Geo-Resolution 2022 Poster Session

Welcome to the Geo-Resolution 2022 poster session. Students from the region submitted posters on a wide range of research topics connected to geospatial sciences or using geospatial tools.

The award committee, led by Professor Ness Sandoval, selected the following posters for the top three prizes:

• First place: Laura Gray, University of Illinois Urbana-Champaign, “Impacts of global climate change on total runoff”
• Second place: Pengfei Ma, Missouri University of Science & Technology, “Pipeline Leakage Detection by Mapping Vegetation Stress Indices from Hyperspectral Imaging”
• Third place: Jenni Nugent, University of Illinois Urbana-Champaign, “Monthly Virtual Blue and Grey Water Transfers on the U.S. Electric Grid”

We would also like to give special recognition to the winners of the “audience favorite” voting:

• Rana Das, University of Missouri-Columbia, “Actionable AI System for Animal Farms Identification”
• Harris-Stowe State University, “The Darkside of Light Series”: 
  • • Trena Mirzani Harris, The Darkside of Light: Geospatial Analysis of Maternal Depression as Affected by light pollution and proximity to Maternal Health Resources
    • Camara Macon, “The Darkside of Light: Geospatial Analysis, the intersection of Gestational Light Pollution, and Childhood Anxiety”
    • Inaya Smith, “The Darkside of Light: Geospatial Analysis of Gestational effects of light pollution on low birthweight and prematurity in the Chicago-metro area”

Congratulations to our 2022 winners!

We hope that you enjoy reading about the interesting research of students in our geospatial ecosystem.

Geo-Resolution 2022 Student Posters

Poster 1: A Framework to Evaluate Resiliency of Water Resource Systems Using Biophysical Models and Machine Learning

Authors: Umanda Abeysinghe, University of Missouri-Columbia
Abstract: “Maintaining healthy ecosystems is vital for sustainably accessing land and water resources for human appropriation. Hydrologic cycle, which involves the continuous circulation of water in the land-ocean-atmosphere system, helps in supporting various ecosystem functions and services. Of the many processes involved in the water cycle on the land surface, the most important are evaporation and transpiration, precipitation, and runoff. Although the total amount of water within the cycle remains essentially constant, its distribution among the various processes is continually changing both spatially and temporally. With rapid environmental change that affect the water cycle, assessment of key processes, both natural and managed, that affect the water cycle is an essential first step. In this presentation, we propose to (i) develop a process-driven watershed-scale hydrologic model to simulate key processes, (ii) evaluate the performance of the model in predicting key fluxes such as runoff and plant water use by comparing with observed values, and (iii) implement Artificial Intelligence (AI) based techniques to identify vulnerable regions across the landscape. While the model we propose here is process-based, the key processes and fluxes cannot be tested and validated everywhere due to limited observations. Because the biophysical processes driven by the water cycle are not fully understood, and the available data are incomplete and noisy, the AI techniques we propose will provide alternative and complementary ways. We propose to implement our methods in two continental-scale River Basins: (i) the Mississippi River Basin – a heavily managed landscape in the US, and (ii) the Congo River Basin – a largely pristine landscape but vulnerable to human appropriation. Our goal is to develop a modeling framework that combines process-based models and AI techniques to understand key processes that affect the water cycle, their impact on ecosystem services and develop plans to manage earth resources sustainably. Disclaimer: Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the U.S. Department of Agriculture. USDA is an equal opportunity provider and employer.“
Poster Link: https://drive.google.com/file/d/19PUAX4SUXfz1EZZmcbZYV1CJIyd9I3UT/view
Video Link: https://drive.google.com/file/d/19GW7ttgPg6V4k2VblPNR4Idsno5NBJu0/view

Poster 2: AgLapse: A spatiotemporal visualization tool for agricultural crops in the United States

Authors: Sourav Bhadra, Saint Louis University.
Abstract: The United States is a major producer of Corn, Soybean, Barley, and Oats within the world. Due to the extreme variation in climate, landscape, and the availability of natural resources in the United States, these crops show distinct spatial pattern on planting area, harvesting rate and the yield. The variation can also be observed in the temporal scale as the pattern of cultivation, management practices and genetic diversity of crops has significantly shifted from the last century. The United States Department of Agriculture (USDA) has been collecting the cultivation and yield data about different crops at county level from as early as 1900. However, the availability of such intense dataset is limited to static maps and tables, which are hard to interpret along the spatiotemporal circumstances. Dynamic and interactive visualization of such dataset in terms of both spatial and temporal context can reveal interesting patterns of crop productivity across the United States. AgLapse is a web-application that provides easy-to-use spatiotemporal visualization tools for the four major agricultural crops in the United States. It is developed with open-source geospatial tools in Python, and has two vital features, i.e., Ag-Status that shows the spatial pattern in terms of different years, and Ag-Trend that visualizes the temporal trend of productivity in different locations. Policy makers along with growers can make informed decisions by utilizing the spatiotemporal trend of such major crops through using the AgLapse. Additionally, AgLapse showcases the applicability and efficiency of open-source geospatial tools in similar use-cases regarding environmental sustainability.
Poster Link:https://souravbhadra-ag-lapse-ag-lapse-app-1l9sgf.streamlit.appVideo Link: https://drive.google.com/file/d/1k5aiwXMg__4ldcVTZ7Wve7adlTZ-A3un/view

Poster 3: An inferential GIS assessment of vegetative community dynamics

Author: Christian Blue, Aliya Fleeks, Childress, Bonsasa, Harris-Stowe State University.
Abstract: “A vegetated community benefits the earth through the capacitation of the environment, such as moderating climate, temperature, humidity, moisture and buffering winds, which protect built environments. For example, trees play a huge role in regulating community temperatures, soaking up toxins within the environment, and supports healthy ecosystems. This leads to the quality of citizens’ health, air quality, and neighborhood vibrancy. A New York Times article showed a correlation between tree discrepancies in Black and Brown neighborhoods with lower median salaries compared to more affluent and white neighborhoods. Locally, the St. Louis Mayor’s Office is facilitating a community-driven planning process designed to promote equitable development in six neighborhoods surrounding the new National Geospatial-Intelligence Agency (NGA) West site in Saint Louis, Missouri: St. Louis Place, JeffVanderLou, Carr Square, Columbus Square, Old North St. Louis, and Hyde Park. Our team hypothesizes that these lower socioeconomic status neighborhoods lack the proper amount of foliage which can lead to temperature degradation, lower quality of life, and declining property values. We will use geographic information system (GIS) mapping to measure the amount of tree canopy and crown density. We will also evaluate and characterize micro-climates in the neighborhoods. As a result of our study, we plan to design an app based on the parameters of the metrics that were measured in the neighborhoods identified for this project. This app will be supported by a device that will advise the residents in the six neighborhoods what appropriate trees and vegetation could monitor climate and lower temperature extremes.“
Poster Link: https://drive.google.com/file/d/1POsurZSQKQR0_Ha06BdtPQPvVZKHC7zb/view
Video Link: https://drive.google.com/file/d/1TMdBvlIIAqchf5A7hGKsYTfVy96NFb41/view

Poster 4: GIS as experimental archaeology: a soil erosion model of a prehistoric pastoral corral in Tibet

