About

Dr. Magaly Koch is a geologist specializing in the application of remote sensing and geographic information systems in the study of groundwater resources and environmental change of arid lands. She is particularly interested in projects addressing global water scarcity and flooding problems as related to extreme weather conditions through an interdisciplinary approach. In recent years, she has become interested in studying climate impact on coastal zones and earthquake-related hazards. She holds a Ph.D. in Geology from Boston University, a Diploma in Ground Water Hydrology from the Polytechnic University of Catalonia, and an M.Sc. in Geology from the University of Cologne. Dr. Koch is a Research Associate Professor at Boston University, where she contributes to the Earth & Environment department, focusing her research on environmental and geological challenges using advanced remote sensing techniques.

Research topics

• Geography
• Remote sensing
• Environmental resource management
• Artificial Intelligence
• Computer Science
• Environmental science
• Agroforestry
• Natural resource economics
• Ecology
• Cartography
• Environmental planning
• Geotechnical engineering
• Geology
• Meteorology
• Economics
• Civil engineering
• Engineering
• Business

Selected publications

• Bibliometric Analysis of Ecosystem Services and Stakeholder Engagement in the Mangrove Communities of the Gulf of Mexico and the Caribbean

  Preprints.org · 2025-12-17

  preprintOpen access

  Understanding the services provided by coastal ecosystems is vital for their study, preservation, and restoration. Mangrove forests, in particular, provide key ecosystem ser-vices: they sequester carbon, support fisheries and biodiversity, and enable sustainable tourism. In the Caribbean and Gulf of Mexico, mangrove-related services have been stud-ied extensively, but often in fragmented ways. This meta-analysis combines literature re-view, bibliometric tools, and thematic mapping to identify emerging trends and long-standing gaps. We analyzed 61 peer-reviewed studies across 21 sovereign and U.S. states, highlighting shifting research priorities and the lack of convergence across ecosys-tem service categories. While early research emphasized supporting services such as fish-ery nurseries, recent studies focus on regulating services, especially carbon sequestration. Stakeholder engagement remains limited, with only 18% of studies incorporating local or institutional perspectives. We argue for greater integration of stakeholder input and con-vergence across service categories to enhance the scientific basis for mangrove manage-ment and policy design.

  PublisherOA PDFDOI

• Optimized YOLOv8 with multi-level attention for satellite image-based landslide detection

  Advances in Space Research · 2025-06-09

  article
  PublisherDOI

• Tracing the 2018 Sulawesi Earthquake and Tsunami’s Impact on Palu, Indonesia: A Remote Sensing Analysis

  Journal of Marine Science and Engineering · 2025-01-19 · 1 citations

  articleOpen accessSenior author

  The 2018 Sulawesi Earthquake and Tsunami serves as a backdrop for this work, which employs simple and straightforward remote sensing techniques to determine the extent of the destruction and indirectly evaluate the region’s vulnerability to such catastrophic events. Documenting damage from tsunamis is only meaningful shortly after the disaster has occurred because governmental agencies clean up debris and start the recovery process within a few hours after the destruction has occurred, deeming impact estimates unreliable. Sentinel-2 and Maxar WorldView-3 satellite images were used to calculate well-known environmental indices to delineate the tsunami-affected areas in Palu, Indonesia. The use of NDVI, NDSI, and NDWI indices has allowed for a quantifiable measure of the changes in vegetation, soil moisture, and water bodies, providing a clear demarcation of the tsunami’s impact on land cover. The final tsunami inundation map indicates that the areas most affected by the tsunami are found in the urban center, low-lying regions, and along the coast. This work charts the aftermath of one of Indonesia’s recent tsunamis but may also lay the groundwork for an easy, handy, and low-cost approach to quickly identify tsunami-affected zones. While previous studies have used high-resolution remote sensing methods such as LiDAR or SAR, our study emphasizes accessibility and simplicity, making it more feasible for resource-constrained regions or rapid disaster response. The scientific novelty lies in the integration of widely used environmental indices (dNDVI, dNDWI, and dNDSI) with threshold-based Decision Tree classification to delineate tsunami-affected areas. Unlike many studies that rely on advanced or proprietary tools, we demonstrate that comparable results can be achieved with cost-effective open-source data and straightforward methodologies. Additionally, we address the challenge of differentiating tsunami impacts from other phenomena (et, liquefaction) through index-based thresholds and propose a framework that is adaptable to other vulnerable coastal regions.

