Geographic Units Research Articles

GLocal: A global development dataset of subnational administrative areas

The purpose of the GLocal dataset is to enable research in international development that requires both global scope and local precision. Leveraging modern geospatial analysis tools, we process a diverse array of sources to provide researchers with a growing set of economic, demographic, ecological and socio-political variables for geographic units relevant to public policy. We provide separate data files for different levels of administrative and periodic aggregation, along with ad-hoc files with more detailed information on specific topics. In this data descriptor paper, we discuss both our data processing methodologies and validation pipelines, and provide a short case study to illustrate the research potential of the dataset. We also introduce a simple web app, glocal.streamlit.app, which offers a user-friendly interface for exploring and visualizing the dataset. Given the growing number of public and granular sources of relevance for international development research, we hope to continue adding features and expand the GLocal dataset in the future.

Assessment of Chinese urban land-use efficiency (SDG11.3.1) utilizing high-precision urban built-up area data

Assessment of SDG11.3.1 indicator of the United Nations Sustainable Development Goals (SDGs) is a valuable tool for policymakers in urban planning. This study aims to enhance the accuracy of the SDG11.3.1 evaluation and explore the impact of varying precision levels in urban built-up area on the indicator's assessment outcomes. We developed an algorithm to generate accurate urban built-up area data products based on China's Geographical Condition Monitoring data with a 2 m resolution. The study evaluates urban land-use efficiency in China from 2015 to 2020 across different geographical units using both the research product and data derived from other studies utilizing medium and low-resolution imagery. The results indicate: (1) A significant improvement in the accuracy of our urban built-up area data, with the SDG11.3.1 evaluation results demonstrating a more precise reflection of spatiotemporal characteristics. The indicator shows a positive correlation with the accuracy level of the built-up area data; (2) From 2015 to 2020, Chinese prefecture-level cities have undergone faster urbanization in terms of land expansion relative to population growth, leading to less optimal land resource utilization. Only in extra-large cities does urban population growth show a relatively balanced pattern. However, urban population growth in other regions and cities of various sizes lags behind land urbanization. Notably, Northeast China and small to medium cities encounter significant challenges in urban population growth. Our comprehensive framework for evaluating SDG11.3.1 with high-precision urban built-up area data can be adapted to different national regions, yielding more accurate SDG11.3.1 outcomes. Our urban area and built-up area data products provide crucial inputs for calculating at least four indicators related to SDG11.

Using Vascular Deserts as a Guide for Limb Preservation Outreach Programs Successfully Targets Underserved Populations

Vascular deserts, regions without vascular providers, previously described targets for limb salvage efforts. The Comprehensive Heart and Multidisciplinary Limb Preservation Outreach Networks (CHAMPIONS) programs targeted regions for outreach and evaluated the population using desert maps. At 2 events targeting underserved regions between 2022 and 2023, providers screened and educated participants on peripheral arterial and cardiovascular disease (PACD). Demographics and cardiovascular risk factors were collected. Using Arc geographic information system, vascular surgeons, and Vascular Quality Initiative (VQI) participating facilities were mapped with a 30-mile buffer. Participants were mapped with census data, and the healthy places index (HPI) was overlayed for population and social determinants of health data analysis in medical service study areas (MSSA), a geographical analysis unit. (Figure1) Results were compared to prior statewide deserts. Outreach program participants' mean age was 56 (range 6-88); 39% were male, and the majority were Hispanic (86%). 27% had no primary care provider (PCP). 30% had diabetes, 10% undiagnosed before the event, 38% had hypertension, 40% undiagnosed prior to the event, and 21% described intermittent claudication. 81% made <$30,000 annually, and 28% reported no health insurance. Similarities were observed when comparing program participant demographics to the population-level data from the targeted regions. Patients were more frequently Hispanic than other desert regions (68% vs. 36%, P<0.001). Compared to other vascular desert regions, the target population was more disadvantaged in all HPI domains, including economic (18 vs. 38%, P<0.001), education (21 vs. 39%, P<0.001), and transportation (30 vs. 40%, P<0.001). Worse education, financial, and transportation resources correspond to decreased care access due to poor literacy and travel burdens. CHAMPIONS programs successfully targeted populations needing care based on vascular care desert maps, demonstrating that at-risk populations can be successfully identified and screened for cardiovascular disease.

