Spring 2022 - Project Abstracts
Selu Adams - Vegetation and Land Use Change from 1987 - 2019 in Lumpkin County, GA
In recent decades, rapid development in areas of high biodiversity, such as the southern Appalachians, has raised concerns about land use change and how it affects the ecology of those regions. One such location that has experienced rapid growth over the last 30 years is Lumpkin County, a county in northeast Georgia. Using NDVI and supervised classification methods, I explore the extent to which land use and vegetation density and distribution has changed between 1987 and 2019. The results of the NDVI comparison suggest that the areas that have experienced the largest losses in vegetation density are in the central and southern portions of the county due to developments in infrastructure and industry. Both the maximum likelihood and the Forest Service’s LCMS (Landscape Change Monitoring System) and classifications of land uses within the county suggest that the total amount of forested area within the county is around the same for both years, and that the largest changes in area have involved Further supervised classifications using higher spatial resolution imagery and data (considerations) should be implemented for higher accuracy of land uses within the county, which could result in more well informed land management decisions by both public and private landowners.
Marco Alvarez - Compare Northern Florida Plantation Pine Stands in the Panhandle Region to Pine Stands in Preserves in Maine
Longleaf pine forests in the US have decreased by 97% since in the United States. Removal for buildings, fuel, and paper led to this large decrease in longleaf pine forests since the 1800s. These longleaf pine forests in the southeastern United States provide habitat for keystone species, such as gopher tortoises, which create burrows for other animals to stay cool when temperatures are high. These forests also have high biodiversity and are home to 900 endemic plants, as well as 29 threatened and endangered species (The Conservation Fund, 2021) . Longleaf pine forests are fire dependant habitats, and the soil requires scarification in order for the proper nutrients to give rise to the wide variety of understory plants. Healthy longleaf forests appeared more like savannahs than thick, densely foliaged vegetation (Nature Conservancy, 2020). Areas that have even aged stands were likely formed by a large disturbance, while areas with uneven aged standards were likely formed by a variety of small disturbances. This can impact NDVI values, since savannas have more exposed bare soil than dense vegetation. NDVI stands for normalized difference vegetation index, and it represents how healthy a forest is based on how green the leaves are. The NDVI is calculated by subtracting the near infrared reflectance of leaves by the red color of the visible spectrum and diving this value by the sum of the two. NDVI is always between -1 and 1, with soil having an NDVI of about 0, water in the negatives, conifer forests around 0.5, and healthy deciduous forests around 0.8. The greener a leaf is, the higher its reflectance in the near infrared wavelength, and the lower its absorbance of what we perceive as red in the visible light spectrum.
Anna de Hostos - Harmful Algal Blooms in Alameda County during the 2011-2017 California Drought
LHarmful Algal Blooms (HABs) are incidences of excessive algae growth in fresh and salt water environments that can pose a threat to local vegetation and wildlife as well as humans and their pets. The health hazard of these algal blooms occurs due to the toxins released by the algae. The most common freshwater algal blooms, of importance in this study, are the result cyanobacterial blue-green algae that can produce cyanotoxins (Denchak & Sturm, 2019). Factors of climate change, such as warmer water temperatures and lower water volume can have a large impact on the scale of HABs, making HAB monitoring an increasingly important part of environmental management. Worsening droughts can therefore contribute to HAB levels, especially in states that have been highly impacted by severe drought (Poole, 2021). California experienced one of its worst droughts in history between 2011-2017. In this study, satellite images of Alameda County, CA lakes and reservoirs were analyzed in order to quantify HABs during the 2011-2017 drought. The Chlorophyll Reflection Peak Intensity Algorithm (CRPI), used in previous studies to calculate chlorophyll present in a body of water to quantify algae levels, was used to track HAB trends (Ma et al., 2021). Landsat 8 RGB-432 (2013-2017) and Landsat 5 RGB-321 (2011) images obtained from USGS Earth Explorer were processed by obtaining RGB band stacks, creating ROI inverse masks of lakes and reservoirs of interest and applying the CRPI algorithm to water pixels. The goal of the study was to measure the correlation between droughts and HAB incidence, as well as to employ remote sensing as a technique for HAB monitoring. Results show a significant correlation between rising temperature and falling water quantities during drought years, and an increase in CRPI values in some lakes and reservoirs. The trends CRPI trends between 2011-2017 indicate a notable impact of drought on HABs.
