2026 Conference Student Presentations

2026 ASPRS Gulf South Region Geospatial Conference Student Presentations

_____________________________

Graduate Student Presentations

_____________________________

Quantifying Measurement Uncertainty in Repeat UAS-SfM and UAS-Lidar Surveys

José A. Pilartes-Congo^1,2 and Michael J. Starek^1,2

¹College of Engineering and Computer Science, Texas A&M University-Corpus Christi; ²Conrad Blucher Institute for Surveying and Science, Texas A&M University-Corpus Christi

Abstract: Uncrewed aircraft system (UAS)-based photogrammetry and light detection and ranging (lidar) workflows provide a cost-effective approach for repeat roadway surveys at localized scales. The resulting geospatial products can support applications such as construction monitoring, pavement distress identification, and preventive maintenance planning. However, reliable distinction between true surface change from measurement induced variability remains challenging, as apparent elevation differences may result from uncertainties related to sensor calibration, global navigation satellite system quality, and other sensor-specific artifacts. Motivated by this challenge, this study applies error propagation methods to quantify uncertainty in digital terrain model (DTM) differencing results derived from repeat UAS surveys. The research stems from a Texas Department of Transportation project focused on developing data collection strategies to support the agency’s digital delivery initiative. The study leverages UAS-based structure-from-motion / multi-view stereo photogrammetry (or UAS SfM) and UAS-Lidar point cloud datasets acquired over a concrete runway surface mimicking a typical two-lane Texas State right-of-way (ROW) corridor. Point clouds were georeferenced using post-processed kinematic (PPK) with four ground control points, interpolated into DTMs, and differenced to generate DTMs of difference (DoDs) used to evaluate sensor repeatability. Given the unchanged nature of the test surface, observed elevation differences were attributed primarily to measurement uncertainty rather than true surface change. Accordingly, error propagation was incorporated into the DoD calculations to account for accumulated vertical uncertainty. Results yielded propagated vertical uncertainties of ±3.42 cm for UAS-SfM and ±2.37 cm for UAS-Lidar at the 1σ confidence level, increasing to ±6.85 cm and ±4.74 cm, respectively, at 2σ. These thresholds were then used to enhance DoD interpretation by improving discrimination between noise and meaningful surface change. The findings underscore the value of incorporating error propagation into DoDs derived from repeat UAS surveys and may interest surveyors and transportation engineers involved in UAS deployment for transportation applications and ROW monitoring.

Bio: José A. Pilartes-Congo is a Ph.D. candidate in the Geospatial Computer Science program at Texas A&M University-Corpus Christi (TAMU-CC). He holds a bachelor’s degree in Geographic Information Science and a master’s degree in Geospatial Systems Engineering. For more than a decade, he has contributed to multiple laboratories within the Conrad Blucher Institute (CBI) for Surveying and Science, including roles as undergraduate and graduate student research scientist. He currently works in the Measurement Analytics (MANTIS) Laboratory at CBI and his dissertation research focuses on multimodal remote sensing and computational workflows for digital delivery, with particular emphasis on transportation corridor mapping and monitoring. He also serves as President of the ASPRS Student Chapter at TAMU-CC.

_____________________________

From Visualization to Surveying: 3D Scene Representations with Novel View Synthesis for Coastal Mapping

Esther Oladoyin¹

¹College of Geospatial Computer Science, Texas A&M University-Corpus Christi;

Abstract: Advances in 3D point-cloud reconstruction have become foundational to emerging technologies such as the metaverse, digital twins, autonomous systems, and robotics. Beyond passive remote sensing techniques like uncrewed aerial system (UAS)-based Structure-from-Motion (SfM) and Multi-View Stereo (MVS) photogrammetric methods, active remote sensing technologies such as light detection and ranging (LiDAR) and radio detection and ranging (Radar) are increasingly used for 3D data capture. Recently, novel view-synthesis techniques, including Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS), have emerged as significant advances in 3D vision, enabling real-time rendering, novel view generation, and enhanced photorealism. While NeRF and 3DGS were initially popularized for view synthesis, recent simultaneous localization and mapping (SLAM) systems optimize them for visual realism, camera pose estimation, and robustness to adverse conditions. However, adoption in practical survey workflows remains limited.

