Climate change causes cities to deal with increased temperatures and more frequent weather extremes. Heat waves will occur more often, becoming a more prevalent issue in especially urban areas. The quantification of heat stress is a first step to define mitigation measures. For that purpose, a standardised method to assess the spatial influence of surfaces on the Physiological Equivalent Temperature (PET) was developed. This study aims to reshape this model into a statistical dependence model which is more flexible regarding missing data. To this end, we used a Non-Parametric Bayesian Network (NBPN). We created a model driven by both data and expert knowledge, that is capable of dealing with input data layers with a grid resolution up to 20 m. Results show that training the model with only 20 sample points did not affect the performance considerably, compared to using 2,000 data points. Inclusion of a layer with sky view factor mainly improves the estimation of observations in the tails of the distribution. The model predicts the PET with a Mean Absolute Error (MAE) of 1 to 2 ºC, dealing adequately with missing data layers. With this limited amount of necessary input, the NPBN in our study helps in standardising the assessment of heat stress outside the borders of the Netherlands. Also, our model offers a framework to make a first assessment regarding the effect of NBSs on heat stress.

In this paper we discuss PyBanshee, which is a Python-based open-source implementation of the MATLAB toolbox BANSHEE. PyBanshee constitutes the first fully open-source package to quantify, visualize and validate Non-Parametric Bayesian Networks (NPBNs). The architecture of PyBanshee is heavily based on its MATLAB predecessor. It presents the full implementation of existing tools and introduces new modules. Specifically, PyBanshee allows for: (i) choosing fully parametric one-dimensional margins, (ii) choosing different sample sizes for the model-validation tests based on the Hellinger distance, (iii) drawing user-defined sample sizes of the NPBN, (iv) sample-based conditioning sampling (similarly to the closed-source proprietary package UNINET by LightTwist Software) and (v) visualizing the comparison between the histograms of the unconditional and conditional marginal distributions. New detailed examples demonstrating new features are provided.@en