Matthew Duyst - Studying the Effect of Converting Paddied Rice to Urban Land on Atmospheric Methane Concentration Along the Yangtze River Delta (YRD) in China
Atmospheric CH4 emissions along the Yangtze River Delta (YRD) will be carried out in two phases. First, I will identify acreage of agricultural loss due to urban expansion through spatiotemporal mappings. These visualizations will explore China’s contribution of atmospheric CH4 over the past three decades (1990 to 2018) through the acquisition of Landsat imagery and Global Artificial Impervious Area (GAIA) data. Land classification mappings (identifying cropland along the YRD) will be achieved through MODIS imagery and Global Land Surface Satellite Climate Data (GLASS-GLC). A next step for this study will be to create a paddied riceclassification (currently, we are examining all cropland within the study region). Future ambitions will also project the rise in NGV presence and quantify atmospheric CH4 emissions of faulty tailpipe leaks through various scenario analyses. Forecast modeling NGV emission measurements at the Fleet-Level will identify trends in China’s atmospheric CH4 outputs at a scale equal to NGV production rates. This observational study will be conducted in early 2022 and will conclude upon graduation in mid 2023. Overall, my research will measure atmospheric CH4 emissions along the YRD in response to the rise in NGV production, a proxy for urbanization, and the continued loss of paddied rice fields for urban development. Areas along the YRD experiencing the highest volume of atmospheric CH4 concentration are likewise regions with the highest portion of cropland and urban area extents. The YRD is a good study area for this multi-faceted project due to the region’s complexity – cities teeming with millions of people, infrastructure ill equipped for transportation emissions, and area rich for agricultural productivity. This study’s complexity (urban expansion, agricultural loss, rise in new transportation technologies) may be used for future regional analyses and scenario forecasts for areas with similar intricacies.
Brianna Fernandez - Comparison Between Before and After the 2016 Flash Flood in Greenbrier, West Virginia
West Virginia is known for its gorgeous Appalachian Mountains and breathtaking views, but that mountainous terrain creates conditions ideal for flash flooding which are only exacerbated by the increases in extreme precipitation resulting from climate change. Flash floods have affected large swaths of the state, most notably and recently in 2016. This flood caused twenty-three deaths and extensive statewide property damage. In Greenbrier County, fifteen people died, and many small towns were left in ruins. This project studied White Sulphur Springs, a city within the affected area that has been subject to major flooding events and was particularly hit by the 2016 flood. By performing change detection analysis through image transformation, band difference analysis, and topographical modeling, I found a clear correlation between the terrain and the extent of damage in the region. Further, I used the West Virginia GIS Technical Center’s WV Flood Tool to examine at-risk structures and the potential cost that flood damage could have on them, furthering my analysis by seeking the best flood mitigation efforts for the area and the economic effects that they could have.
Derek Fucich - Mapping Gopher Tortoise Habitat Through RSPF Application
The Florida Scrub is a rare and exceptionally biodiverse ecosystem of central Florida where fires routinely maintain the ecology of the region. The federally threatened gopher tortoise (Gopherus polyphemus) maintains a sizable population in the Florida Scrub and excavates burrows into the soil which many scrub species use for refuge from fires. Due to the heavy reliance of other species on the gopher tortoise and their burrows, Archbold Biological Station engages in the longest gopher tortoise population study in history. Since the study involves manual surveys, efforts are restricted to a small portion of Archbold’s 5,193-acre property. In order to locate additional areas of encountering an active gopher tortoise burrow with high probability, the methods of a recent study were adopted to implement a Resource Selection Probability Function (RSPF) on the region of interest. The final RSPF gave an estimate of the probability of finding a gopher tortoise mound at each 10 x 10m pixel of Archbold’s property based off of a logistic regression model. While the final result is in need of a greater sample size in order to be more informative, additional suitable habitat for gopher tortoise burrow construction were identified and can serve to guide future explorations in search of additional tortoise populations throughout Archbold.
Megan Grimes - Quantify Glacial Retreat, Grey Glacier
This report analyzes the recent retreat of Grey Glacier through satellite imagery in conjunction with temperature data. Images were classified in ENVI 5.6.1 and surface area loss was quantified through differences in classified pixels. Increasing temperatures in the southern Patagonia region as a result of climate change appear to be responsible for the acceleration of the retreat of Grey Glacier as well as other glaciers in the Andes.
