Discover the latest academic research and innovations during this interactive poster session. Engage with students directly to discuss new ideas and emerging technologies shaping the field.
10:00 AM – 10:05 AM – Tracking Invasive Hydrilla in Florida Freshwater Ecosystems using a Sentinel-2–Based spectral Index (SVIH) in Google Earth Engine
This study presents the Submerged Aquatic Vegetation Index for Hydrilla (SVIH), a new spectral index developed using Sentinel-2 imagery to detect invasive hydrilla in Florida freshwater ecosystems. Implemented in Google Earth Engine, SVIH enables efficient, large-scale, and multi-year monitoring. Field validation confirms its accuracy, making it a valuable tool for aquatic vegetation management and conservation.
Ayesha Malligai M., University of Florida
10:05 AM – 10:10 AM – High-Resolution Land Cover Mapping with Geographic Object-Based Image Analysis (GEOBIA) for Neighborhood Level Understanding of Urban Environments.
Land cover maps catalog and represent complex earth observation products by categorizing surface features into multipurpose reference maps. Neighborhood level high resolution (HR) land cover maps give valuable information to cities, states, and governmental agencies. (Szantoi et al., 2020) Analysis of HR urban land cover maps quantify fields such as: bio-geophyical, social-economic, hazards mitigation, and change through time (Griffith & Hay, 2018; O’Neil-Dunne et al., 2014). Geographic Object Based Image Analysis (GEOBIA) is a methodology to extract information from remotely sensed imagery. The process of GEOBIA applies segmentation algorithms to imagery by grouping similar pixels into objects. Homogenous objects store information about each pixel group for example: spectral values (RGB, reflectance values), texture (contrast), spatial information (area, height) and contextual properties (length of shared border) (Hossain & Chen, 2019; Kucharczyk et al., 2020). GEOBIA was performed on high resolution 0.5 feet, 8-bit, RGB-NIR, NAD83(2011) Illinois East aerial imagery from 2021 for the Justice and Willow Springs neighborhoods of Cook County, Chicago, Illinois. This data was combined with lidar derivatives to develop high spatial resolution landcover maps. The segmentation was implemented with eCognition Developer 10.4 software. Preprocessing of imagery, lidar derivatives (DEM and nDSM), map schemes, and statistics analyses were prepared in ArcGIS Pro 3.3. Statistical analysis of accuracy was completed with a comparison of land cover maps to the imagery. This project is an example of best practices for development of high resolution neighborhood scale land cover maps through GEOBIA. The HR urban maps produced are valuable as basemaps for several types of GIS analysis. As an example, an impervious surface per parcel has been developed for the study area.
Christine Moen-Crabtree, Penn State
10:10 AM – 10:15 AM – Geomorphic Change and Inundation Associated with Hurricane Helene on the Nolichucky River, TN through DEM Differencing
Hurricane Helene made landfall in Florida on September 24th, 2024, and pushed large amounts of precipitation north into East Tennessee and Western North Carolina during the following week. The high volume of rain falling in these mountain regions produced extreme flooding events both locally and farther downstream. One such river impacted by the storm runoff was the Nolichucky, located in northeastern Tennessee. The storm produced a record-breaking flood that devastated communities along the river, especially Erwin, TN, where the flood destroyed major transportation infrastructure and inundated much of the city. The deluge also eroded the channel banks and modified the planform of the river. This study investigates the impacts of Hurricane Helene on the riparian corridor and floodplains of the Nolichucky River in East Tennessee through a DEM of Difference (DoD) analysis to investigate the effects of this large magnitude flood. The DEMs being compared were derived from lidar data collected before the flood in 2016 and just after the flood in late September 2024. Positional uncertainties will be propagated according to the lidar system specifications and mission parameters, to distinguish observed landscape changes from noise in the DoD. This study will examine changes to both the natural and built environments.
Briar Pierce, University of Florida
10:15 AM – 10:20 AM – Quantifying Local Vertical Land Motion to Improve Sea-Level Rise Projections for Coastal Massachusetts
Global sea-level rise (SLR) represents one of the most pressing consequences of climate change, threatening low-lying coastal communities and ecosystems worldwide. The interaction between eustatic SLR and vertical land motion (VLM) determines the rate and extent of relative sea-level rise (RSLR), yet the magnitude of VLM can vary over short spatial scales due to diverse natural and anthropogenic processes. Current projection frameworks, including the Intergovernmental Panel on Climate Change (IPCC) Sixth Assessment Report, often underestimate local RSLR impacts by relying on coarse regional linear VLM estimates that fail to capture fine-scale spatial heterogeneity and temporal variability.