To bridge the identified gap, this study evaluates how novel view-synthesis methods, especially 3DGS, compare with and enhance UAS-based SfM/MVS photogrammetric workflows in terms of geometric fidelity, particularly under degraded data quality and reduced image overlap.

The purpose of the research is to examine whether 3DGS can move beyond photorealistic rendering to produce mapping‑useful geometry that meets survey standards. By targeting conditions that challenge photogrammetric pipelines such as reduced image overlap, motion blur, lower image quality, etc., the study determines where novel view-synthesis methods offer advantages or tradeoffs relative to UAS-SfM/MVS photogrammetric techniques. The study benchmarks 3DGS against conventional workflows to reveal whether splatting can deliver survey‑grade reconstructions that are both visually compelling and geometrically accurate. The evaluation also considers how 3DGS may enhance existing SfM/MVS workflows. In doing so, this research seeks to establish whether novel view‑synthesis methods can transition from tools of visual realism to components of geospatial survey pipelines. The studies are performed on RGB-UAS data of Ward Island, Texas A&M University – Corpus Christi campus.

Bio: Esther is a second year PhD student in Geospatial Computer Science at Texas A&M University – Corpus Christi. She holds both a Bachelor’s and Master’s degree in Electrical Engineering. Her research investigates the integration of emerging novel view synthesis techniques such as Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS) into UAS based photogrammetric workflows for survey grade geospatial applications.

Her work evaluates how these methods perform relative to conventional Structure from Motion and multi view stereo photogrammetry under degraded imaging conditions. She also studies the impact of sensor fusion and investigates how dynamic scene elements influence geometric fidelity and reconstruction accuracy of novel view synthesis techniques in real world environments compared with conventional SfM and MVS photogrammetric reconstructions.

Prior to beginning her doctoral studies, Esther worked as a software engineer in the oil and gas industry, where she developed firmware for sensor enabled wired drill pipe systems used for real time subsurface data acquisition. She also worked in the consumer electronics industry, conducting firmware validation and security testing for network connected camera systems.

_____________________________

A Time-Series Approach to Forecasting Land Area Dynamics in the Mississippi River Delta

Basant Awasthi¹, Xuelian Meng¹, Thanos Gentimis², and Manisha K C¹

¹Department of Geography and Anthropology, Louisiana State University; ²Department of Experimental Statistics, Louisiana State University

Abstract: Coastal wetlands are vital ecosystems that provide essential services, including wildlife habitat, flood mitigation, and carbon storage. Louisiana's coast, which holds 40% of the wetlands in the continental U.S., is experiencing substantial land loss at roughly 75 square kilometers per year. To understand and anticipate these changes, this study uses satellite remote sensing to project land area changes in the Mississippi River Delta (MRD). Landsat Collection 2 data from 1984 to 2024 were analyzed in Google Earth Engine, with only cloud-free images selected. Land and vegetated areas were determined using spectral indices (mNDWI, EVI, NDVI, LSWI). The resulting monthly time series had gaps filled with imputation methods. The last 5% of the dataset was set aside for model testing. Several forecasting models were compared: ARIMA, Time Series Linear Model (TSLM), Prophet, a Neural Network Autoregressive (NNAR) model, and an ARIMAX model that included vegetated area as an external variable. Model accuracy was evaluated using Root Mean Square Error (RMSE) and Mean Absolute Error (MAE). The ARIMAX model achieved the highest accuracy (RMSE=0.23, MAE=0.16), underscoring the benefit of including environmental covariates. Prophet performed the worst, likely due to its limited handling of weak seasonality. Although the ARIMAX was most precise, the NNAR model was chosen for future predictions because it does not require external predictors, such as future vegetated area, that are unavailable. Using the NNAR model, land area was projected for 24 months, with 80% and 95% confidence intervals generated via bootstrapped simulations. The MRD is expected to see steady but significant land gain after 2024. These findings support the development of adaptive management strategies tailored to specific site trajectories and offer valuable insights into the geographic variation of wetland changes.