Author: Xinzhou Chen, Washington University in St. Louis.
Abstract: “Recent archaeological research of ancient pastoralists’ settlement pattern and their long-term ecological legacy greatly enriches our understanding of the human-environment interactions across the globe. Following this line of inquiry, we conducted an archaeological survey in Lhoka, Tibet, and discovered a prehistoric pastoral corral in the mountainous regions of the Yarlung river valley. Based on a high-resolution drone DEM, we simulated the dynamics of soil erosion around the archaeological site. To investigate the anthropogenic impacts of corral building activities on the local landscapes, we further removed the archaeological sites on the DEM using Kriging interpolation. The soil erosion model was run again under this hypothetical scenario that the ancient corral never existed. Our results suggest that the architectural building activity of ancient humans may create a small niche by altering local soil erosion and deposition dynamics. The anthropogenic geomorphological stability may facilitate the emergence of pastoral hotspots that are used for millenniums.”
Poster Link: https://drive.google.com/file/d/1IKheuF3twDoSwq42JyYUsiOTie6oeD8e/view

Poster 5: Actionable AI System for Animal Farms Identification

Author: Rana Das, University of Missouri-Columbia.
Abstract: “Too many animal farms (units) in certain areas can cause imbalanced manure nutrients and higher biosecurity risk, leading to the need of practical tools for farmers and engineers to improve farmstead planning and watershed-based manure management. A team of university faculty and professionals has been developing a robust artificial intelligence (AI) system to create timely information from massive remote sensing image data and layers of geospatial information, to identify animal farms and perform modeling work. This project aims to develop novel computational tools inspired to recognize animal farms by their shapes, dimensions and other factors, such as near-by structures and distance between them. Dataset of animal farm type and sizes collected from Missouri Department of Natural Resources were analyzed, and manually checked to confirm the farm structures, animal type, and barn sizes and to develop a machine-learning tool for farm recognition. The team is expanding the machine learning method to computationally link visual patterns with poultry farms. Using the results of such deep learning methods for classification of the components of animal farms, we will employ spatio-temporal modeling from experts’ gaze activities (using human tacit knowledge with minimal human-in-the-loop intervention) to characterize farm complexes and activities as factors in deciding animal farm size, and density. A web-based tool is being setup to present the capability of the tools and showcase the results. These types of information are important for new farm construction or expansion consideration, and effort and costs needed to optimize animal farming management and decisions.“
Poster Link: https://drive.google.com/file/d/1g-VkyyG_5dK9YzyH0yJnMyQaaAklKVyU/view
Video Link: https://drive.google.com/file/d/1CwW_iSZDcKzoj1HfbyjAt0C89TvBbdty/view

Poster 6: Hey! What type of drone is that? What it doing? Estimating the Location, Size, Geometry and Trajectory of a Drone from Two Video Streams

Authors: Matt Dreyer, Saint Louis University.
Abstract: “In recent years, drones have solidified as a common tool in several fields like farming, package delivery, construction, etc., and with their operations expected to grow, a more drone infested future is eminent. Now, methodologies need to be developed to prevent unauthorized drones from encroaching restricted airspace, like an airport. What is preventing a drone from potentially being sucked into an aircraft engine causing danger to passengers and crew onboard? Whether a drone has malicious intent or not, a counter unmanned aerial system (C_UAS) is needed to monitor and protect the airspace.

In this research, we proposes a low-cost C‑UAS solution by merely utilizing two video feeds to autonomously estimate characteristics (like GPS, size and geometry) of a drone flying in the field of view (FOV). Potentially, these characteristics can be utilized to estimate the future trajectory of the drone.”
Poster Link: https://drive.google.com/file/d/18XJ2o7ploJv_ll1bT8A-qqorJbDOTbX9/view
Video Link: https://drive.google.com/file/d/1DeETn93-DYdnVwcVyNztVg5fTw_1lZPP/view

Poster 7: From desert to oasis: A study of food deserts in urban environments using geospatial technologies with a goal of creating sustainable living communities.

Authors: Fischer, Reed, and Brown, Harris-Stowe State University.
Abstract: “Urban environments face numerous challenges for access to fresh, healthful food. Compared to all other census tracts, food desert tracts tend to have smaller populations, higher rates of vacant homes, and residents with lower levels of education, incomes, and higher unemployment.
In this study, we selected South (Lafayette Square) and North (Fairground Park) of the new NGA West facility. Both locations are less than 3 miles away from the new facility; this type of study will provide valuable information to NGA and residents in nearby communities.
We generated a manually-curated dataset of food stores coded with attributes such as walking distance and type of store (convenience versus grocery). Our results showed a stark divide: Lafayette square with a total population of 3032 residents had five grocery stores within a one mile radius versus one grocery store for Fairground neighborhood with 1559 residents.
Comparing census data in both neighborhoods seems to substantiate the correlation of food deserts with other social determinants. Median income of residents in the Lafayette Square neighborhood is $71,386 +/- 14,989 compared to median income of $26,042 +/- 12,598 in Fairground. Education levels showed a similar trend: 32.6% of all adults over the age of 25 in Fairground have not graduated from high school compared to just 6.2% in Lafayette Square.
Our research is ongoing and using ArcGIS for mapping and analytics will contribute to understanding and mitigating the impacts of food deserts. This data can be used to engage residents and create more sustainable communities.”
Poster Link: https://drive.google.com/file/d/1UuL3cRR_Vbr_DE4Pluu-r6kpUz5GSFkQ/view
Video Link: https://drive.google.com/file/d/1xPfAV3_75ccU0JgNr43g__CedXd5-1b_/view

Poster 8: 911 Response Times in Lafayette Square and Fairground Park

Author: Franklin, Kiara, Harris-Stowe State University.
Abstract: “Our goal is to create more sustainable and safer communities within the St. Louis area. This project will evaluate the current state of response times that are below the national and the state averages. Based on some research completed on the program, the findings show this program has actually hindered response times more than before it’s implantation. We will examine the response times beyond the current realm of study that speaks to these differences only based on state, county and local levels. This will create a Segway into addressing other community needs as crime is a deficiency symptom. Our project will correct the need for protection by providing better response times. This allows community members to feel safe and know assistance will arrive quickly when they call for help. This project will allow people within the community to participate in crime prevention and lead to a rise in community resources. We will use various GIS tools to analyze groups within our treatment and control group and pinpoint where crime hot spots are located. We will do a cost surface analysis, which uses GIS layers to find the path of least resistance. We will create a map and plot points where every police station, hospital, fire station, and EMT station are located. We can. use GIS to show where crime happens within the community. A historical analysis will be done to show where buildings are falling into dilapidation, so when we make a sight selection, we’ll have areas for a sub-station.”
Poster Link: https://drive.google.com/file/d/1bYwhlAVVOC-wz-VmImblaDHrzgGd2LUj/view
Video Link: https://drive.google.com/file/d/1q_RrTu1u9mY5-G6KpDjLcoew935Jllp2/view

Poster 9: Spatial Distribution of Tuberculosis (TB) and HIV incidences in Sub-Saharan African countries in 2018, using GIS technologys

Author: Gbaja-Biamila, Titilola, Saint Louis University.
Abstract: “GIS, although beneficial in improving health information systems, many countries are missing out on this technology, especially in Africa. Still, it is unclear how much GIS has been used in Tuberculosis (TB), and the human immunodeficiency virus/acquired immune deficiency syndrome (HIV/AIDS) research in Africa. TB and HIV/AIDS have significant reciprocal interactions, and global public health hazards, more than twenty-five percent of all TB deaths occur in Africa. Although HIV is not one of the world’s leading causes of mortality, it is still one of Africa’s top five. This study aims to inform researchers, and program implementers on the trends of incidence of TB and HIV spread in Africa, using GIS to map out the geographical distribution and analyze the trend while indicating the disease hotspots in Africa. This descriptive study was done between the 25^th and 6^th of May, 2022. It includes sub-Saharan African data from only 2018 WHO database on HIV, TB incidences, and the Proportion of HIV-positive new and relapsed TB cases on Antiretroviral therapy (ART) during TB treatment. The study concluded by showing an association between HIV and TB. Also noted was that there is no association between the TB detection rates and incidence rates of TB in Sub-Saharan Africa (SSA) countries. The study confirms the associations between HIV and TB trends in SSA countries using GIS.“
Poster Link: https://storymaps.arcgis.com/stories/cfcfb43919624961ab50ac23533e364c 
Video Link: https://drive.google.com/file/d/1ZweJ9CEvU-RAFfM-P3BITcIvKVrsjwnN/view