Maggie Hart - Examing the Agricultural Impact in Northern Afghanistan Following Taliban Occupation
The Taliban took over political control of Afghanistan in August of 2021. Since then, the health and well-being of the nation has declined as foreign aid pulled out of the country and a devastating drought continues to impact agriculture – Afghans main source of income and food.1 Given security concerns, international researchers cannot enter the country and, due to lack of funding, internal researchers have limited ability to assess the country’s agriculture state. Using simple methodology and applying NDVI vegetation indices as a proxy for agricultural integrity, I use MODIS data to examine regions of change in Afghanistan’s northern provinces over a 16-day period in the years 2019 to 2022. Policy – both globally and within Afghanistan – must be strengthened to address the current, dire needs of the population. To this, I recommend increased methodology and improved result dissemination plans. This policy reform can increase external aid as well as provide insight to steer the country’s food security planning.
Teddy Horangic - Drone-based Classification of 800 Untouched Islands
The Myeik Archipelago, Myanmar is a group of 800 tropical islands along the nation’s Southern coastline. It is home to some of the most biodiverse and intact coastal ecosystems in the world. The islands have been sequestered from large-scale human impacts for over 60 years due to the isolationist policies of Myanmar’s ruling military regime; but as of early 2015, that was rapidly changing. A sweeping transformation in government policy sparked interest in developing the area for tourism. In the years between 2015 and 2020, 16 new resorts were permitted for construction, and 5 began operating in the islands. Rapid and uncontrolled development – especially tourism-driven development – is not new in Southeast Asia, and can result in inordinate losses to biodiversity and habitat. However, careful planning and negotiation of environmental and developmental goals can help avoid such losses, while also promoting improved economic outcomes for local populations. The first step in any environmental monitoring and modeling effort is securing baseline data on the state of the local ecology. Due to its isolation, impoverishment, and the regime’s suppression of academia, there is a dearth of baseline data on the Myanmar region. New and affordable imagery tools, like drones, offer an incredible opportunity for the creation of high-accuracy remote sensing datasets to aid in environmental monitoring and development policy assessment. In this paper, I investigate the methodology for using drone-imagery captured from 5 years of annual surveying in the Myeik Archipelago to enhance the accuracy of a Sentinel-2 land classification map of the Northern region of the Myeik Archipelago, Myanmar, with the goal of comparing land-use change over the years from 2017 to 2021. I discuss the challenges with using such imagery for classification, as I was unable to complete my own supervised classification. However, I believe that with a few more weeks of time and some tweaks to my drone surveying process, the tool could be useful for aiding increasing the accuracy of classification. I also conducted an unsupervised classification of the region to investigate what classes were captured without training data. I then use VIIRS Nighttime Light data to map changes in electricity usage – a proxy for human habitation – to understand if heightened human impacts might be reflected in a comparative classification scheme. This is ultimately a methods paper documenting the trial-and-error lessons from using DJI recreational drone imagery for land-use classification to understand human impacts.
Camilla Ledezma - Quantifying Land-Use/Cover Changes (LUCC) in the Central Venezuelan Llanos
As demonstrated by the scientific literature, land use patterns in Venezuela’s Llanos del Orinoco region—part of the second-largest savanna region in South America—have never been mapped comprehensively using remote sensing techniques. In an attempt to begin to fill this gap, this project comprises an initial assessment of land use/cover changes (LUCC) in a central section of the Venezuelan Llanos from 1997 to 2017. Employing Landsat 5 (TM) and Landsat 8 (OLI) imagery, I conduct a supervised classification of the study area to visually depict land-cover and subsequently quantify changes among a number of land-cover classes, including savanna, forest, agriculture, and infrastructure. By evaluating the resulting transition matrix, I find that the largest transition in this area has occurred in savanna regions that have been converted for agriculture. I also measure other attributes of the land classes, including their gross gains and losses, their total and net changes, and their tendency to persist over time. Still, there remains much to be understood about the scale and implications of such landscape transitions in the Llanos. I thus encourage future research to undertake a more detailed and comprehensive assessment of land-use changes and present practices in this region, whose savanna and forest ecoregions demand conservation attention as they are increasingly exploited for agriculture.