Bio: Basant Awasthi is a Ph.D. student in Geography in the Department of Geography & Anthropology at Louisiana State University. He works as a Graduate Research Assistant in the Technology-Intensive Geospatial and Remote Sensing (TIGeRS) Lab under the supervision of Dr. Xuelian Meng. His research focuses on satellite remote sensing, GIS, and time-series analysis to understand long-term coastal wetland dynamics in Louisiana. His current work integrates machine learning-based land classification with statistical modeling to quantify land-area change and evaluate hydro-environmental drivers using multi-decadal Landsat data.

Basant received his Bachelor of Engineering in Geomatics Engineering from Kathmandu University, Nepal, in 2018. After graduation, he served as a Teaching Assistant in the Department of Geomatics Engineering at Kathmandu University for three years, where he supported courses and research in GIS, remote sensing, and spatial data analysis.

His broader research interests include coastal monitoring, environmental change detection, geospatial data science, and Earth observation for climate and ecosystem studies. Basant expects to complete his Ph.D. in December 2026 and plans to pursue an academic career in geospatial science.

_____________________________

Mapping Phragmites australis Dynamics Through Deep Learning Techniques Based on Time Series Sentinel-2 Imagery and Evaluating its Temporal Transferability

Manisha K C¹, Xuelian Meng¹, and Basant Awasthi¹

¹Department of Geography and Anthropology, Louisiana State University

Abstract: Phragmites australis is the most dominant emergent plant in the freshwater and brackish marshes in the lower MRD, significantly contributing to the stability of the marsh ecosystem. However, the widespread dieback of Phragmites has raised serious concerns about the stability and resilience of wetlands. Traditional mapping methods, which rely on single-date imagery and concurrent ground samples, are costly and time-consuming and often do not generalize across years. This study explored an integrated approach to assessing the robustness and temporal generalization of deep learning models for Phragmites classification when trained on multi-year historical datasets, without concurrent ground-truth data. We used a U-Net deep learning model to classify wetlands as Phragmites, non-Phragmites, and water using time-series Sentinel-2 Level-2 imagery from 2019 to 2024. The results demonstrated that while single-year models performed well with concurrent training samples (with an overall accuracy of 90%-98%), the cross-year transferability model's accuracy depended on the choice of training sample time and vegetation condition similarity, achieving an overall accuracy above 80% for most input scenarios. However, integrating training datasets from multiple past years yielded accuracies comparable to those of models trained on concurrent samples. The overall accuracy was above 90%, and Phragmites accuracy was above 80% in most years, indicating the potential to use historical datasets without concurrent data. This deep learning classification approach facilitates model training using historical data, thereby reducing reliance on costly concurrent field surveys and enhancing the long-term efficiency of wetland monitoring. By linking classification accuracy to disturbance history, this research collectively established a transferable framework for long-term monitoring in other coastal areas that experience significant phenological variation and rapid change.

Bio: Manisha K C is a recent Geography Graduate from the Department of Geography & Anthropology at Louisiana State University, where she served as a Graduate Research Assistant in the Technology Intensive Geospatial and Remote Sensing (TIGeRS) Lab under the supervision of Dr. Xuelian Meng. Her research focuses on applying remote sensing, GIS, and deep learning techniques to monitor coastal vegetation dynamics and environmental change. She investigated the health dynamics of Phragmites australis in the lower Mississippi River Delta coastal wetlands using Sentinel-2 time-series imagery and deep learning methods to improve vegetation monitoring and change detection.

She received her Bachelor of Engineering in Geomatics Engineering from Kathmandu University, Nepal, in 2022. Her broader interests include remote sensing, GIS, UAV applications, coastal ecosystem monitoring, agriculture, and analysis of climate-related environmental change.

Manisha aims to pursue a career applying geospatial technologies, remote sensing, and data-driven analytics to support environmental monitoring, climate resilience, and sustainable resource management.