Poster 10: Impacts of global climate change on total runoff

Authors: Gray, Laura, University of Illinois Urbana-Champaign.
Abstract: “In this study, we first validate total (grid cell-level) runoff from the fully coupled Community Earth System Model (CESM) historical simulations against two observational-based datasets, and further analyze both grid cell-level runoff and urban subgrid runoff under future climate change scenarios. We calculated global annual average of monthly runoff from the period 1986–1995 for the validation and calculated bias and correlation coefficients between CESM and each of the datasets. Additionally, we analyzed future grid cell and urban runoff across three CMIP6 coupled Shared Socioeconomic Pathways and Representative Concentration Pathways — 2−4.5, 3−7.0, and 5−8.5 — and evaluated changes between the future period of 2041–2050 and the same past period of 1986–1995 for each scenario. Results show spatial consistency and robustness between the CESM simulations and both datasets. However, there is some spatial inconsistency in the areas highlighted as major runoff producers, such as the Amazon basin and Southeast Asia, as well as mountainous regions outside the United States. Grid cell-level runoff and urban runoff projections suggest that future hydroclimatic conditions will vary depending on the climate scenario. However, certain locations, such as Madagascar, Indonesia, and the Himalayan mountain range, consistently see decreases in both grid cell-level runoff and urban runoff across all scenarios, and locations such as Nigeria and Ecuador consistently see increases in both grid cell runoff and urban runoff across all scenarios. Our findings provide quantitative insights on hydrology representation in the global Earth system model and advance the understanding of the impacts of large-scale climate change on future local-scale urban runoff.“
Poster Link: https://drive.google.com/file/d/1z0GpRPs-wyHkL2hdSandlPmVblh_jSV1/view
Video Link: https://drive.google.com/file/d/1ty4hNxQ9bYlN8GJ15ny82eny301sUXFY/view

Poster 11: A Step Towards Building Inclusive Infrastructure

Authors: Grover, Savitri, Saint Louis University.
Abstract: “People with disabilities are often left behind when it comes to access to infrastructure. The situation is not different in India, where the geography is so varied that defining inclusive infrastructure can be challenging. Nevertheless, the most pressing question is how to achieve universal inclusive infrastructure in a country as large and populated as India. I am currently engaged in a dissertation project in which I am attempting to understand how to bridge the infrastructure gap among people with disabilities. In my project, I use both traditional data analysis and GIS tools to explore population, existing infrastructure, and geography to propose informed policy decisions, particularly relating to disability policies. Especially when it comes to building inclusive infrastructure, my work focused on identifying areas where there might be a need for evaluating, improving, and developing infrastructure.”
Poster Link: https://storymaps.arcgis.com/stories/164fd23bd2d2403aaf03303241696f7f
Video Link: https://drive.google.com/file/d/1UnmdLeQ_xyiah0VlUCaQQ7gwN7AXZpYP/view

Poster 12: Searching for Spatial Patterns on Venus: Kernel Density Analysis of Volcanoes on the Second Planet

Author: Hahn, Rebecca, Washington University in St. Louis .
Abstract: “Venus is home to thousands of volcanic landforms that range in size from less than 5 km to well over 100 km in diameter. With the NASA Magellan SAR (synthetic-aperture radar) FMAP (full-resolution radar map) left- and right-look global mosaics at 75 meter-per-pixel resolution, we developed a global catalog of volcanic edifices on Venus that contains ~85,000 features, ~99% of which are volcanoes <5 km in diameter. The distribution and orientation of these volcanoes may give insight into the size and shape of the underlying magma sources, the mechanisms of magma production, and the states of stress in the crust. One method for analyzing the distribution of edifices on a global level is by using kernel density functions. Here, we employ two different optimized bandwidth algorithms within our kernel density estimation, the sum of the asymptotic mean square error (SAMSE) and least squares cross validation (LSCV). The resulting spatial density maps reveal concentrations of volcanoes across Venus, which can then be correlated to local geological structures and geophysical datasets to identify patterns. Analysis of the high volcano density regions and Venus altimetry, geoid anomaly and crustal thickness datasets reveals that many larger volcanoes are associated with high topography, a high positive geoid, and thicker crust, whereas smaller volcanoes are associated with lower topography, a low positive geoid and thinner crust. These findings suggest a difference in magma source depth and volume between areas occupied by differing sized volcanoes on Venus.”
Poster Link: https://drive.google.com/file/d/1uv12pgN1vpuvoSfJCIH5zi75jAArKWVp/view
Video Link: https://drive.google.com/file/d/1uJEmrIjzVOEvT-VRSA0uHMLAI4IY3t3R/view

Poster 13: UAS-LiDAR Predicts 2D and 3D Spatiotemporal Patterns of Change in Retrogressive Thaw Slumps in Northern Alaska

Authors: Hall, Emma, University of Illinois Urbana-Champaign.
Abstract: “Predicting where and when retrogressive thaw slump (RTS) initiation and expansion will occur across permafrost terrain is a predominant research challenge, yet important for understanding the potential implications of landscape dynamics. Here, we use Uncrewed Aerial System Light Detection and Ranging (UAS-LiDAR) data to characterize fourteen active to stabilized RTS scars to evaluate the potential climate and topographic forces controlling decadal time scale RTS areal change and associated volumetric losses of ice and sediment on the Foothills of the Brooks Range. High-resolution UAS-LiDAR (acquired during 2021) and historic aerial and satellite image analysis (i.e., 1949–2020) was used to calculate the rate of (1) 2D terrain deformation, and (2) 3D volumetric loss of ice and sediments, for predicting the susceptibility and rate of continued RTS expansion. We used Generalized Boosted Regression Tree (BRT) models to predict patterns of RTS change using topographic and climatic predictor variables. Results indicate BRT models were able to predict 2D RTS expansion (R2 = 0.89, CV = 0.60) and 3D volumetric loss (R2 = 0.85, CV = 0.71) with high accuracy, where the most influential predictors were dElevation and headwall height, and dWetness and dPISR, respectively. Interestingly, climate variables only explained a relatively small proportion of the overall variance in both models, whereas topographic controls such as the hydrologic influence (wetness) and dElevation were the most important predictors of change. These findings demonstrate that high-resolution UAS-LiDAR or topographic information can be a powerful predictor of future patterns of Arctic landscape evolution within recently deglaciated terrain.”
Poster Link: https://drive.google.com/file/d/1c3sspVWVpE1yBDK5nW-B5u4uq6fmDJS1/view
Video Link: https://drive.google.com/file/d/1nD_UKNeHfaonLrYqpiEurW_1lfxkruO4/view

Poster 14: A spatial analysis of food inequality: correlations between food access and residential segregation in the St. Louis, MO-IL metropolitan statistical area

