up technical support, to help you continue developing PEST skills.
Please contact us if you would like to discuss a PEST emersion similar to that described above, tailored to your specific needs.