_____________________________

Seasonal Beach Morphology and Shoreline Change Detection Using Mobile LiDAR at Padre Island National Seashore

Vamshi Gindam^1,2, Michael J. Starek^1,2, and Jacob Berryhill²

¹College of Engineering and Computer Science, Texas A&M University-Corpus Christi; ²Conrad Blucher Institute for Surveying and Science, Texas A&M University-Corpus Christi

Abstract: Padre Island National Seashore (PAIS), Texas, encompasses over 70 miles of barrier island coastline along the Gulf of Mexico and provides critical nesting habitat for the endangered Kemp's ridley sea turtle (Lepidochelys kempii). Effective nest management requires high-resolution, repeatable measurements of beach elevation and shoreline position during nesting and hatching seasons; however, traditional surveying methods are time-intensive with limited spatial coverage, and airborne lidar often lacks the temporal frequency needed to capture short-term coastal change. This project implements vehicle-mounted Mobile LiDAR Scanning (MLS) to collect high-density elevation data along the PAIS beach corridor. Repeat MLS surveys were conducted during the 2024 and 2025 nesting seasons across three study sites, including Malaquite Beach, Mile Marker 15–20, and Mile Marker 30–35. The MLS system integrates a high-resolution lidar sensor with a tightly coupled GNSS/IMU navigation solution and is supported by ground control targets and

RTK-GNSS beach profile surveys for quality control and vertical accuracy assessment. Survey data were processed to generate georeferenced point clouds and bare-earth digital elevation models (DEMs) referenced to NAD83(2011) and NAVD88. These DEM products support shoreline proxy extraction and elevation change analysis to identify erosion and accretion patterns occurring between the beginning and end of nesting seasons. Vertical accuracy and repeatability are evaluated using independent RTK-GNSS measurements to establish confidence limits for change detection. This presentation will show results from MLS surveys conducted during the 2024 and 2025 nesting seasons, including DEM products, vertical accuracy assessment, and observed changes in beach morphology and shoreline position across the three PAIS study sites.

Bio: Vamshi Gindam is a Graduate Research Assistant in the Measurement Analytics Lab (MANTIS) at the Conrad Blucher Institute for Surveying and Science at Texas A&M University–Corpus Christi (TAMU-CC), where he is pursuing a Master of Science in Geospatial Systems Engineering in the College of Engineering and Computer Science under the advisement of Dr. Michael Starek.

Vamshi received his Bachelor of Science in Civil Engineering from Jawaharlal Nehru Technological University Hyderabad (JNTUH), India. His research focuses on coastal geomorphology, shoreline change detection, and beach morphology monitoring using mobile LiDAR systems and geospatial analysis at Padre Island National Seashore (PAIS), as part of a National Park Service–funded project investigating the impacts of sea level rise on Kemp’s ridley sea turtle (Lepidochelys kempii) nesting habitat.

_____________________________

A Geospatial and Machine Learning Framework for Modeling Freshwater Inflow Variability in the St. Marks–Apalachee Bay Watershed

Christianah Adegboyega¹

¹School of the Environment, Florida Agricultural and Mechanical University

Abstract: Characterizing freshwater inflow variability is fundamental to understanding hydrologic connectivity and salinity dynamics in coastal estuarine systems. Quantifying these inflows remains challenging due to the combined and scale-dependent influences of surface runoff, streamflow, and groundwater discharge. In the St. Marks – Apalachee Bay watershed, limited spatial characterization of freshwater source pathways constrains hydrologic assessment and estuarine salinity modeling. This study develops an integrated framework combining Geographic Information System (GIS) analysis, Long Short-Term Memory (LSTM) machine learning, and Unmanned Aerial System (UAS)–based thermal infrared (TIR) remote sensing to characterize and model hydrologic variability of freshwater inflows to Apalachee Bay. 