Authors: Heinrichs, Ellie, Saint Louis University.
Abstract: “A person’s well-being is, at least partially, derived from what they consume on a daily basis, and although healthy diets are promoted by government officials and healthcare workers, they are not possible when nutritious foods are not within reach of a community. In order to promote the health of communities through increased food access, one must first understand the underlying factors that are correlated with the lack of food accessibility. Access to supermarkets is used as a measurement of healthy food access in the United States, due to the widespread availability of supermarkets and high diversity of food options present in stores in comparison to smaller shops and convenience stores. This project seeks to explore the spatial relationship between access to healthy food and residential segregation in the St. Louis, MO-IL metropolitan statistical area, using data from the USDA Food Access Research Atlas 2019 and the American Community Survey 2019. Poverty, entropy, percent white, and percent black were used as independent variables in this analysis, with tracts identified as low-income with low access to food (LILA) as the dependent variable. Spatial autocorrelations were able to determine that the spatial distributions of both the independent and the dependent variables were not random but clustered. Geographically weighted regressions showed that LILA was positively associated with poverty and percent black, and negatively associated with percent white. LILA was only slightly positively associated with entropy.”
Poster Link: https://drive.google.com/file/d/1yO-4rOhF0mkR_CHilSynESrWNMDsmTJy/view
Video Link: https://drive.google.com/file/d/1f3aBXG6HB5ALEp6JufdpY4vz3NxO-uwj/view

Poster 15: On the effects of increasing spatial resolution of residential water monitoring systems through smart water meters: An experimental analysis

Author: Heydari, Zahra, University of Illinois Urbana Champaign.
Abstract: “Water flow in residential sectors is usually only monitored at the main inlet and little is known about how water is used among different end-uses since monitoring every water fixture can be both costly and intrusive. Smart water meters provide high temporal resolution data which integrated with non-intrusive machine learning methods can emerged as promising techniques to estimate the disaggregated contribution of water end uses. We explore how we can leverage high resolution smart water meter data to increase special resolution of the water monitoring system. We first collected 1‑second resolution indoor water use data from a residential single-point smart water metering system installed at a 4‑person household, in combination with ground-truth end-use labels based on a water diary recorded over a 4‑week study period. Then, we trained a supervised machine learning model to classify six water end use categories across different temporal resolutions and two different model calibration scenarios. Our findings show that data collected at 1- or 5‑second intervals allow for better end-use classification, particularly for toilet events; however, certain water end uses (e.g., shower) can still be predicted with acceptable accuracy even in coarser resolutions, up to 1 minute. Next, we develop a method to link fine resolution smart water metering system data to stagnation time in the same monitored study home. Our study advances current water stagnation time monitoring that often neglects the temporal stagnation variations among different household end-uses in premise plumbing, revealing areas of future work to integrate water quantity and water quality.”
Poster Link: https://drive.google.com/file/d/1fhXIVt_nB9zlWHMuDMCCro9kvj1-2ybX/view
Video Link: https://drive.google.com/file/d/1LCWl8L8iOT7yw51Xye32TsdfCNpOnDD8/view

Poster 16: Transfer Learning with Convolutional Neural Networks for Hydrological Streamline Detection

Authors: Jaroenchai, Nattapon, University of Illinois at Urbana-Champaign
Abstract: “Streamline network delineation is essential for various applications, such as agriculture sustainability, river dynamics, watershed analysis, etc. Recently, machine learning techniques such as the U‑net model have been applied for streamlining delineation and shown promising performance. However, their performance drop significantly when applied to different location. In this paper, we explore whether fine-tuning neural networks that have been pre-trained on large label dataset (i.e., ImageNet) will improve transferability from one geographic area to another. Specifically, we collected smaller NHD label datasets from the Rowan County and Covington River areas in the eastern United States. First, we fine-tune pre-trained ImageNet models on the Rowan County area and compared them with an attention U‑net model that is trained from scratch on the same dataset. We found that the DenseNet169 model achieved an F1-score of 85% which is about 4% higher than the attention U‑net model. Then, to compare the transferability of the models to new geographic area, we selected the three highest F1-score models in Rowan County area and further fine-tuned them with the data in the Covington area (for model transfer). Similarly, we fine-tuned the attention U‑net model from Rowan County area with the data in the Covington area. We see that fine-tuning ImageNet models achieve an F1-score of 71.87% in predicting the steam pixels in the Covington area, which is significantly higher than training the models from scratch in the Covington area or fine-tuning attention U‑net model from Rowan to Covington.”
Poster Link: https://drive.google.com/file/d/15oN5mvr84BdMaso4aCk84A0ROTqNImrK/view
Video Link: https://drive.google.com/file/d/1WLkyMC6QF-70S4CxD2YfRfzNlEWTc39o/view

Poster 17: Learning-Based Shadow Detection in Aerial Imagery Using Automatic Training Supervision From 3D Point Clouds

Authors: Kavzak Ufuktepe, Deniz, University Of Missouri — Columbia.
Abstract: “Wide area motion imagery (WAMI) poses many challenges for vision tasks due to large shadows. For object detection and tracking, when an object of interest moves under a hard shadow, the track is easily lost. Although there are many successful shadow detection methods that work well in indoor scenes, close range outdoor scenes, and spaceborne satellite images, they tend to underperform for oblique view WAMI data. We propose automatic shadow masking using self-supervised learning—removing the need for manual labeling to greatly increase the amount of training data. 3D point clouds and known solar angles are combined to create analytical ground-truth shadow masks with a two-pass shadow map technique from computer graphics. FSDNet, a recent deep network for shadow detection, is evaluated on our aerial imagery. Results show our automatic shadow mask self-supervision improves performance, which provides insight and a potential path to develop new deep architectures for shadow detection to complement vision tasks in WAMI.”
Poster Link: https://drive.google.com/file/d/15GVRgXEMgDdKJUo60UST0q1URTjEngm7/view
Video Link: https://drive.google.com/file/d/1znrdehfyqMCeCxGKHVDp08GyLlZXlnK6/view

Poster 18: Designing a Space-Time-Indexed Geospatial Big Data Repository for Climate and City Fitness Monitoring: Scalable Storage and Management

Authors: Khan, Solaiman, University of Missouri Columbia.
Abstract: “Geospatial big data offers great prospects for advanced research in different scientific spectrums, including urban planning, smart city development, climate monitoring, disaster management, public health, etc. Ever-growing geospatial big data, especially in the age of the Internet of Things (IoT), brings great opportunities to human society as well as different challenges to utilize those data in a useful way. The massive volumes of data require large and reliable data storage with high data scalability and easy data accessibility. This poster demonstrates the robust design, development, and techniques of open-source data repository systems for hierarchical space-time-indexed time-series data from sensor systems. This repository will leverage high-scalability database systems and facilitate efficient queries, easier access, analytics, and integration with other systems, especially with web applications, without arising any computational complexity. We have evaluated the performance characteristics of our proposed systems. On average, it reduces 97% of storage space without any data loss.”
Poster Link: https://drive.google.com/file/d/13SR_ZHtSWX_H_Hf8IsRs_Qx_wvw8sje_/view

Poster 19: Adding Context to Plaintext: Interfaces for Quickly Exploring Aerial Data

Authors: Krock, Timothy, University of Missouri.
Abstract: “Students often find themselves attached to a linux terminal searching a cluster for some arcane dataset known only to somebody who just graduated. We recently started to address that with easily accessible and human readable data collection synopses and bundle adjustment reports. The at-a-glance readability and convenient exploration has lead to the available data being analyzed much more than it would have been otherwise. Some of the summaries are hand-generated, but for incoming datasets, we have automated the process to meet the demands of a huge influx of drone data.”
Poster Link: https://drive.google.com/file/d/1Uw-_V08oyM2vVaZpmkJrf_hjcLVnYHim/view