  PublisherDOI

• Pixel-based classification method for earthquake-induced landslide mapping using remotely sensed imagery, geospatial data and temporal change information

  Natural Hazards · 2024-02-04 · 15 citations

  article
  PublisherDOI

• Mapping the dynamics of aquatic vegetation in Lake Kyoga and its linkages to satellite lakes

  Science of Remote Sensing · 2024-08-13 · 2 citations

  articleOpen access

  Lake Kyoga is a shallow, young, flooded basin just north of and about 30m lower than Lake Victoria. The catchment encompasses Lake Kyoga itself, and a constellation of several dozen small satellite lakes following valley contours mostly to its east. The Kyoga basin fish fauna shares many non-cichlid species plus a spectacular, partially endemic radiation of haplochromine cichlids most similar to but still largely distinct from those in Lake Victoria. This fish fauna is of high conservation concern, as it preserves remnants of the regional species flock that have disappeared from Lake Victoria and Lake Kyoga, leaving small remnant populations in some of the satellite lakes. Now, these too are imperiled by limnological dynamics, including fluctuations in the nature and extent of aquatic vegetation. The water bodies in the Kyoga Basin are highly dynamic due both to fluctuation in water level and large amplitude variation in marginal and floating vegetation. This variation has profound evolutionary and conservation implications, since it can create and destroy critical aquatic habitat. It can also alternately anneal and cleave gene flow over time, both between the main lake and its satellites, and among the satellite lakes. The aquatic vegetation cluttering these linkages can create a spatial refugium for many native fish species that are more tolerant of hypoxia than an introduced macropredator, the Nile perch. Anthropogenic impacts to this region have greatly increased in recent years, altering relationships between aquatic vegetation and endangered species, fisheries and other ecosystem services provided by the lake. Understanding these dynamics require a means of mapping aquatic vegetation, connectivity, and habitat through time. Here we develop a new and improved algorithm to map the spatial distribution and dynamics of floating and emergent aquatic vegetation via remote sensing. We utilize a time series of 440 Landsat images dating from 1986 to 2020. A series of water and vegetation indices are designed to reveal change in the aquascape over time. First, two types of water masks are derived using a majority rule - a separate water mask for each image and a composite water mask of the region over the study period. Second, the difference between the two masks is then used to delineate the potential location of macrophytes over the image. Third, an algorithm is developed to separate the floating vegetation from emergent vegetation; this algorithm uses Landsat spectral bands and two additional spatial and temporal metrics that considerably improve classification accuracy. A more extensive analysis of aquascape trajectories using remote sensing can inform fish conservation strategies and fisheries management and illuminate the role of landscape dynamics in macroevolutionary patterns of aquatic taxa. • Lake Kyoga Lake system is highly dynamic and complex. • Classification accuracy of aquatic vegetation is usually low. • Open water areas are delineated from water indices and adaptive binarization. • Spatial and temporal metrics derived from Landsat Archive greatly improve accuracy. • Application of the maps guides sustainability between stakeholders and wildlife.

  PublisherDOI

• One versus all: identifiability with a multi-hazard and multiclass building damage imagery dataset and a deep learning neural network

  Natural Hazards · 2024-05-08 · 4 citations

  article
  PublisherDOI

• Recurrent Flooding and Household Food Access in Central Java, Indonesia

  International Journal of Environmental Research and Public Health · 2024-10-17 · 2 citations

  articleOpen access

  It is unknown how recurring flooding impacts household diet in Central Java. We aimed to assess how recurrent flooding influenced household food access over 22 years in Central Java by linking the Global Surface Water dataset (GSW) to the Indonesian Family Life Survey. We examined linear and nonlinear relationships and joint effects with indicators of adaptive capacity. We measured recurrent flooding as the fraction of district raster cells with episodic flooding from 1984-2015 using GSW. Food access outcomes were household food expenditure share (FES) and dietary diversity score (DDS). We fit generalized linear mixed models and random forest regression models. We detected joint effects with flooding and adaptive capacity. Wealth and access to credit were associated with improved FES and DDS. The effect of wealth on FES was stronger in households in more flood-affected districts, while access to credit was associated with reduced odds of DDS in more flood-affected districts. Flooding had more predictive importance for FES than for DDS. Access to credit, a factor that ordinarily improves food access, may not be effective in flood-prone areas. Wealthier households may be better able to adapt in terms of food access. Future research should incorporate land use data to understand how different locales are affected and further understand the complexity of these relationships.