This study addresses this critical knowledge gap by generating improved high-resolution RSLR projections for coastal Massachusetts through the integration of spatially and temporally variable VLM estimates derived from advanced remote sensing techniques. We utilize nine years of Sentinel-1 Synthetic Aperture Radar (SAR) interferometry data (2016-2025) to quantify local VLM rates across coastal Massachusetts and assess their contribution to present and future RSLR scenarios. Our analysis reveals significant spatial variability in VLM, with some coastal areas experiencing subsidence rates that cause regional RSLR estimates to substantially understate actual SLR impacts. This work provides critical scientific insights for planners, engineers, and policymakers working to protect vulnerable coastal communities from accelerating SLR hazards.
Anurag Sharma, Florida International University
10:20 AM – 10:25 AM – Detection of Woody Plant Community Change on Wetland Tree Islands from Remotely Sensed Data – A Historic Change Analysis
Vegetation communities in the Everglades are arranged along a narrow topographic gradient creating a patch mosaic of graminoid marshes and prairies, hardwood hammocks and pine forests where small differences in elevation determine inundation frequency and duration driving the plant community distribution. Hydrologic restoration of the Everglades increases water levels over longer periods throughout the year which is expected to affect species-specific survival of trees, resulting in compositional and structural changes in woody vegetation communities. To understand the spatiotemporal changes of woody vegetation, we detected vegetation communities in selected tree islands in Everglades National Park and quantified horizontal change in extent and change in community height of tree island communities over a specified time period within islands. Major vegetation community outlines were hand digitized from overlapping 1973 Near Infrared Aerial Stereo Photography (ASP). Random forest classifiers were trained to map the tree island plant communities from bi-seasonal Worldview 2012 imagery. The variable set composed of spectral bands, vegetation indices and the canopy structure and vegetation height metrics derived from 2017 LiDAR data, were evaluated. The overlay of the two maps generated a horizontal change of community classes. A minimum mapping unit of 100 m2 was applied as a morphological filter considering the four neighbor pixels. The spatially explicit changes of major communities across islands were mapped and tabulated for each island. To analyze change in community height, Digital Surface Models (DSM) were generated from 2012 ASP and 2017 LiDAR point clouds for one tree island. Accuracy of each ASP derived DSM was estimated using the Root Mean Square Error with LiDAR derived DSMs as reference. Long-term change detection for coarse morphological classes was possible when using ASP. Historic change detection of more detailed plant communities is difficult. Interpolation of point cloud data derived from aerial photography was successful and can be used in change analysis of tree islands.
Ximena Mesa, Florida International University
10:25 AM – 10:30 AM – Predicting Chlorophyll-a Concentration Using Deep Learning and Remote Sensing Data
Monitoring chlorophyll-a (chl-a) concentration is critical for understanding the health of freshwater ecosystems and mitigating the impact of harmful algal blooms. Traditional methods of water quality monitoring are often limited by spatial and temporal constraints. To address this, our study explores the integration of long-term environmental data and deep learning to accurately predict Chl-a dynamics at high temporal resolution.
The primary aim of this study is to develop a deep learning model capable of predicting chl-a concentration based on environmental variables data. Our model utilizes the Gated Recurrent Unit with Decay (GRU-D) architecture, which is well-suited for capturing temporal dependencies in time-series data and designed to handle missing values. We investigate the influence of 13 environmental variables, including water temperature, air temperature, turbidity, pH, dissolved oxygen, CDOM, nitrate concentration, and meteorological factors, on chl-a fluctuations over time. This study used ArcGIS Pro software to extract important information for model training, and Python for model development.
The study focuses on Muskegon Lake, Michigan, a freshwater estuary of Lake Michigan with a rich history of ecological restoration and continuous monitoring. We use over 40,000 temporally aligned observations collected from 2011 to 2022 at a single observation buoy. These measurements were obtained through high-frequency monitoring initiatives and compiled into a daily median time series for model development.
The preliminary results demonstrate that the model successfully captures the seasonal and interannual variability in Chl-a concentrations. The model exhibits promising generalization ability and robustness, making it particularly valuable for real-world water quality management scenarios.
This study contributes a scalable and interpretable framework for chl-a prediction that can complement satellite-based monitoring in data-scarce periods. Also, it highlights the potential of combining deep learning with environmental observations to generate continuous water quality estimates that can be integrated into geospatial decision support systems for lake management and restoration planning.
Acknowledgement: This material is based on work supported by the EEG and NSF grant # 2401942.
Eden Wasehun, North Carolina A&T State University