Poster 20: Progress Toward Mapping and Modeling Arctic-Boreal Peatland Dynamics

Authors: Ludden, Caroline, University of Illinois Urbana-Champaign.
Abstract: “Northern peatlands contain between 300 and 600 PgC, representing a globally important pool of carbon and nutrients. Rising temperatures and changes in precipitation increase the likelihood of disturbances that threaten northern peatlands. Disturbances, such as fire, have direct implication for carbon-climate feedbacks. Therefore, it is imperative we understand the current and future carbon dynamics and the spatial distribution of these peatlands. However, our fundamental understanding of the spatial distribution has been largely limited to coarse spatial scales (≥500 mresolutions), thus representing a disconnect between typical heterogeneous patterns of burns and the heterogeneous distribution of peatlands across the Arctic Boreal Zone(ABZ )of Alaska. Here we developed a new ABZ high-resolution (10 m resolution) peatland map and parameterize the DOS-TEM model to predict changes in peatlands due to changing climatic conditions.”
Poster Link: https://drive.google.com/file/d/1sJ3CzzlsuGt2t6PASUR6UqW5tthSUgAr/view
Video Link: https://drive.google.com/file/d/1RE9hdXwARyZu9I1jUZd49VoqfaQFn_Q9/view

Poster 21: An integrated cyberGIS and machine learning framework for fine-scale prediction of Urban Heat Island using satellite remote sensing and urban sensor network data

Authors: Lyu, Fangzheng, University of Illinois at Urbana-Champaign.
Abstract: “Due to climate change and rapid urbanization, Urban Heat Island (UHI), featuring significantly higher temperature in metropolitan areas than surrounding areas, has caused negative impacts on urban communities. Temporal granularity is often limited in UHI studies based on satellite remote sensing data that typically has multi-day frequency coverage of a particular urban area. This low temporal frequency has restricted the development of models for predicting UHI. To resolve this limitation, this study has developed a cyberGIS framework encompassing multiple machine learning models for predicting UHI with high-frequency urban sensor network data combined with remote sensing data focused on Chicago, Illinois, from 2018 to 2020. Enabled by rapid advances in urban sensor network technologies and high-performance computing, this framework is designed to predict UHI in Chicago with fine spatiotemporal granularity based on environmental data collected with the urban sensor network and Landsat‑8 remote sensing imagery. Our computational experiments revealed that a random forest regression (RFR) model outperforms other models with the prediction accuracy of 0.45 degree Celsius in 2020 and 0.8 degree Celsius in 2018 and 2019 with mean absolute error as the evaluation metric. Humidity, distance to geographic center, and PM 2.5 concentration are identified as important factors contributing to the model performance. Furthermore, we estimate UHI in Chicago with 10-min temporal frequency and 1‑km spatial resolution on the hottest day in 2018. It is demonstrated that the RFR model can accurately predict UHI at fine spatiotemporal scales with high-frequency urban sensor network data integrated with satellite remote sensing data.”
Poster Link: https://drive.google.com/file/d/1r68dE3Ws6b9lfcsWzQnxyN5PyYojnyIX/view
Video Link: https://drive.google.com/file/d/1er_D0TYkTrIwjI1WIo2pOwxxUMH0ys2t/view

Poster 22: Pipeline Leakage Detection by Mapping Vegetation Stress Indices from Hyperspectral Imaging

Authors: Ma, Pengfei, Missouri University of Science and Technology.
Abstract: “Detection of natural gas leakage from underground pipelines is traditionally a time-consuming and labor-intensive task especially in vegetation areas. The ground vegetation, however, can serve as a natural sensor for leakage detection due to its response (stress) to the alteration in soil microenvironment. Hyperspectral imaging can record the physiological changes of plant leaves/canopies due to the stress mitigation activities. A field test was conducted to simulate the natural gas seepage to monitor periodically the health condition of the ground grass using an unmanned aerial system including a hyperspectral camera. The most effective plant stress indicators were selected and mapped in hyperspectral images to detect the potential difference between the control and gas leakage impacted grass trenches. MCARI was found to be the most sensitive stress indicator and NDVI showed a notable increase on the stress grass. But PRI that reflects the photosynthesis activities seemed insensitive to gas leakage induced stress. Overall, hyperspectral stress indicator can definitively detect the gas-induced stress on ground vegetation after 42 days of treatment.”
Poster Link: https://drive.google.com/file/d/19XW4GGrMe7idE_gacjTgYneoQG2OSJ7p/view
Video Link: https://drive.google.com/file/d/16Pj8kjALkSY94HAVYFlW59L5fNDj0V3G/view

Poster 23: The Darkside of Light: Geospatial Analysis the intersection of Gestational Light Pollution and Childhood Anxiety

Author: Macon, Camara, Harris-Stowe State University.
Abstract: “Circadian rhythms are a 24-hour biological clock that the body runs on to know when to give energy to and relax the body and is key to balanced mental and physical health. Pregnant women and their progeny represent especially important populations affected by light pollution. Circadian rhythms can be interrupted by shift work that involves working at night or being exposed to bright lights at nighttime. Previous research indicates that disruption of circadian rhythm in pregnant women leads to negative changes in health aspects such as metabolic status, feeding, and sleep in both the parent and offspring. These effects have been linked to mood disorders such as anxiety. We hypothesize that mothers with who live in zip codes with high levels of light pollution during gestation will have offspring that are more susceptible to anxiety disorders because of unhealthy light environment. We will therefore analyze geospatial light pollution and distribution of early childhood anxiety disorders. Data from zip codes with extreme and low light pollution in the Saint Louis metro area will be analyzed. These data seek to reveal the relationship between maternal light pollution and early onset anxiety in adolescence.“
Poster Link: https://drive.google.com/file/d/1GPHuaEccAM63eVNgrJjibycWtvMJsayE/view
Video Link: https://drive.google.com/file/d/18_N2g3lluInQeUNfsTtBACNaW4ssfk5b/view

Poster 24: Tick-Borne Disease in a Dynamic Landscape: Does Exposure Risk Change Over Time?

Authors: McFarland, Derek, University of Illinois Urbana-Champaign.
Abstract: “Human-mediated landscape change is a major driver of tick-borne disease (TBD) risk. Amblyomma americanum, the lone star tick, is the most prevalent vector of emerging TBDs in the southeastern and central U.S. It is responsible for the transmission of several pathogens to humans, which are supported by populations of white-tailed deer (Odocoileus virginianus) that serve as key reservoir hosts (i.e., help pathogens to reproduce and transmit to uninfected ticks). An opportunity to explore the effects of landscape change on TBDs transmitted by lone star ticks exists in the St. Louis, MO region, which represents a typical urban-to-rural human land-use gradient. In 2009–2011, a research group led by my doctoral thesis advisor repeatedly sampled ticks and tick-borne pathogens from 31 sites spanning this gradient in human land-use. For my dissertation research, I will sample these same 31 sites across this human land-use gradient to investigate changes in the abundance and pathogen infection rate of lone star ticks, and the abundances of key reservoir hosts such as white-tailed deer; I will compare my findings to measurements taken a decade prior to understand how ongoing changes in land-use have influenced disease risk to humans. I will use ArcGIS to quantify changes in the distribution of human land-uses surrounding each study site, and Landsat imagery to quantify changes in vegetation cover between the two sampling periods. This research will produce novel insights into how ongoing land-use change may influence vector-borne disease risk relative to historical estimates, a topic that has not been investigated previously.”
Poster Link: https://drive.google.com/file/d/15fIigg2EA1lX3aJPO6cmDJygxR72zcES/view
Video Link: https://drive.google.com/file/d/1qTPuDphefVjhNTln7sIuJGI409IC8wFN/view

Poster 25: The Darkside of Light: Geospatial Analysis of Maternal Depression as Affected by light pollution and proximity to Maternal Health Resources

Authors: Mizani Harris, Trena, Harris-Stowe State University.
Abstract: “Postpartum depression is a debilitating mental disorder with a prevalence between 5% and 60.8% worldwide. The disease manifests as sleep disorders, mood swings, changes in appetite, fear of injury, serious concerns about the baby, much sadness and crying, sense of doubt, difficulty in concentrating, lack of interest in daily activities, thoughts of death and suicide. Feelings of hopelessness in severe cases of illness can threaten life and lead to suicide it is a factor that causes 20% of maternal deaths in the course after giving birth. In addition, issues such as fear of harming the baby (36%), weak attachment to the baby (34%) and even, in extreme cases, child suicide attempts have been reported These symptoms have serious effects on family health. Therefore, susceptible people need to be identified before delivery to receive proper care measures and. We hypothesize that the risk of post-partum depression is higher in St. Louis city versus St. Charles County dependent on the accessibility to resources.