Hydrologic features were derived from hydrographic and topographic datasets, including the Watershed Boundary Dataset (WBD), National Hydrography Dataset (NHD), and Digital Elevation Models (DEM). Flow direction, flow accumulation, and stream network layers were generated within a GIS environment to delineate drainage pathways, identify upstream contributing areas, and characterize surface-water connectivity within the watershed. Land use/land cover data derived from satellite imagery were analyzed for change detection between 1990 and 2024, revealing a gradual shift from natural land cover to developed land uses. 
Streamflow observations from two gaging stations, together with meteorological forcing variables, were used to train an LSTM neural network for streamflow prediction. The model captures nonlinear relationships and temporal dependencies inherent in the coastal watershed hydrology. Preliminary results indicate stable convergence and effective learning behavior, as demonstrated by decreasing training and validation loss, supporting the applicability of deep learning methods for streamflow prediction in estuarine watersheds. 

Groundwater contributions will be addressed through spatially explicit remote sensing and GIS analysis. Planned UAS-based TIR surveys will be used to identify thermal anomalies associated with shoreline springs and groundwater seepage zones along Apalachee Bay. These features will be georeferenced and integrated into the GIS framework to improve spatial representation of groundwater inputs and complement surface-water-based streamflow predictions. This study demonstrates the value of integrating GIS, remote sensing, machine learning, and UAS technologies to advance hydrologic modeling and spatial analysis in coastal estuarine systems.

Bio: Christianah Adegboyega is a Ph.D. candidate in the School of the Environment at Florida Agricultural and Mechanical University. Her research integrates geospatial analysis, machine learning, and remote sensing to investigate freshwater inflow variability in coastal and estuarine systems. To support this work, she has developed computational and data science skills for large-scale environmental data processing, time-series modeling, and spatial analysis. Her expertise includes watershed hydrology, GIS-based spatial analysis, land use/land cover change detection, and the use of long short-term memory (LSTM) neural networks for streamflow prediction. She also uses GLDAS datasets for groundwater feasibility assessments and contributes to living shoreline suitability modeling to support coastal resilience planning. In addition to her research, she mentors students in geospatial methods and contributes to the development of applied tools for watershed and coastal management. Her work promotes the integration of GIS, remote sensing, and machine learning to inform quantitative environmental and water resource management.

_____________________________

Species-Specific Phenology of Phragmites australis in Coastal Louisiana Using Machine Learning and Multi-Sensor Satellite Data

Bijaylaxmi Sahoo¹

¹Department of Geography and Anthropology, Louisiana State University

Abstract: Louisiana’s coastal wetlands are disappearing at an alarming rate of about one football field of land per hour due to subsidence, sea-level rise, and reduced sediment input. Phragmites australis, a dominant wetland species, plays a vital role in stabilizing land by trapping sediments through its complex rhizome network. However, widespread dieback of Phragmites australis has recently been observed, and its phenological dynamics in Louisiana remain poorly understood. This study presents a detailed phenological assessment of Phragmites australis in the lower Mississippi River Delta using satellite remote sensing from 2016 to 2025. To ensure species-specific monitoring, a machine learning model was developed to classify and isolate Phragmites australis stands from other wetland vegetation, restricting phenological analysis exclusively to Phragmites australis-dominated areas rather than the broader vegetation assemblage. The study employed a multi-sensor framework, integrating Landsat 8 and 9 imageries with Sentinel-2 imageries downscaled to match Landsat’s 30 m spatial resolution, creating a consistent time series of vegetation indices. A threshold-based phenology estimation approach with cubic interpolation was applied to derive the start of season (SoS), end of season (EoS), and seasonal growth trajectories for the extracted Phragmites australis stands. Results indicate that Phragmites australis exhibits peak growth between May and November. Additionally, decadal fluctuations in vegetation indices were examined relative to the mean phenological curve averaged over the 10-year period to identify anomalies and potential drivers of declining plant vigor. This study provides the first phenological characterization explicitly derived for Phragmites australis in coastal Louisiana using species-specific remote sensing classification. The findings offer new insights into Phragmites australis growth dynamics and potential dieback mechanisms, highlighting the value of integrating machine learning with multi-year satellite observations for long-term wetland monitoring. These results provide critical information to support restoration and conservation strategies aimed at stabilizing Louisiana’s rapidly eroding coastal wetlands.