  PublisherOA PDFDOI

• A framework and pilot study for assessing usability of flood data portals for interdisciplinary research

  PLOS Climate · 2024-11-04 · 5 citations

  articleOpen accessSenior authorCorresponding

  There is a lack of datasets to study the climate and human outcomes nexus. There are many flood data portals due to recent improvements in flood identification using satellites, providing opportunities to study the human impacts. The development of these portals is rapid and there is currently no standard for evaluating their usability for interdisciplinary research. This paper addresses this important data gap. We put forth a usability framework that includes data availability, approaches to flood identification, alignment, velocity, variety, and user feasibility aspects. We piloted it through an in-depth review and user survey of NASA Worldview (NW), Global Flood Awareness System (GloFAS), Global Flood Monitoring System (GFMS), Global Surface Water Explorer (GSWE), and Dartmouth Flood Observatory (DFO) GSWE and GloFAS were rated most favorably. Respondents had discrepancies in their opinions on the clarity of the goals and platform accessibility for GFMS, DFO, and NW, and in data and visualization quality for all portals. Historical data and measures of flood recurrence and other characteristics are needed. Flood data products should be provided in multiple formats, aggregated by sub-national boundaries, with mechanisms that delineate incomplete or unreliable data. Flood data portals should include interdisciplinary research as part of their mission. Their longevity and maintenance should be secured to preserve these important data sources for future research. This framework can be adapted and used to enable interdisciplinary spatial and survey data linkages.

  PublisherDOI

• Regional landslide mapping model developed by a deep transfer learning framework using post-event optical imagery

  Georisk Assessment and Management of Risk for Engineered Systems and Geohazards · 2024-01-02 · 16 citations

  article

  Landslides are major natural disasters in mountainous areas, often caused by earthquakes and heavy rainfalls. Traditional manual delineation methods for identifying landslide features using optical imagery are inefficient, highlighting the need for automated detection techniques. Deep Convolutional Neural Networks (CNNs) have emerged as advanced solutions in computer vision for this purpose. Despite the reliance on pre-event and post-event imagery or various data sources like digital elevation models (DEMs), the success of deep learning models largely depends on the quality and availability of training data. This poses a challenge for their immediate application after a landslide. This study explores the transferability of a CNN model trained on data from the 2016 Kumamoto Earthquakes for detecting landslides in different events, specifically the 2018 Hokkaido earthquake and the 2017 Asakura Rainfall in Japan. These cases were chosen for their geographical similarities. The proposed deep transfer learning model, based on a DeepLabV3 + architecture built on a pre-trained ResNet50, automatically identifies landslide features without needing specific training data or model adjustments for each event. It achieved high accuracy in both cases, demonstrating CNNs’ potential for broad application in landslide detection and enhancing disaster response efforts.

  PublisherDOI

• Examining the Effects of Fracking on Groundwater Using the United States Fracking Well Data

  Journal of Student Research · 2024-05-31

  articleOpen accessSenior author

  Ever since fracking technology, drilling downward and then horizontally, has been developed to extract oil and gas from underground bedrock, the number of fracking wells has been drilled at an accelerating speed, and the amount of underground water consumed has also increased enormously. The goal of this paper is to address this alarming environmental and climate threat from the proliferation of fracking wells accompanied by a surge in groundwater consumption, through analyzing the up-to-date registry data available as of October 2023 from FracFocus.org on the fracking wells nationwide . The paper starts with a literature review to study the existing research findings on fracking water consumption issues. Next it provides a comprehensive data analysis of the water consumption by fracking wells at both national and state levels. It also explores the relationships between the water consumption with various other factors such as the vertical depth of the wells, the chemical ingredients of the fracking fluid, and the purposes of the additives, and aims to provide insights from the correlation and causality analyses that may offer potential strategies to reduce groundwater consumption in the fracking industry. In addition, the paper employs machine learning techniques such as Random Forest model to explore using predictive models to identify wells that have high likelihood of causing a water contamination issue so that proactive controls can be developed to reduce the occurrence of water quality issues.

  PublisherOA PDFDOI

Recent grants

• US Egypt Cooperative Research: Assessing the potential of natural resources for land use planning and development west of Aswan, Egypt

  NSF · $103k · 2010–2013

• IRES Track I: Collaborative Research: U.S.-Indonesian Research Experience for Students on Sustainable Adaptation of Coastal Areas to Environmental Change

  NSF · $199k · 2018–2023

Frequent coauthors

• Thomas Schmid

  Centro de Investigaciones Energéticas, Medioambientales y Tecnológicas

  45 shared

• Elena N. Naumova

  34 shared

• Ahmed Gaber

  Minia University

  24 shared

• Laurie G. Baise

  Tufts University

  19 shared

• Farouk El‐Baz

  National Research Centre

  18 shared

• Erin Coughlan de Perez

  Tufts University

  15 shared

• Sucharita Gopal

  Boston University

  15 shared

• Alexander Liss

  Paderborn University

  13 shared

Education

• PhD, Geology

  Boston University

  1993