Using the density and distance of local facilities within the area of interests and reported percentage of mothers with postpartum depression at a county level within these areas. We will compare on a county level using ARCGIS. We will also measure the distance it takes to from mothers to these facilities as a measure a accessibility. This data will shed light on the impact of the new mothers who don’t have access to post-partum care and find ways to make these resources accessible.“
Poster Link: https://drive.google.com/file/d/1mRsYg2dPgFyzo6N2gKAIdEC0ChrTCCu0/view
Video Link: https://drive.google.com/file/d/1M7TLvAn7zo3KcwvcH4_zPigtoBGFvfG-/view

Poster 26: Local Indicator of Spatial Agglomeration between Newly Opened Outlets and Existing Competitors on a Street Network

Author: Morioka, Wataru, University of Illinois at Urbana-Champaign.
Abstract: “Distance from competitors is a key factor in retail site selection and profitability. To understand the locational tendency that each newly opened outlet locates close to or far from existing competitors in a target area, a specific method is needed. Hence, this study aims first to develop a statistical method to discover the local spatial associations between newly opened and existing point-like outlets on a street network. We achieve this objective by extending the network local cross K function. The second objective is to evaluate the practicality of the proposed method by applying it to restaurants in a trendy district in Tokyo. Specifically, this study focuses on answering two questions: first, whether each newly opened restaurant is closely located to existing ones or not and, second, whether each existing restaurant attracts newly opened restaurants or not. The results show that the method is useful for revealing the location tendencies of retail outlets toward competitors.”
Poster Link: https://drive.google.com/file/d/1K0HkODcOsg7KRfB_8vWLE9uVg9YMCVjW/view
Video Link: https://drive.google.com/file/d/1FCJbiD8gxxb1qvbnX4_hJGg1aIvN0Gbp/view

Poster 27: Monthly Virtual Blue and Grey Water Transfers on the U.S. Electric Grid

Author: Nugent, Jenni, University of Illinois Urbana-Champaign.
Abstract: “Water used by power plants is transferred virtually from producers to consumers on the electric grid. This network of virtual transfers varies spatially and temporally on a sub-annual scale. In this study, we focused on cooling water used by thermoelectric power plants and water evaporated from hydropower reservoirs. We analyzed blue and grey water virtual water flows between balancing authorities on the United States electric grids from 2016 to 2020. Transfers were calculated using water use volumes reported in Form EIA-923, water consumption factors from literature, and electricity transfer data from Form EIA-930. Virtual blue and grey water transfers follow seasonal trends, reaching a maximum peak during the summer months and a smaller peak during the winter. Notable blue water transfers occur between MISO and PJM in the east and from Arizona to surrounding regions in the southwest. Virtual grey water transfers are greatest in the eastern United States where older, once-through cooling systems are still in operation. Understanding the spatial and temporal transfer of water resources has important policy, management, and equity implications.”
Poster Link: https://drive.google.com/file/d/1I9DWH7BzIBk7UnzAakXME3KKHwaOJbml/view
Video Link: https://drive.google.com/file/d/1EJ2NAzJw1jSReHXTTl5CHeX8jwhpLhNq/view

Poster 28: Experimenting with Localization Management Functions in 5G Core Networks

Authors: Pinto, Andrea, Saint Louis University.
Abstract: “Localization has achieved great attention in 5G networks, pushed by standardization. However, experimentation in 5G networks lacks the integration of network function modules designed for localization. We present our implementation of the 5G Localization Management Function. It complies with the 3GPP standard and OpenAirInterface, the most advanced framework that implements a full 5G-New Radio stack. We show that we are able to extend the functionality of OpenAirInterface, enabling location services. Finally, we demonstrate that the tool’s performance satisfies the 5G Key Performance Indicators required by 3GPP for localization.“
Poster Link: https://drive.google.com/file/d/1HasNcjUxoXcxpdb0U2YN7I62DVhT2_ui/view
Video Link: https://drive.google.com/file/d/1Nac4TW5DS7365ckVUrTCwc4F3YHlNFab/view

Poster 29: Decadal-scale modeling of patch-scale shrub cover change in the Seward Peninsula, Alaska

Authors: Schore, Aiden I. G., University of Illinois at Urbana-Champaign.
Abstract: “Warming at high latitudes has been linked with increasing plant productivity as widespread patterns of “greening” (i.e., increased NDVI trends) have been observed across the Arctic, due in part to the expansion of tall shrubs. Understanding the spatiotemporal patterns that control shrub expansion is key to predicting future biophysical and biogeochemical climate feedbacks. Here, we map patch-scale shrub expansion between 1950 and 2017 using high-resolution 3 m USGS historical aerial photography and PlanetScope‑3 data, respectively, across the central Seward Peninsula of Alaska. Patch-scale tall shrub map products were used to model the spatially explicit patterns of shrub change over time using climate, topographic, and environmental predictor variables within a probabilistic quasi-logistic regression model. Patch-scale shrub mapping and models were validated by high-resolution 1m WorldView 3 imagery and site-level observations. Preliminary results indicate (1) high correspondence between shrub maps and models with independent reference data, and (2) patch-scale shrub cover change is controlled by permafrost presence, elevation, and proximity to waterbodies (i.e., streams, ponds, lakes), as new shrubs tend to grow in permafrost-poor, wet soils running downslope on mountainous terrain. This research improves our understanding of local to regional-scale trends in shrubification using widely available remote sensing products, while providing a framework for widespread implementation of such patch-scale shrub models spanning topographically diverse tundra landscapes spanning the pan-Arctic.”
Poster Link: https://drive.google.com/file/d/16ZhBzSioWSciCH5mwrgsDHzzI6XdXLq2/view
Video Link: https://drive.google.com/file/d/12MPYn3AhKz-mRgDW3TCJ5uMeKWSUmx‑F/view

Poster 30: The Relationship of Underreporting of COVID-19-Related Mortality and County Death Investigation Offices across Missouri

Authors: Scroggins, Stephen, Saint Louis University.
Abstract: “COVID-19 deaths are estimated to be 2 to 3 times higher than reported. Often, cause of death is established and reported by a county’s medicolegal death investigation office. Underreporting of COVID-19 deaths undermines prevention measures and results in less effective resource allocation. We applied spatial methodologies to identify counties in the state of Missouri where deaths due to COVID-19 were significantly lower than expected. Logistic regression modeling then revealed these counties were more likely to have a smaller number of death investigation staff members even when controlling for hospital locations and total populations. Our results suggest underreporting of COVID-19 may be due to limitations in workforce and providing additional resources such as extra staff or supplemental protocols may alleviate this problem.”
Poster Link: https://drive.google.com/file/d/1w3eNoOPNTWwn2DOS0YjhhonQ9BB1NF2n/view