_____________________________

The use of LiDAR and Hyperspectral data to identify vegetative cover and geomorphic profile to monitor Flotant in the Barataria-Terrebonne Estuary System

Noah Wurtzel¹, Ivy Norton¹, Alex Himel¹, Samuel Landry¹, Reece Toups¹, Chris Bonvillain¹, Jonathan Willis¹, Justine Whitaker¹, Gary LaFleur¹, Balaji Ramachandran¹

¹Department of Biological Sciences, Nicholls State University 

Abstract: Flotant marsh is a unique wetland ecosystem that occurs as a thin vegetative mat (<30cm to 60cm) consisting of plants loosely rooted above peat that float on top of the water column. Flotant makes up a large portion of the freshwater and intermediate marsh habitat found in Terrebonne and Barataria basins. Eutrophication, sea level rise, and natural disasters can degrade flotant. The floating characteristic of this unique mat makes repair and re-creation of flotant nearly impossible. To preserve and protect flotant, characteristics of thick healthy flotant are being compared to thin degraded flotant in our project. We are surveying sites at Lake De Cade, Lake Boeuf, and Lac des Allemands using un-crewed aerial vehicles (UAV) carrying LiDAR and hyperspectral sensors four times per year. In order to protect the flotant from damage, we developed a floating launch pad that allows UAV to be deployed on the water. The LiDAR sensor has captured site topography, allowing us to highlight mat breakages and thickness of cover. Hyperspectral sensors have allowed us to resolve specific vegetation composition by the unique wavelengths of various floral species. Through our initial missions we have identified the presence of several native plants such as Sagitarria lancifolia, Sagitarria latifolia, Hibiscus moscheutos, Cephalanthus occidentalis, Typha latifolia, Althaea officinalis, Zizaniopsis miliacea, and Phragmites australis, along with invasive plants such as Pontederia crassipes, and Ipomoea lacunose. Winter months will allow LiDAR to better penetrate the mat and detect breakages, while also allowing hyperspectral sensors to detect plant species that are hidden by taller canopy in the summer. We expect that by combining results from LiDAR and hyperspectral sensors, we may be able to better characterize sections of flotant that may be in danger of separation during large storm events.

_____________________________

A Multi-Index Sentinel-2A and Machine-Learning Approach for High-Resolution Flood Mapping: Insights from Kerr County’s 2025 Floods

Mohammad Sohail^1,2 and Tianxing Chu^1,2

¹Department of Computer Science, Texas A&M University-Corpus Christi; ²Conrad Blucher Institute for Surveying and Science, Texas A&M University-Corpus Christi

Abstract: Flood events continue to pose an escalating risk to communities, infrastructure, and agricultural systems worldwide. On July 4, 2025, severe flash floods struck Central Texas—particularly Kerr County—after rapid rises in the Guadalupe River, where water levels increased by nearly 6 m within three hours, reaching major flood stage. The disaster resulted in more than 130 fatalities, underscoring the urgent need for rapid and reliable flood detection systems.

In response to this event, this study presents an automated flood-mapping framework that integrates Sentinel-2A satellite imagery with multiple spectral water indices—Normalized Difference Water Index (NDWI), Modified Normalized Difference Water Index (MNDWI), Automated Water Extraction Index (AWEI), Enhanced Water Index (EWI), Modified Bare-Soil Water Index (MBWI), Automated Normalized Difference Water Index (ANDWI), and Normalized Channel Index for Water Identification (NCIWI) alongside machine learning classifiers such as Random Forest (RF), Gradient Boosting Machine (GBM), Support Vector Machine (SVM), and k-Nearest Neighbor (kNN). Implemented by Google Earth Engine (GEE), the workflow enables rapid delineation of inundated areas, with particular focus on identifying impacts within residential and urban zones. Feature-importance analysis indicates that MBWI and AWEI contribute most significantly to accurate flood classification.

