PAGOS: A New Python Library for Fast Implementation of (Tracer-Based) Hydrological Models

      In order to make generalised, quantitative statements and predictions about hydrological systems, numerically-estimated mathematical parameterisations are essential. Few-parameter models, optimised using measurements of physical/chemical tracers, are ubiquitous in the hydrological sciences. Examples include groundwater noble gas paleothermometry, determination of groundwater pollution sources from toluene measurements and determination of water mass fractions/transformation processes using hydrographic variables and gas/mineral solute concentrations. All of these applications involve minimising a cost function against a small (~10 or fewer) set of tracer observations for each sample. However, the computational implementation of any new tracer exchange model is time-consuming and often involves many redundant steps: a programmatically-represented mathematical model often must be retroactively suited to previous code, not necessarily written by the same person. With increasing project complexity, even small editions to a model may necessitate many changes propagating throughout a program, costing the programmer time and more easily introducing errors. With time-pressure and greater emphasis on the scientific results of a project rather than the code used to obtain them, accessibility and readability of software (i.e. its FAIRness) suffers.

      We have developed PAGOS (Python Analysis of Groundwater and Ocean Samples), a Python package which serves to streamline the development and testing of hydrological models, reducing the “scientist-side” time and effort required while also providing an environment conducive to highly accessible code development. Any hydrographic variable/tracer-based model representable as a function in Python can be quickly implemented in as few as 3 lines of code, whereafter it can immediately be forward-run with known parameters, or those parameters can be estimated by inverse-modelling against a user-provided dataset. PAGOS also automatically handles units, sparing the user the task of writing and re-writing methods to account for different units, and avoiding unit-conversion errors (to which hydrological investigations are particularly prone). A plotting subroutine is also provided by the package.

      The case study for which PAGOS was initially developed is an investigation of surface gas exchange processes in the Arctic Ocean, parameterising new models with respect to noble gas measurements. Additionally, parameter estimates for selected models in groundwater and ocean sciences literature have been reproduced by PAGOS. These applications all use 3–5 noble gas tracers, but any number of tracers of any kind may be employed by the user.

      Collaborative extension of PAGOS’s scope to more areas in hydrology is encouraged through a public GitHub repository (https://github.com/TeamPAGOS/PAGOS).