Poster 31: Using CyberGIS to Predict Solar Fields Sizes for Glycerol Conversion to Value Added Chemicals

Authors: Sibal, Adam, University of Illinois Urbana-Champaign.
Abstract: “Historically, biodiesel produced from plant and animal oils has been utilized and blended with fossil fuel diesel to reduce emissions and is produced at 75 facilities broadly spaced across the U.S. each with varied levels of production. The primary process converts oils to compounds known as biodiesel, fatty acid methyl esters (FAME), with about a 10-weight % glycerol. Since biodiesel production in the U.S. ramped up after 2005, the value of further purified co-produced glycerol has plummeted from $0.90 per kg to $0.24 per kg. Electrochemical conversion of this glycerol to value-added fuels and chemicals can both increase the economic viability of biodiesel production as well as decrease the greenhouse gas emissions associated with chemical production. The greenhouse gas emissions can be further reduced by utilizing distributed renewable energy to supply the electricity required to convert the glycerol to value-added chemicals. This work utilized Python integrated CyberGIS, a mathematical electrolyzer model with predicted power requirements and geospatial data from the National Renewable Energy Lab’s (NREL) National Solar Radiation Database (NSRDR) to model the required solar array size to generate enough electricity to complete the electrochemical conversion of glycerol and water to value added fuels and chemicals at each U.S. biodiesel facility. The resultant code returns an interactive Folium map showing the location of each facility and the size of the solar array required.”
Poster Link: https://drive.google.com/file/d/1XA37EXznM18A4aBQWw3eRkVQCAAlRqTD/view
Video Link: https://drive.google.com/file/d/1Zd1ze5wfou7A7M3Pkp3AyIpA5cXjmpZF/view

Poster 32: The Dark Side of Light: Geospatial Analysis of Gestational effects of light pollution on low birthweight and prematurity in the Chicago-metro area

Authors: Smith, Inaya, Harris Stowe State Universiy.
Abstract: “Light pollution is the presence of unwanted, inappropriate, or excessive artificial light and is a growing threat to biodiversity. Global light pollution has been increasing exponentially during the past century. At all levels the industrialization of artificial light in everyday use has become normalized. Unfortunately, the spread of anthropogenic light has unintended physiological consequences, especially in urban areas. Light pollution (LP) disrupts circadian rhythms by altering downstream signaling pathways that can trigger a variety of chronic disease states from infertility to metabolic syndrome in mammalian species.

The city of Chicago, IL is highly light polluted. The March of Dimes reports, the state of Illinois scores a C‑, the county of Cook scores a D+, and DuPage County scores a B‑, highlighting between county differences in preterm birth and low birthrates. Our study will include Markham, Evanston and Naperville, IL. We will investigate the role of light pollution in the distribution of preterm birth and low birth weight rates. We will determine the interactions and correlations between geographic location, and urbanization as affected by light pollution. We hypothesize that the socioeconomic standings of an area will contribute to the rate of low birth weights and preterm births. We also hypothesize that the proximity to the main source of light pollution will influence low birth weight and preterm births. With these findings, we will be able to bring awareness to this growing problem, possible introducing new policies and resources to help combat this global threat to biodiversity.
”
Poster Link: https://drive.google.com/file/d/17CwMipLSlB91CysXXOtShN9v1sbBuitK/view
Video Link: https://drive.google.com/file/d/1s7POYSide6gQNPPBiMaQQwjupiTzah4S/view

Poster 33: Nitrogen Risk Hazard: National to State Scalar Differences

Authors: Spell, Sam, University of Missouri.
Abstract: “Nitrogen in the form of fertilizer is a vital part of farming today and ensures that our farms can continue to produce enough food for the population. Whenever Nitrogen is not utilized efficiently by crops, or is too abundant in an area, there is an increased probability that it might end up in the water supply – either surface or groundwater. A machine learning estimation was completed at the national level of the United States to produce a nitrogen risk map in drinking water. In this models development, there were certain variables that were not available at a national level able and thus not represented or incorporated. The purpose of this phase of the project is to develop a more detailed nitrogen risk map at the state level for Missouri and compare the results between the national and state levels. This will give us an idea of the confidence associated with the national level map, as well as start to inform this research of where hotspots of nitrogen risk are located. Understanding nutrient application, as well as runoff, can inform inquiries into rural adverse health effects. This research is part of a larger Nitrogen risk landscape analysis project that includes socioeconomic disparities as well as health effect components using medical data. The goal of this project is to use it as a foundation for further research and analysis of these interrelationships in agricultural areas and associated rural community health.”
Poster Link: https://drive.google.com/file/d/18v06qTeFOVQzohfbA1MwlhiO2xvpv8TP/view
Video Link: https://drive.google.com/file/d/1fT898ByhUE9sV0K5pE1XhIe_tGSBZ4iw/view

Poster 34: Where are long-acting reversible contraceptive (LARC) users in Ethiopia? An exploratory spatial study

Authors: Teni, Mintesnot, Saint Louis University.
Abstract: “Introduction: One in four pregnancies in Ethiopia were unintended, and the unmet need for family planning stands at 22%. One strategy to prevent unintended pregnancy is to increase access and utilization of an effective contraceptive method like long-acting reversible contraceptives (LARCs). However, the uptake rate of LARC in Ethiopia is still relatively low compared to the lesser effective short-acting contraceptive methods.
Objective: To conduct an exploratory geospatial study on the uptake of LARC in Ethiopia.
Methods: Data from the 2019 Performance Monitoring for Action (PMA) Ethiopia survey was used to assess the spatial variation of LARC uptake. Spatial lag regression model was developed to assess the association between zonal level demographic variables and LARC uptake in Ethiopia.
Results: Clustering of low uptake of LARC was observed in the eastern regions (Afar and Somali) of the country. A statistically significant spatial autocorrelation was found for LARC uptake (Moran’s I= 0.308, p<0.001). Furthermore, the spatial lag regression model showed that the percentage of Muslim population (β=-0.075, p=<0.001) and no formal education (β=-0.127, p=0.01) are associated with a lower uptake of LARC.
Conclusions: The Ethiopian government, policymakers, and non-governmental organizations could develop interventions targeting areas (Afar and Somali regions) with low uptake to improve access to LARC. Religious leaders of Muslim population could play a significant role in increasing the acceptance of LARC among their followers. Other stakeholders could also develop health education regarding contraceptive methods tailored to the Muslim population and population with no formal education to increase their knowledge of LARC.”
Poster Link: https://drive.google.com/file/d/1udrj9olJEgOLofRbyl2fSfeY3nFtc7_C/view
Video Link: https://drive.google.com/file/d/1S1HsMPSB1bqZhaAaeYA6PLTCb5uCUewZ/view

Poster 35: CyberGIS-ABM for Large Scale Emergency Evacuation

Authors: Vandewalle, Rebecca, University of Illinois Urbana Champaign.
Abstract: “Recent crises, such as the COVID pandemic and western U.S. wildfires, underscore complex interrelationships between human actions and impacts for processes such as disease spread and emergency evacuation. Agent-based models (ABMs) are powerful tools for modeling these types of complex human-environment interaction processes. In ABMs, actions and interactions are encoded for individual entities, and higher processes develop as a result of interactivity. As networks are often fundamental to human dynamics, such as contact networks driving disease spread or transportation networks channeling evacuation traffic, spatial structures of such networks need to be considered for geospatial modeling. To represent realistic problems, it is also important to be able to represent large geographic areas, massive agents, and numerous dynamic interactions between agents and their environment. However, scaling up ABMs requires significant computational resources – especially for spatially explicit models.