New perspectives, additions, and amendments to plant endemism in a North African flora

BackgroundEndemism is essential in biodiversity, biogeography, and conservation tasks. Based on herbarium specimens kept in some local herbaria, many published literature, and available information, we compiled a comprehensive list and an updated assessment of the Egyptian endemic and near-endemic taxa. The application of quantitative approaches to the distribution patterns, conservation status, and habitat preference of endemic taxa in Egypt was provided. Comparisons of the near-endemic taxa with other neighbouring flora were explained. For each taxon, the distribution patterns, most preferable habitat, biological spectrum, and taxa among 14 phytogeographical regions (Operational Geographical Units; OGUs) of Egypt were determined.ResultsIn this study, 19 endemics (out of 70) and 76 near-endemics (out of 181) are newly added taxa. Differentiation indices represented the taxonomic degrees of differentiation among endemic taxa. Two different indices were used to assess endemism: single-region endemic taxa (SRET) and multiple-region endemic taxa (MRET). Most endemic and near-endemic taxa were recorded from the mountainous Sinai (S) and the Mareotis sector of the Mediterranean coastal land (Mm). Generally, the most represented families in endemic and near-endemic areas were Asteraceae, Caryophyllaceae, Lamiaceae, and Fabaceae. More than 60% of the endemic taxa occurred in the sandy plains, wadis (desert valleys), and rocky plains and mountains. Applying hierarchical cluster analysis to the occurrences of 70 endemic taxa in the 14 studied OGUs revealed five main floristic groups (I–V), each characterized by certain OGUs. We provided eight groups of near-endemic taxa that represented their extension in neighbouring countries.ConclusionsThe presented data will help to fill the gap in our knowledge of endemism, provide baseline information to understand biogeographical processes and facilitate further cooperation toward conservation purposes.

Geographic representation and requests for federal funds in the U.S. senate

AbstractAn essential component of a representative's job centers on deciding where to allocate scarce resources. How do legislators attempt to allocate spending across the geographic units of their constituencies, and what determines a legislator's success in securing these resources? To answer these questions, we create a new dataset of all earmark requests made by U.S. senators in the 117th Congress. This data was recently made available as part of reforms, allowing researchers to examine both requests and outcomes. We show that senators prioritize seeking earmarks for areas of their constituencies critical to their electoral success. We then match these requests to earmarks funded by the Appropriations Committee and show that a legislator's institutional standing and the behavior of their same‐state colleague—but not electoral concerns—shape whether projects are successfully funded. We provide insight into a behind‐the‐scenes aspect of representation and illuminate the strategic dynamics that shape federal spending.

Coupling Hyperbolic GCN with Graph Generation for Spatial Community Detection and Dynamic Evolution Analysis

Spatial community detection is a method that divides geographic spaces into several sub-regions based on spatial interactions, reflecting the regional spatial structure against the background of human mobility. In recent years, spatial community detection has attracted extensive research in the field of geographic information science. However, mining the community structures and their evolutionary patterns from spatial interaction data remains challenging. Most existing methods for spatial community detection rely on representing spatial interaction networks in Euclidean space, which results in significant distortion when modeling spatial interaction networks; since spatial community detection has no ground truth, this results in the detection and evaluation of communities being difficult. Furthermore, most methods usually ignore the dynamics of these spatial interaction networks, resulting in the dynamic evolution of spatial communities not being discussed in depth. Therefore, this study proposes a framework for community detection and evolutionary analysis for spatial interaction networks. Specifically, we construct a spatial interaction network based on network science theory, where geographic units serve as nodes and interaction relationships serve as edges. In order to fully learn the structural features of the spatial interaction network, we introduce a hyperbolic graph convolution module in the community detection phase to learn the spatial and non-spatial attributes of the spatial interaction network, obtain vector representations of the nodes, and optimize them based on a graph generation model to achieve the final community detection results. Considering the dynamics of spatial interactions, we analyze the evolution of the spatial community over time. Finally, using taxi trajectory data as an example, we conduct relevant experiments within the fifth ring road of Beijing. The empirical results validate the community detection capabilities of the proposed method, which can effectively describe the dynamic spatial structure of cities based on human mobility and provide an effective analytical method for urban spatial planning.

