Spatiotemporal association pattern mining can discover interesting interdependent relationships among various types of geospatial data. However, existing mining methods for spatiotemporal association patterns usually model geographic phenomena as simple spatiotemporal point events. Therefore, they cannot be applied to complex geographic phenomena, which continuously change their properties, shapes or locations, such as storms and air pollution. The most salient feature of such complex geographic phenomena is the geographic dynamic. To fully reveal dynamic characteristics of complex geographic phenomena and discover their associated factors, this research proposes a novel complex event-based spatiotemporal association pattern mining framework. First, a complex geographic event was hierarchically modeled and represented by a new data structure named directed spatiotemporal routes. Then, sequence mining technique was applied to discover the spatiotemporal spread pattern of the complex geographic events. An adaptive spatiotemporal episode pattern mining algorithm was proposed to discover the candidate driving factors for the occurrence of complex geographic events. Finally, the proposed approach was evaluated by analyzing the air pollution in the region of Beijing-Tianjin-Hebei. The experimental results showed that the proposed approach can well address the geographic dynamic of complex geographic phenomena, such as the spatial spreading pattern and spatiotemporal interaction with candidate driving factors.

Building footprints are among the most predominant features in urban areas, and provide valuable information for urban planning, solar energy suitability analysis, etc. We aim to automatically and rapidly identify building footprints by leveraging deep learning techniques and the increased availability of remote sensing datasets at high spatial resolution. The task is computationally challenging due to the use of large training datasets and large number of parameters. In related work, You-Only-Look-Once (YOLO) is a state-of-the-art deep learning framework for object detection. However, YOLO is limited in its capacity to identify small objects that appear in groups, which is the case for building footprints. We propose a LOcally-COnstrained (LOCO) You-Only-Look-Once framework to detect small and densely-distributed building footprints. LOCO is a variant of YOLO. Its layer architecture is determined by the spatial characteristics of building footprints and it uses a constrained regression modeling to improve the robustness of building size predictions. We also present an invariant augmentation based voting scheme to further improve the precision in the prediction phase. Experiments show that LOCO can greatly improve the solution quality of building detection compared to related work.

This paper seeks to evaluate and calibrate data collected by low-cost particulate matter (PM) sensors in different environments and using different aggregated temporal units (i.e., 5-s, 1-min, 10-min, 30 min intervals). We first collected PM concentrations (i.e., PM₁, PM_2.5, and PM₁₀) data in five different environments (i.e., indoor and outdoor of an office building, a train platform and lobby of a subway station, and a seaside location) in Hong Kong, using five AirBeam2 sensors as the low-cost sensors and a TSI DustTrak DRX Aerosol Monitor 8533 as the reference sensor. By comparing the collected PM concentrations, we found high linearity and correlation between the data reported by the AirBeam2 sensors in different environments. Furthermore, the results suggest that the accuracy and bias of the PM data reported by the AirBeam2 sensors are affected by rainy weather and environments with high humidity and a high level of hygroscopic salts (i.e., a seaside location). In addition, increasing the aggregation level of the temporal units (i.e., from 5-s to 30 min intervals) increases the correlation between the PM concentrations obtained by the AirBeam2 sensors, while it does not significantly improve the accuracy and bias of the data. Lastly, our results indicate that using a machine learning model (i.e., random forest) for the calibration of PM concentrations collected on sunny days generates better results than those obtained with multiple linear models. These findings have important implications for researchers when designing environmental exposure studies based on low-cost PM sensors.

Regional co-location patterns represent subsets of feature types that are frequently located together in sub-regions in a study area. These sub-regions are unknown a priori, and instances of these co-location patterns are usually unevenly distributed across a study area. Regional co-location patterns remain challenging to discover. This study developed a multi-level method to identify regional co-location patterns in two steps. First, global co-location patterns were detected, and other non-prevalent co-location patterns were identified as candidates for regional co-location patterns. Second, an adaptive spatial clustering method was applied to detect the sub-regions where regional co-location patterns are prevalent. To improve computational efficiency, an overlap method was developed to deduce the sub-regions of (k + 1)-size co-location patterns from the sub-regions of k-size co-location patterns. Experiments based on both synthetic and ecological data sets showed that the proposed method is effective in the detection of regional co-location patterns.