Model evaluation indicates that RF and GBM outperform the other classifiers. RF achieved an F1-score of 0.9828 with an overall accuracy of 98.6%, while GBM achieved an F1-score of 0.9779 with an overall accuracy of 97.0%. Estimated flood extents from RF and GBM were approximately 5.31 km² and 4.88 km², respectively. Overall, the results demonstrate that combining multispectral Sentinel-2A data with advanced machine-learning classifiers offers an effective and timely approach for flood detection, supporting improved emergency response, recovery planning, and long-term flood-risk mitigation.

_____________________________

Mapping and Evaluating Fine-Scale Variations in Crop Growth Parameters of Corn Using UAV-Based Remote Sensing

Bikash Ghimire¹

¹Department of Natural Resources and Environmental Sciences, Alabama A&M University

Abstract: Understanding fine-scale variability in crop growth and canopy development plays a critical role in agricultural management. Most conventional methods can be labor-intensive, time-consuming, and inefficient at large scale studies. Remote sensing has emerged as an effective alternative for monitoring crop growth at high spatial and temporal scale studies. This study evaluated the applicability of UAV-based high-resolution multispectral remote sensing data to detect fine-scale variations in crop growth.

This research was conducted at the Winfred Thomas Agricultural Research Station (WTARS) of Alabama Agricultural and Mechanical University in Hazel Green, Alabama. UAV-based multispectral imagery, and field measurements were collected at weekly intervals throughout the growing season of corn (Zea mays). Field observed vegetation cover and plant height were used to develop and validate predictive models using UAV-based MS RS data to predict plant height and cover across different stages of the crop growth. Vegetation indices (Vis) including normalized difference vegetation index (NDVI), normalized difference red edge index (NDRE), and normalized difference water index (NDWI), soil-adjusted vegetation index (SAVI) were analyzed to characterize spatial and temporal variations in plant cover and height. Both simple and multiple linear regression models were evaluated using VIs as predictor variables. Results demonstrated that NDWI best explained fine-scale variations in plant cover (R² = 0.8), while NDRE explained variability in plant height (R² = 0.6), when used in simple linear regression models. Combination of NDRE and NDWI worked as a superior model when evaluated in multiple linear regression models to predict plant cover (R²=0.9) as well as plant height (R²=0.8).  Our findings highlight the applicability of UAV-based multispectral remote sensing data for fine-scale mapping and monitoring of crop parameters that could be beneficial in implementing precise crop management practices as in precision agriculture.

_____________________________

Quantitative assessment of surface urban heat island intensity over major cities of Alabama, USA using remotely sensed land surface temperature and land-use/ land-cover data

Ruwini Rathnayaka¹ 

¹Department of Natural Resources and Environmental Sciences, Alabama A&M University

Abstract: Surface temperatures over the developed areas are generally higher than their surrounding rural areas leading to Surface Urban Heat Islands (SUHI). Intensity and magnitude of the SUHIs can be quantified using surface temperature differences between urban areas and their surrounding non-urban/ rural areas and is a widely used indicator of urban thermal stress. Increased SUHI Intensities (SUHIIs) can negatively impact human health, urban ecosystem functioning, and energy consumption, particularly during warm seasons. The overall goal of this study was to quantify spatial and temporal variations of SUHIIs using remotely sensed land surface temperature (LST) and land-use/land-cover (LULC) data. As the study sites, five (5) major cities of Alabama, USA were selected. Developed and non-developed LULCs were delineated using National Land Cover Database (NLCD) LULC products, with developed classes representing urban areas and vegetated and natural classes representing non-urban areas. LST data were obtained from the MODIS MOD21A2 product. SUHII over each city was quantified as the mean surface temperature difference between urban and surrounding non-urban pixels. Seasonal average SUHIIs during day- and night-times of the summer and winter seasons were used to assess temporal variability in SUHII over the period from 2000 to 2025. Results indicate significantly stronger SUHIIs during summer relative to winter, highlighting enhanced thermal contrasts between natural and constructed areas during warmer periods. Temporal analyses further suggest increasing thermal differences (as reflected in SUHII) over time in regions experiencing sustained urban development over the study period. These findings demonstrate the applicability of long-term LST and LULC datasets for developing a reliable framework for detecting seasonal and temporal SUHII patterns at a regional scale. Our findings also highlight the importance of seasonal considerations in urban climate assessments and support the use of remote-sensing-based approaches to inform climate-resilient urban planning and heat-mitigation strategies.