Distortions in Judged Spatial Relations in Large Language Models

We present a benchmark for assessing the capability of large language models (LLMs) to discern intercardinal directions between geographic locations and apply it to three prominent LLMs: GPT-3.5, GPT-4, and Llama-2. This benchmark specifically evaluates whether LLMs exhibit a hierarchical spatial bias similar to humans, where judgments about individual locations’ spatial relationships are influenced by the perceived relationships of the larger groups that contain them. To investigate this, we formulated fourteen questions focusing on well-known U.S. cities. Seven questions were designed to challenge the LLMs with scenarios potentially influenced by the orientation of larger geographical units, such as states or countries, whereas the remaining seven targeted locations were less susceptible to such hierarchical categorization. Among the tested models, GPT-4 exhibited superior performance with 55 percent accuracy, followed by GPT-3.5 at 47 percent and Llama-2 at 45 percent. The models showed significantly reduced accuracy on tasks with suspected hierarchical bias. For example, GPT-4’s accuracy dropped to 33 percent on these tasks, compared to 86 percent on others. The models identified the nearest cardinal direction in most cases, however, reflecting their associative learning mechanism, thereby embodying human-like misconceptions. We discuss avenues for improving the spatial reasoning capabilities of LLMs.

Geographical analysis of the impact of road networks on traffic accidents in Pishdar District

An increase in the population and its needs, the development of transportation means and the diversification of transportation networks, as well as the differences in the geographical location of the population, are reasons for paying more attention to the road and transportation sector. In addition to all this, the increase in traffic accidents at the local and global levels is another reason for the attention of different fields. Pishdar district, as a geographical unit in the Kurdistan Region, has seen significant progress in the field of transportation in recent years, which has caused it to have its share of traffic accidents. The main goal of this study is to identify the characteristics of road networks in Pishdar district as well as a geographical analysis of traffic accidents and determine the level of solving these problems that are observed in the district in a scientific way. The research plan is divided into three main parts, each of which consists of two main sections: Part 1. This part describes the characteristics of transport networks and consists of two main sections. The first section includes the classification of road networks, and the second section includes the distribution of road networks at the level of administrative units in the study area. Part 2. This part is dedicated to the role of geographical principles in the time and place of traffic accidents in Pishdar district. The first is about the difference in time according to (month and year), and the second is the difference in location according to (city and outside the city). Part 3. This part is dedicated to how the most significant natural factors, which also include the most significant human factors, affect traffic accidents.

Comprehensive Risk Assessment of Farmer Households in Flood and Earthquake Prone Areas Based on Multidimensional Information Diffusion Model

In recent years, the global practice of regional natural disaster reduction and comprehensive risk prevention has promoted the integration of multiple disciplines, and the theory and technology of comprehensive risk assessment for multiple natural disaster chains urgently need innovation. This article is based on multiple natural disaster chain scenarios and sets secondary indicators for multiple influencing factors based on the first level indicators composed of basic risk factors such as hazard factors, vulnerability of disaster bearing bodies, and disaster prevention and reduction capabilities. By using multidimensional information diffusion technology, multiple hazard factor hazard combination factors, vulnerability combination factors of disaster bearing bodies, and comprehensive disaster prevention and reduction capability combination factors reflecting various natural disaster scenarios are constructed, A household comprehensive risk assessment index system for multi-dimensional information diffusion assessment of natural disaster comprehensive risk was designed by quantifying risk factors such as the intensity of comprehensive disaster causing factors, vulnerability of disaster bearing bodies, and comprehensive disaster prevention and reduction capabilities in disaster scenarios. This paper expands the traditional information diffusion comprehensive risk assessment model, makes up for the lack of information on geographical units, establishes a multi impact factor multi dimension natural disaster risk comprehensive assessment model for multi risk factors of complex disaster scenarios with small samples, obtains the theoretical value of multi-dimensional information Diffusion model comprehensive risk assessment for complex disaster scenarios, breaks through the sample size limit of natural disaster risks, and makes up for the lack of information on small samples of small probability events, Improved the reliability and accuracy of comprehensive risk assessment for multiple natural disaster chains. Then, take the disaster scenario formed by the comprehensive risk of earthquake and flood in a place in the upper reaches of the Fujiang River basin as the analysis sample, calculate the comprehensive risk assessment value of flood and earthquake damage in the study area's micro samples (rural households in villages and towns). By comparison, it is found that the comprehensive risk value assessed by the multi-dimensional information Diffusion model has good reliability and high accuracy, which is significantly better than the traditional Kriging interpolation method and geographical weighted regression method. This article provides a comprehensive risk assessment of natural disasters at the micro household level, which can better reflect the diversity and regionality of risks, help clarify the hidden risks of rural natural disasters, and provide scientific basis for government emergency management decision-making.

