Cartographic map generalization involves complex rules, and a full automation has still not been achieved, despite many efforts over the past few decades. Pioneering studies show that some map generalization tasks can be partially automated by deep neural networks (DNNs). However, DNNs are still used as black-box models in previous studies. We argue that integrating explainable AI (XAI) into a DL-based map generalization process can give more insights to develop and refine the DNNs by understanding what cartographic knowledge exactly is learned. Following an XAI framework for an empirical case study, visual analytics and quantitative experiments were applied to explain the importance of input features regarding the prediction of a pre-trained ResU-Net model. This experimental case study finds that the XAI-based visualization results can easily be interpreted by human experts. With the proposed XAI workflow, we further find that the DNN pays more attention to the building boundaries than the interior parts of the buildings. We thus suggest that boundary intersection over union is a better evaluation metric than commonly used intersection over union in qualifying raster-based map generalization results. Overall, this study shows the necessity and feasibility of integrating XAI as part of future DL-based map generalization development frameworks.

A growing number of citizens and local governments have embraced the use of Twitter to communicate during natural disasters. Studies have shown that online communications during disasters can be explained using crisis communication taxonomies. However, such taxonomies are broad and general, and offer little insight into the detailed content of the communications. In this paper, we propose a semi-automatic framework to extract and compare, in retrospect, the digital communication footprints of citizens and governments during disasters. These footprints, which characterize the topics discussed during a disaster at different spatio-temporal scales, are computed in an unsupervised manner using topic models, and manually labelled to identify specific issues affecting the population. The end objective is to offer detailed information about issues affecting citizens during natural disasters and to compare these against local governments' communications. We evaluate the framework using Twitter communications from 18 snowstorms (including two blizzards) on the US east coast.

Big GPS trajectory datasets can have redundant spatio-temporal information for applications, which requires simplification as a key preprocessing for modeling. Many existing simplification methods focus on the geometric information from a trajectory per se. Conversely, methods considering geographic context often fail to provide spatially adaptive simplification, or require complex parameter settings to achieve this task. This study proposes a novel two-stage adaptive trajectory simplification method embedding spatial indexing, enrichment, and aggregation in an integrated process. The first stage employs a quadtree for the subdivision depending on the density of geographic context features (i.e. POIs), leading to a variable-resolution representation of the area. The second stage aggregates trajectory waypoints locating in the same quadtree leaf node into a representative point, making the aggregation adapting to the spatial layout of the geographic feature in the first stage. Evaluation with a real-world vehicle trajectory dataset shows that the proposed approach can automatically simplify trajectory segments at variable compression ratios with greater simplification in areas with sparse context features (e.g. rural) and less simplification in areas with dense context features (e.g. urban). More importantly, the method can still preserve inter-trajectory distances between original trajectories and simplified ones, while significantly reducing the computing time.