Marcin Mazur (p), Piotr Rosik (p)

Miroslav Despotovic

Centrography deals with the average location of the spatial distribution of a given phenomenon. It has become a widespread field of geographic research in an era when there arose a need for scientific confirmation of the process of colonization of Siberia and the Wild West, followed by the expansion of various socio-economic phenomena. Therefore, it is clear that the definition of the centroid and the methods of its determining initially were a key issue, since tracking and analyzing its trajectory was a convenient tool for the purposes of that time. This paper comprehensively reviews one-dimensional descriptive statistics of central tendency, which provide a starting point for generalizing to the "various centroids" of a two-dimensional Cartesian map space. Among these statistics, one can indicate not only numerous types of averages, including the most commonly applied arithmetic mean and weighted mean, but also such simpler but no less informative ones as mode, median, midrange or midhinge. However, after the era of geographic exploration and tracking the trajectory of "centroids", there has emerged a need to measure dispersion. Similarly, as very different sets of quantitative data may be described by a very similar mean value, the centroid can also be placed in a likely location, regardless of the concentration of the mass (weight) of phenomena, whether it is concentrated centrally (e.g., Hungary), polycentrically (e.g., Poland) or cumulatively on peripheries (e.g., Spain). Hence, although dispersion measures have so far been rather overlooked in the literature and often confined to Standard Distance and Standard Deviation Ellipses, they ought to be perceived as complementary in relation to measures of central tendency and equally important in the synthetic description of spatial structures. Thus, this paper goes beyond a review of measures of central tendency and aims to provide a list of dispersion measures that may also be generalized to two-dimensional Cartesian map space and be applied in geographic research. Selected pairs of central tendency and dispersion measures have been empirically verified on the example of European spatial units, different definitions of “mass” and “distance”, and varying temporal intersections. The results were presented cartographically and cross-referenced. Conclusions focus on the difference in findings obtained with diverse pairs of measures, indicating the potential impact of the selection of the measure on the emphasized features of spatial patterns.

Hiroaki Shirayanagi (p), Yukisada Kitamura

Marcin Mazur

　The population of Japan has been declining since 2004 and the average age of the population is expected to become increasingly older. The number of emergency calls in Japan has been increasing every year, and has increased by approximately 1.59 times in the past 19 years (from 4.18 million calls in 2000 to 6.64 million calls in 2019). Moreover, the travel time from the fire station to the site of the emergency has increased by 3.2 minutes over the past 20 years (from 6.2 minutes in 2001 to 9.4 minutes in 2021), and the travel time to the hospital has increased by 14.3 minutes over the past 20 years (from 28.5 minutes in 2001 to 42.8 minutes in 2021), so the burden on emergency transport operations has increased significantly in recent years.
　Based on this background, the purpose of this study is to quantitatively discuss the optimal allocation of fire stations from the viewpoints of emergency transport workload and disaster prevention and in consideration of the relocation of the Misato sub-branch station of the Kochi City Fire Department.
　First, we constructed a system to calculate the travel time from the fire station to the site of an emergency located on a road network in Kochi City, and the total travel time from the fire station of the Kochi City Fire Department.
Second, three potential sites for relocation (P1: near the entrance intersection of Kochi Medical Center, P2: near the entrance of Nozomigaoka, and P3: near the east side of Totsu Elementary School) were selected, satisfying the following conditions: a tsunami is not expected in the event of a Nankai Trough earthquake and a fire station more than 2,000 m2 could potentially be constructed.
　Finally, the total travel time from the Kochi City Fire Department fire station to the emergency stations at the three potential relocation sites (P1, P2, and P3) was calculated.
　As a result, the total arrival time from P1 is the shortest at 147,216 minutes, and P1 was selected as the optimum place for relocation of the current Misato sub-branch station.

Jeong-Il Park (p)

Hiroaki Shirayanagi

Recent digitalization of production may increase the urban location tendency of industry by enabling downsizing factory and improving productivity. However, studies on this have been mostly conceptual approaches, and there are very few empirical case studies on this. Therefore, this study aims to empirically examine whether the digitalization of production at the firm level increases the urban location tendency by downsizing factory and improving productivity. It extracted information on the digitalization level of production, the area of production space, and productivity of 1,750 companies by using raw data of Survey of the Information Level of SMEs conducted by the Korea Technology and Information Promotion Agency. In addition, the urban location index was calculated using the modified gravity model, and location pattern for each firm was analyzed through GIS analysis. As a result of the analysis, firms with high level of digitalization, small the production size, and high land productivity have exhibited a considerably strong urban location tendency compared to other types of firms. In addition, the analysis results of the multi-level regression model showed that the digitalization of production not only had a positive effect on the downsizing of factory and productivity, but also increased the tendency of urban location. Based on these findings, this study discussed several policy implications for urban industrial location and urban industrial spaces.

Miroslav Despotovic (p), Eris Stumpe

Jeong-Il Park

Real estate location quality assessment methodologies typically work with data from a single modality (i.e., mostly structured data). Today's data analysis techniques allow for the combination of multiple data modalities, which basically increases the information content and, if a suitable methodology is applied, could allow for a more accurate assessment. To this end, we leverage the capacity of a functional neural network model with shared layers of multiple networks and use variable number of inputs per instance from text and image data modalities. Thus, we apply a flexible and robust extraction of location-based features with two different visual representations (satellite and street view images) along with the textual descriptions of the location and estimate the quality classes. For inference, we compare the results to quality classes provided by experts and test the consistency and variability of the predictions. On top of the satisfactory classification results of the multimodal model, we also found significant effects of the estimated classes on the price, compared to the baseline models containing the ground truths. Our work contributes to the development of multimodal information extraction techniques for text and the variable number of image data, which could be applied in the context of a practically applicable unified assessment model that integrates all available modalities and thus can utilize their complementary information.