Does immigration increase crime? The advantage of dynamic threshold models with finer geographic units

This study investigates whether immigration and crime rates are positively related, by applying dynamic panel threshold models to fine geographic units. This is distinct from previous studies, in which static and continuous models are applied to coarse geographic units. This analysis reveals that the immigration-crime relationship has structural breaks (or discontinuities). The favorable ethnic network externalities on crime emerge after the immigrant share reaches a certain level in the community. This analysis also shows that the immigration-crime relationship observed in fine geographic units disappears when coarsely classified units are used. If geographic aggregation obscures heterogeneity among cities, the immigration-crime relationship is underestimated. These results suggest the advantage of dynamic threshold models with fine geographic units compared to traditional static and continuous models with coarse geographic units when discussing the immigration-crime relationship.

Examining multiscale built environment interventions to mitigate travel-related carbon emissions

While established studies have explored interventions in the built environment (BE) and transportation sector to mitigate travel carbon emissions (TCE), planners still struggle to determine the most effective units of intervention, identify key variables, and determine their optimal values. This study addresses the gap by employing the extreme gradient boosting (XGBoost) model to create a multi-scale comparative framework. This study revealed that the relationship between the built environment and travel-related carbon emissions varies depending on the zoning and scale of the BE measurement unit. The explanatory power of TCE varies across different geographic units, with the 15-min walk distance buffer of residents being the most effective in explaining TCE. Most variables were nonlinearly associated with TCE, and the precise threshold of the association between BE attributes and TCE was quantified. Based on these findings, we provide precise and nuanced insights into BE interventions to reduce TCE.

Phylogeography of Morella nana: The Wumeng Mountains as a natural geographical isolation boundary on the Yunnan-Guizhou Plateau.

The Yunnan-Guizhou Plateau (YGP) is characterized by the distinctive isolated habitat of the limestone Karst Islands and features the Wumeng Mountains, which divide the YGP into the two Plateaus of Yunnan and Guizhou. This study aimed to assess the effects of geographic isolation and past climate fluctuation on the distribution of flora in the YGP. To achieve this, we carried out the phylogeographical pattern and genetic structure based on chloroplast and nuclear ribosomal DNA sequence in relation to past (Last Glacial Maximum) and present distributions based on ecological niche modeling for Morella nana, an important wild plant resource and endemic to the YGP once considered a vulnerable species. The results suggest that the genetic and chlorotype network structures of M. nana are divided into at least two groups: cpDNA chlorotype H2 (or dominant nrDNA haplotypes h1 and h2), distributed primarily to the east of the Wumeng Mountains, and cpDNA chlorotypes H1 and H3-H10 (or dominant nrDNA haplotype h2 and h3), distributed to the west of the Wumeng Mountains. A deep genetic split was noted within the two groups to reach 25 steps, especially for the cpDNA fragment variation. This east-west divergence reveals the existence of a natural geographical isolation boundary in the form of the Wumeng Mountains, and supports the existence of at least two glacial refuges during the Quaternary glacial period, along with two genetic diversity center, and at least two large geographic protection units for the important species of M. nana. This study indicates that the phylogeographical pattern of M. nana can be attributed to geographic/environmental isolation caused by the Wumeng Mountains and climate fluctuation during the last glacial maximum, and proposes an effective strategy to protecting this important plant resource.

Site selection and prediction of urban emergency shelter based on VGAE-RF model

As urban development accelerates and natural disasters occur more frequently, the urgency of developing effective emergency shelter planning strategies intensifies. The shelter location selection method under the traditional multi-criteria decision-making framework suffers from issues such as strong subjectivity and insufficient data support. Artificial intelligence offers a robust data-driven approach for site selection; however, many methods neglect the spatial relationships of site selection targets within geographical space. This paper introduces an emergency shelter site selection model that combines a variational graph autoencoder (VGAE) with a random forest (RF), namely VGAE-RF. In the constructed urban spatial topological graph, based on network geographic information, this model captures both the latent features of geographic unit coupling and integrates explicit and latent features to forecast the likelihood of emergency shelters in the construction area. This study takes Beijing, China, as the experimental area and evaluates the reliability of different model methods using a confusion matrix, Receiver Operating Characteristic (ROC) curve, and Imbalance Index of spatial distribution as evaluation indicators. The experimental results indicate that the proposed VGAE-RF model method, which considers spatial semantic associations, displays the best reliability.