Ella Lubin - Compare NDVI before, during, and after 2019 extreme drought in Australia
Between 2017 and 2019, Australia experienced the worst drought in the country’s history. In 2019, the country recorded the lowest average annual rainfall since 1902, and temperatures were 1.50C higher than average.1 In a country heavily dependent on massive agriculture and livestock industries, this drought meant steep declines in profits and brought to light perpetuated environmental and water impacts of the industries. This project compares the NDVI at the start, at the peak, and two years after the drought occurred using Landsat 8 satellite images. The region being observed is in the middle of the Murray-Darling Basin in New South Wales, an area that contains several of Australia’s largest rivers and supplies many people, livestock, and crops with water. The images from 2017, 2019, and 2021 were transformed using a tasseled cap transformation to better emphasize the differences between water sources, bare soil and vegetation, three aspects of the landscape that widely vary throughout the course of and after the drought. To further investigate the differences in vegetation growth and NDVI as an indicator of drought impact in the region chosen, an image of differences was produced and analyzed. The project presented some difficulties and challenges, as the image available for 2021 contained slightly different types of data, yet most analyses were still able to be completed. The NDVI comparisons and disparities between the three years cannot be explained solely by examining the drought’s impacts, however, as water usage in the area is also due to human activity and industry.
Adriana Maciel Metal - Map Seasonal Distribution of Urban Green Space in Bay Area, CA
This report maps the vegetation coverage over a region of the Bay Area using zonal NDVI statistics (mean and standard deviation) and demographic data at the city level and analyzes the significance of correlations between vegetation coverage and demographic variables at the city-level using Ordinary Least Squares regression analysis. This report also entails spatial autocorrelation analysis across this region using the Univariate Local Moran’s Index. Ultimately, this report suggests the existence of vegetation coverage trends in the Bay Area that simultaneously support and refute the validity of ongoing processes of gentrification, environmental racism, and ecological inequity (in terms of access to green spaces). This report also suggests that there could be either an ecological spillover effecting that leads to clustering patterns in vegetation coverage across the Bay Area, which may or may not be indicative of socioeconomic clustering as well. As issues of environmental injustice, inequity, and climate change become increasingly relevant to the present day, it is essential for researchers to examine how patterns in vegetation coverage between and within Bay Area communities may help to either mitigate or exacerbate human and non-human urban resiliency. Fundamentally, such data will help urban planners, developers, policy makers, and grassroots organizations make strides toward improving the human and non-human ecosystem health and resiliency of urban Bay Area cities.
Urmila Mallick - Examining Landscape Characteristics and Soil Properties in Botswana’s Makgadikgadi Landscape
Animals affect the environment through various biogeochemical processes that must be considered carefully when planning regional carbon budgets. Until recently, most of these zoogeochemical effects across ecosystems remained unacknowledged and therefore may be largely skewing regional and global carbon budgets. Soils store about 75% of the terrestrial carbon pool and can sequester varying levels of carbon depending on the variety of wildlife, livestock, and human activities on the landscape. My thesis research explores the effects of wildlife versus livestock on soil carbon in northern Botswana’s Makgadikgadi landscape by studying the landscape using satellite imagery, collecting soil samples at camera trap sites, and analyzing wildlife and livestock detected in camera trap images. In this project, I used Landsat 8 and ASTER DEM data to examine landscape and soil characteristics in northern Botswana’s Central Kalahari Game Reserve, Central Boteti West region, and lower Okavango Delta. By generating and comparing NDVI, NDMI, and Tasseled Cap indices, overlaying indices on digital elevation models, and classifying images using unsupervised methods, I establish a preliminary understanding of the land cover and characteristics in the region.
Jon Michel - Quantify the Land-Use Change of Big Sagebrush Ecosystems in Central Wyoming
In this study, I used two methods to quantify the area of vegetation lost due to fossil fuel development in the Jonah Field, a site located in Sublette County, Wyoming. I used two methods to perform this analysis: first, supervised classification for a period of years, and second, estimating the number of pixels in an image that experienced unusual changes in NDVI. The first method proved to be the most useful, showing that approximately 50 km2 of vegetation has been lost since development started in 1996. The second method didn’t produce reliable results, although it was effective on a smaller scale. Ultimately, I found that sagebrush decreased ~2.2 km2 yr-1 while bare soil and light vegetation increased 0.61 and 1.56 km2 yr-1 respectively.