_____________________________

Undergraduate Student Presentations

_____________________________

Leveraging GeoAI for Damage Assessment in South Louisiana

Samuel Landry¹ and Balaji Ramachandran¹

¹Geomatics Program, Department of Applied Sciences, Nicholls State University

Abstract: Geographic Information System (GIS) Artificial Intelligence (GeoAI) is a powerful tool that enhances the interpretation, analytics, and scalability of spatial data processing by analyzing large and complex geospatial datasets. By integrating artificial intelligence with traditional GIS methods, GeoAI enables faster decision-making and improved accuracy when working with spatial data collected from remote sensing and other geospatial sources. Within GeoAI, Machine Learning (ML) and Deep Learning (DL) are concepts that allow models to identify patterns, interpret data, and make decisions within spatial datasets that may not be easily recognized through manual interpretation. However, these models require substantial amounts of data and time to train. To reduce the amount of data needed to train a model, pre-trained deep learning models can be used. Pre-trained AI models are ML or DL models that are already trained on large datasets that can be implemented in a GIS environment as needed. These models can be fine-tuned and localized or directly applied to new study areas, allowing analysts to achieve reliable results with less training data and reduced processing time. The objective of this project is to showcase the significance of GeoAI in tasks entailing classification and detection. Using a deep learning architecture SegUNet in ENVI 6.0, post-Ida damage assessment was conducted. The AI model was tasked with locating and extracting blue tarps around Lafourche Parish. The validation accuracy was 99%. After performing “ground truthing” the model successfully detected most if not all blue tarps within the extents, however a considerable number of false positives were also extracted. We are currently working on detecting other types of structural damages. In addition, pretrained models such as the building footprint extraction deep learning model were used to compare residential and commercial building footprints pre and post Hurricane Ida along with the outputs from the trained model. 

Bio: Samuel Landry, born and raised in South Louisiana, is a resident of Thibodaux, LA. He graduated from Thibodaux High School in Spring of 2024, and is pursuing a bachelor’s degree in Geomatics at Nicholls State University. Samuel is currently working as an undergraduate research assistant at the Geospatial Technology Lab under the guidance of Dr. Balaji Ramachandran working on research projects related to coastal sciences. His research interests include adoption of GeoAI into coastal monitoring projects, as well as hyperspectral and lidar remote sensing. Samuel is expected to graduate in Spring of 2028.

_____________________________

Examining Algal Blooms in the Northern Gulf of Mexico using Remotely sensed chlorophyll-a data

McKinley Dunford¹ 

¹Department of Natural Resources and Environmental Sciences, Alabama A&M University

Abstract: Algal blooms are naturally occurring events of large growth of phytoplankton, often resulting from excessive nutrient intake, temperature increase, or other anthropogenic factors. Algal blooms are considered harmful when they negatively affect human health or ecosystems by producing toxins or creating anoxic environments. Observing spatial distribution and trends of occurrence and frequency in the Gulf of Mexico in conjunction with the presence of the brevetoxin Karenia brevis can help predict future events as well as their extent. In this study we examined the applicability of remotely sensed chlorophyll-a data from Moderate Resolution Imaging Spectroradiometer (MODIS) at 1km spatial resolution to map and evaluate harmful algal bloom incidences along the coastal waters of Alabama, USA. Findings from our preliminary evaluations reveal MODIS chlorophyll-a data as a promising data source for mapping spatial distributions of HAB events. However, we were not able to detect significant correlations between chlorophyll-a concentrations and the occurrence of Karenia brevis algal blooms and thus indicate the needs for further analyses to evaluate any potential linkages between the occurrence of Karenia brevis and chlorophyll-a levels before, after, and during algal blooms using time specific data.

_____________________________