An unsupervised machine learning approach to the spatial analysis of urban systems through neighbourhoods’ dynamics

Urban systems’ dynamics are the result of two intertwined processes that operate at different rhythms: their physical structure and underlying social processes. This paper suggests a novel approach to the spatial analysis of urban systems, using neighborhoods as a basic building block. Neighborhoods are usually the minimal homogeneous geographical unit in urban areas, both regarding their physical, and social characteristics, and the availability of governmental data. Using unsupervised machine learning algorithm, an extensive real-estate transaction dataset, and census data, a multi-scale analysis of neighborhoods’ dynamics in England and Wales is performed. The spatial and temporal dynamics of the resulting clusters of neighborhoods highlights the urban challenges faced by entire urban systems. The results suggest that processes triggered by urban inequalities may affect not only the social structure of cities (for example, through gentrification and displacement), but also the environmental sustainability of the whole urban system at a much larger scale: Suburbanization pressures seem to threaten rural areas at an unprecedented magnitude. The identification of the areas where this pressure is acute allows for the design of appropriate urban policy responses. The main message for the analysis of urban dynamics is that physical transformations and socio-political rearrangements that are intrinsically intertwined: Therefore, the joint management of the both aspects is the key to a sustainable future. The analysis also highlights the potential of machine learning algorithms for the benefit of urban science in general, and the study of urban dynamics in particular.

Assessment and mapping of priority areas for transboundary ecological conservation: Suggestions for the protection of the Altai Mountains in Central Asia

Political boundaries often do not coincide with ecological boundaries, and it is common for the same physical geographical unit to span multiple countries. Exploring establishing transboundary protected areas and ecological networks can effectively protect the integrity of natural ecosystems separated by political boundaries. Based on the integrated evaluation of ecosystem services function, ecological vulnerability, and ecosystem integrity, this study explored a methodological system for evaluating ecological protection priority areas in transboundary areas and selected the Altai Mountains, which span four countries, contain China, Kazakhstan, Russia, and Mongolia, as a typical case to conduct empirical research. The study found that the extremely important areas of ecosystem services function, ecological vulnerability, and ecosystem integrity showed transboundary contiguous characteristics. Furthermore, weighted overlay analysis identified the spatial distribution pattern of ecological protection priority areas in the Altai Mountains. We suggested establishing transboundary protected areas in the four countries’ transboundary areas, building the identified ecological protection priority areas and existing protected areas into a transboundary protected area network, and deepening a multi-party transboundary protection cooperation mechanism. This study can provide a theoretical and practical reference to achieve the 2030 biodiversity conservation goals.

DETERMINATION OF THE RELATIONSHIP BETWEEN LEISURE TIME MANAGEMENT PERCEPTION AND ANXIETY LEVELS OF VETERINARY FACULTY STUDENTS

The aim of this study is to determine whether the perception of leisure time management and anxiety levels of veterinary faculty students are related or not. For this purpose, among students attending veterinary faculties during the spring semester of the 2020-2021 academic year, 822 students (410 females and 412 males) randomly participated in a survey. In study, correlation analyses, the independent two-group t test, ANOVA test, Mann-Whitney U test and Kruskal-Wallis are used. In result, a significant positive correlation is detected between the anxiety level and scheduling, whilst significant negative correlations are determined between the anxiety level and leisure time management behavior, anxiety level and goal setting and measurement, and anxiety level and strategy. It is concluded that male students suffer from a higher anxiety level than female students, and both the geographic region and unit of residence affect leisure time management.

Population growth in growing and declining cities: the role of balanced-budget increases in local government spending

Tiebout sorting describes people moving to communities that most closely match people's preferences over taxes and public services. This Tiebout equilibrium is disturbed when cities vote to increase taxes and spending. We study the effect of increased taxes and public spending on population growth in growing and declining cities. Using regression discontinuity to compare otherwise similar cities, we find increasing local government taxes and spending by 15% can increase population growth rates. The increase is only evident the year after the vote. For the general sample, the increase is 0.4 percentage points, but it is 0.8 percentage points (25% of a standard deviation) for growing cities with below-median percentages of elderly residents. In cities with declining population, passing large tax levies increases population growth rates by 0.9 percentage points the year after the vote—33% of a standard deviation. Instead of cutting taxes and services, cities with declining population might instead consider providing additional public services to stem population declines. Most migration studies use a fairly large geographical unit like states, counties, and urban areas; our study contributes to the literature by studying migration at the local government level (cities, villages, and townships).

