Escaping the infinite modelling maze: useful decision pathways to follow when exploring and interpreting thermal history models using QTQt.

Abstract:

Numerical thermal history modelling has become a core approach used for interpretation of low-temperature thermochronometry (LTT) data as modelling programs can derive rock time-temperature (t-T) paths from measured LTT data while incorporating complex factors affecting the kinetics of a particular thermochronometric system (e.g. grain size, radiation damage, and composition). ‘QTQt’ (Gallagher, 2012) is a software commonly used for forward and inverse thermal history modelling (also HeFTy, see parallel study by Murray et al., 2020), however, approaching this tool can be challenging for experienced and novice thermochronologists alike. The modelling process involves making key decisions about (i) data input, (ii) initial set-up of model space and parameters, (iii) kinetic model(s) (i.e. annealing, diffusion, radiation damage model), and (iv) additional t-T constraints. In addition, users need to have an understanding of the statistical methods underlying the modelling approach to be able to interpret the model outputs and the relationship between the observed and predicted data.

Here, we present examples of key decision points users face when modelling LTT data and aim to provide guidance and strategies to make informed and logical decisions. We use a suite of simple t-T paths (Wolf et al., 1998) to test forward and inverse models of various data inputs and model parameters including: data type (AHe vs AFT), data quantity (single-grain vs multigrain), error assignment, t-T prior information, and the use of constraint boxes. We suggest that users should perform similar exploratory modelling first by using forward models, generating synthetic data and following up with inverse modelling to retrieve the known answer. Systematically varying input parameters increase awareness and intuition about how different decisions can modify and potentially bias model outputs. Following this process will enable users to confront and better understand some of the common challenges faced when modelling LTT data including: integrating multiple LTT datasets, assigning uncertainties, the role of geologic constraints, the non-uniqueness of cooling ages, the sensitivity of the parameters in diffusion and annealing models, and limitations of a model to retrieve the ‘correct’ rock thermal history.