This research aims to establish CyberGIS-ABM, a large-scale network-based ABM for exploiting high-performance computing (HPC) to support efficient and large-scale emergency evacuation. In CyberGIS-ABM, spatial network characteristics are targeted to optimize dynamic load-balancing, i.e. the efficient distribution of computational work across multiple computers. Using C++, MPI, an object-oriented design, and parallel software design patterns, the CyberGIS framework runs many iterations of large-scale agent-based simulations as well as collates and presents the simulation results. CyberGIS-ABM was designed to achieve adaptability, extendibility, and reusability by separating code interfaces from implementation aspects. Finally, CyberGIS-ABM is built and evaluated using multiple case studies focused on several wildfire evacuations in the western U.S.”
Poster Link: https://drive.google.com/file/d/1Qmfj4UcpT9yiwpMAPW4LmVwfZ8dBqQRU/view
Video Link: https://drive.google.com/file/d/1KN6tbQrkKtOyY2HIbb5LcJ-fGYqwR4rx/view

Poster 36: Visualising Sorghum Genotype x Phenotype Relationships Using StylEx

Authors: Venkatesh, Varun, Saint Louis University.
Abstract: “Although deep neural networks (DNNs) are remarkably good at complex tasks, how they make decisions can be difficult to understand. For instance, what makes a deep convolutional neural network think an image comes from one class and not another? Perhaps the most standard visualization approach for this classification task is to make a heatmap that highlights the regions that most impacted the classification. We have previously used such heatmap approaches to understand the genotype x phenotype relationship in biomass sorghum, by training DNNs to predict whether an image has a particular mutation or not, and then highlighting what the model focused on to make that decision. These heatmaps, however, only highlight the spatial location that was important, and do not disentangle the characteristics of those regions. Is it the color of the leaves that’s important? Their shape? In this poster, we will compare the previous heatmap-based visualization approach with the StyleEx approach proposed early this year by researchers at Google. StyleEx trains generative models to produce counterfactual examples – given an image that does not have the mutation, what would it look like if it instead did have the mutation? – and then automatically uncovers the most important visual attributes that show up in these counterfactuals to explain what is most important in determining which class an image belongs to. We show this provides richer information than heatmaps and has more potential to help biologists uncover unknown relationships in between plant genotypes and phenotypes.”
Poster Link: https://drive.google.com/file/d/1xh1uwwbUICRNZSwRnqo-X6YBvYT9GEtm/view
Video Link: https://drive.google.com/file/d/1rKcJ0iB3CfU9jQ7R_jWnU8LJSNSjnHBp/view

Poster 37: Assessing the relationships between COVID-19 prevalence and population characteristics: a case study on Chicago, USA

Authors: Wang, Ryan, University of Illinois Laboratory High School.
Abstract: “This research aims to comprehensively examine the relationships between COVID-19 cases and rates and the population size and density with a case study on the Chicago area. The COVID-19 data we used include the weekly case count and rate since March 1^st, 2020 for each ZIP Code Tabulation Area.The population data we obtained is from the Chicago Data Portal. We ran Pearson’s correlation coefficient (CC) analysis and bivariate regression between the cases and the population. Our results show high temporal variability in the association between the two. For example, the CC between the number of new cases per week and the population size ranges from 0.08 to 0.74, with a median of 0.35 during the 128 weeks, and the CC between that case count and the population density ranges from ‑0.19 to 0.38, with a median of 0.06. In the regression models, both population size and density are insignificant for most weeks. In general, the association between the COVID-19 cases/rates and the population is weak. These findings are consistent with previous studies in China and the US but inconsistent with the findings from some other countries such as India, Malaria, Algeria, and Turkey. Our findings suggest that the association of communicable diseases with population size/density differs from that of chronic diseases. Therefore, geospatial analysis methods such as the detection of spatial clusters used for chronic diseases should not be directly applied to communicable diseases.”
Poster Link: https://drive.google.com/file/d/1Bk5xjG2dl2nFXsYUOVDay8VOWoNrPh4Q/view
Video Link: https://drive.google.com/file/d/1Iwb6LrZhOANNw46R3gR7W6vtplwAlDXn/view

Poster 38: Spatial disaggregation per unique temporal sequences in dynamic spatial accessibility to primary care in New York City

Authors: Park, Jinwoo, University of Illinois Urbana-Champaign.
Abstract: “Dynamic spatial accessibility, incorporating temporal dimension into spatial accessibility measurements, is in the spotlight because of significant temporal variations in urban phenomena (e.g., traffic congestion or floating population). Despite its importance, there are two significant shortcomings in healthcare access literature: 1) lack of consideration of the dynamic nature of spatial accessibility, and 2) limited support of dynamic spatial accessibility for policy making. Our study aims to improve the measurements of spatial accessibility with time-dependent data and cluster their temporal sequences to support policy making. Specifically, we measure hourly changes in spatial accessibility to primary care resources in New York City with the generalized two-step floating catchment area (G2SFCA) method. We then employ sequence analysis to cluster unique temporal changes focusing on locations. Our results reveal that the epicenter of sufficient accessibility moves around various locations due to its dynamic changes. In addition, we find that places can be classified based on the combination of temporal changes (i.e., stationary and non-stationary) and the different levels (e.g., sufficient, average, or insufficient) of accessibility. Therefore, our findings can suggest flexible policy implications for various needs of locations. For example, certain areas should have extended operating hours of primary care or additional hospitals in place. Given that decision-making is often based on limited resources, our study sheds light on how to make optimal allocation of healthcare resources to effectively reduce inequality of spatial accessibility.”
Poster Link: https://drive.google.com/file/d/19TZl_hzexjiLm7UU4rTdMTla__gbHdk3/view
Video Link: https://drive.google.com/file/d/1dLocPYSVOm4S8ooKzA6tR5BqhEHUbiBg/view

Poster 39: Investigating of Flood inundation dynamic using MODIS product and MERIT DEM

Authors: Haireti, Alifu, Saint Louis University.
Abstract: “This study investigating the flood inundation trend based on the optical Moderate Resolution Imaging Spectroradiometer (MODIS) product (500‑m Resolution Daily Global Surface Water Change Database during 2001–2016) and MERIT (Multi-Error-Removed Improved-Terrian) digital elevation model. Global floodplain mask and subbasins mask were derived from MERIT DEM in order to remove water pixels which are not caused by flood and detailed analysis of the sub-basin scale of flood dynamic. Flood inundation was investigated based on annual total cumulative inundation extent (ATCIE: the accumulative total spatial extent of the flooded area) and annual maximum inundation extent day (AMIED). Validation of floodplain mask and trend based on AMIED showed a good agreement with TCIE and trend derived from gauging stations data. Final results indicated that Flood trend increased in the North Hemisphere (above 40°N) in the recent 16 years, and this mainly caused by the increase of temperature and precipitation. On the other hand, flood trend was decreased in south Asia and Europe region during the study period.”
Poster Link: https://drive.google.com/file/d/1czsMqqQlsjy8pPfdLYE7PRKyNThaGwDg/view

Poster 40: Event-Aware Spatio-Temporal Modeling

Authors: Wang, Zhaonan, University of Illinois Urbana-Champaign.
Abstract: “The poster is mainly based on the research I’ve conducted during PhD, in which spatio-temporal IoT data and GeoAI techniques are employed to study anomalous events, including extreme weathers, pandemics, and accidents. Such events impact us in various ways. My research aim to measure, detect, and empower predictive models to adapt to these events and consequent non-stationarity, which further benefits the real-world decision makings.”
Poster Link: https://drive.google.com/file/d/1wErOE5zrr_mKK2ly8zzB2Xj-boWS3aHI/view