Mangrove vegetation response to alteration in coastal geomorphology after an earthquake in Andaman Islands, India

The 2004 Sumatra-Andaman earthquake caused a devastating tsunami and shifts in land elevations, which had an instantaneous as well as the long-term impact on the Andaman Islands. Especially, the coastal uplift and subsidence has altered the tidal regime, leading to mass mortality and degradation of mangroves (∼150 sq. km). The tectonic shift in the Andaman archipelago has created three broader geographical units: North Andaman with coastal uplift (∼1.3 m) – sea level drop (SLD) scenario; Middle Andaman with minor to no change in coastal geomorphology – no change scenario; and South Andaman with coastal subsidence (∼1.1 m) – sea level rise (SLR) scenario. How these changes in the tidal regime have influenced mangrove vegetation remains less understood. Therefore, we assessed mangrove vegetation characteristics across the Andaman Islands to examine mangrove response to coastal uplift and subsidence after 18 years since the disturbance. The uplifted sites mimicking the SLD scenario appeared more destructive to mangroves (tree density - 771/ha) in comparison to subsided sites mimicking the SLR scenario (tree density - 937/ha), while no change sites with the highest tree density (1288/ha) showed lowest impacts to mangrove. The species diversity (H’=Shannon-Wiener index) within recruitment cohorts (seedlings: <1 cm DBH and saplings: ≥1 cm and <5 cm DBH) sapling (H’=1.57), and seedling (H’=1.48) was significantly higher than the trees (H’=1.20). Rhizophora apiculata and Ceriops tagal were the dominant species (cumulative IVI 60%) found in survived mangrove forests demonstrating their high resilience for sudden changes in tidal regime. Our study provided a comprehensive account of mangrove vegetation response after a major disruptive event in 2004. Long-term monitoring of these novel sites would enhance our understanding of mangrove response to sea level change, which can also provide significant insights for the restoration and management of mangroves.

The (complex) effect of internet voting on turnout: Theoretical and methodological considerations

AbstractThe adoption of remote internet voting can be a rather complex reform. In theory and praxis, geographical units can either decide to adopt or not to adopt i‐voting. Those that adopt it can differ in the mode of i‐voting adoption (i.e., internet‐only voting, or also in‐person voting), and the timing of adoption (i.e., some geographical units might adopt it earlier than others). How does the decision to adopt internet voting, the mode of adoption and the timing of adoption influence turnout? Using data spanning the municipal elections in Ontario, Canada from 2000 to 2018, we try to answer these research questions. Generally, we find that allowing internet voting, regardless of the availability of in‐person voting, does not influence turnout over the long term. However, we do find a temporal pattern, in that some of the most participatory municipalities adopted internet voting the earliest. We also detect a novelty effect; a sizable increase in turnout during the first ever election in which internet voting was introduced, which vanished in following elections.

Real-time forecasting of COVID-19-related hospital strain in France using a non-Markovian mechanistic model.

Projects such as the European Covid-19 Forecast Hub publish forecasts on the national level for new deaths, new cases, and hospital admissions, but not direct measurements of hospital strain like critical care bed occupancy at the sub-national level, which is of particular interest to health professionals for planning purposes. We present a sub-national French framework for forecasting hospital strain based on a non-Markovian compartmental model, its associated online visualisation tool and a retrospective evaluation of the real-time forecasts it provided from January to December 2021 by comparing to three baselines derived from standard statistical forecasting methods (a naive model, auto-regression, and an ensemble of exponential smoothing and ARIMA). In terms of median absolute error for forecasting critical care unit occupancy at the two-week horizon, our model only outperformed the naive baseline for 4 out of 14 geographical units and underperformed compared to the ensemble baseline for 5 of them at the 90% confidence level (n = 38). However, for the same level at the 4 week horizon, our model was never statistically outperformed for any unit despite outperforming the baselines 10 times spanning 7 out of 14 geographical units. This implies modest forecasting utility for longer horizons which may justify the application of non-Markovian compartmental models in the context of hospital-strain surveillance for future pandemics.