Wyatt Nabatoff - Analyzing the Effects of Rapid Urbanization in Charlotte-Mecklenburg County
The southeastern city of Charlotte, North Carolina has experienced rapid population growth and expansion over the last few decades. To better understand the effects of rapid urbanization and population growth in the region, two Landsat 5 images from the summers of 1984 and 2011 were analyzed and compared. Analysis techniques such as NDVI, Tasseled Cap, and Supervised Classification were utilized to quantify urban growth, assess vegetation change, and understand how the land use has changed over time. The analysis concluded that urban areas expanded by 6%, which is equivalent to 660 square kilometers. These new urban areas expanded in areas that were previously suburbs, vegetation, and bare soil. However, the area’s vegetation metrics had not noticeably changed. This was likely spurred on by the surprisingly expansion of forests and vegetation over this 27-year window.
Jonathan Rigby - Examining Wildfire Detection Through GOES 16 and 17 Imagery
For remote sensing of wildfires, this presence of clouds presents a problem. Using sensors attached to a satellite, how are those sensors, capturing various wavelengths of radiation,supposed to identify wildfires if they are lightning caused, and underneath a wildfire complex.The objective of this paper is twofold. The first is to examine the August Complex through a series of GOES-17 MESO 2 images on August 18, 2020, and see if there are any identifiable markers on band 7 (3.9 micrometers) that might indicate wildfire. The second objective is, with an identified active wildfire location on August 18, 2020, see when it might be visibly identifiable through cloud cover, and attempt various analysis of cloud masks and band math on band 7 to see if I can classify that pixel as a fire pixel while it is still within cloud cover. Presently, detecting a wildfire with a GOES satellite through a wildfire may not be feasible. However, the timescale of GOES images being at the minute level, sometimes as low as a minute between images, presents the opportunity to perform relational analysis across time. As the spatial resolution of GOES imagery continues to improve, increased study should be done on time series analysis for wildfire detection through cloud cover.
Meredith Ryan - Land Use Classification of an Area in Nebraska and Comparing NDVI Values of Three Different Crops
This study took place at the Yale University Center For Earth Observation. Preexisting map data was used to perform a supervised classification on Landsat images from 1982 and 2001. NDVI was calculated for each of the individual classes of vegetation, resulting in seemingly higher values for 2001 presumably caused by severe rains and therefore delays to the growing season. However, the difference between both years and type of crop was not found to be significant. This may be the result of human error, especially for the corn. It is worth investigating more accurate methods for future studies.
Aranzazu Soto - Link Petroleum Extraction in the Ecuadorian Amazon to Deforestation and Reduced Vegetated Area
Deforestation in the Amazon rainforest is on ongoing problem that continues to be a subject of study by the scientific community. This study focuses on a source of deforestation that is petroleum extraction. To do this, geospatial analysis of Landsat satellite images was used. The area of study is the Ecuadorian Amazon, more specifically oil block 43, where the system of petroleum blocks used by Ecuador can be used to study a specific region of interest, and the known history of oil drilling in the blocks throughout time aids in data collection. An important piece of history of this area that is critical to this project is the Yasuni ITT Initiative that was an effort to conserve the landscape inside block 43 by the Ecuadorian government which took place from 2007 to 2013. The initiative fell through in 2013, making the area susceptible to oil drilling thereafter, making this year a critical point in time for this project, as it allows me to study this area’s vegetation in the context of oil drilling. Additionally, this area is of significant importance due to its location being inside a national park reserve. This study analyzes satellite images in the time range from 1998 to 2022. The main objectives of this study are to track deforestation in this sector of the Ecuadorian Amazon and link it specifically to petroleum extraction versus other deforestation sources. This is done by preprocessing Landsat images in order to take the NDVI (vegetation index) to compare the change over time.
Richard Sturtevant - Wetland Restoration in Eel River Headwater Restoration
Several cranberry bogs in the early 21st century was abandoned in Massachusetts as a result of decreased profitability. Government programs, such as the Eel River Restoration Program, offered farmers a “green-exit” strategy that restored the wetland habitat to a natural state. This project found an increase in Near Infrared reflectance (NIR) over a period of 12 years in an 8km2 region surrounding the Eel River in Plymouth Massachusetts. This wetland habitat was restored by the Eel River Headwaters Restoration Project, which restored several cranberry bog wetlands throughout Plymouth Massachusetts, resulting in an increase in vegetation health. Higher NIR reflectance correlates to a higher vegetation index (NDVI), and is a useful metric to observe restoration of natural habitats over a specific time period using remote sensing. A 9.5% increase in mean NDVI was calculated from the start of the restoration in 2009, compared to a nearly restored site in 2017.
Katy Sun - Analyzing the Rapid Rate of Deforestation in Kalimantan, Indonesia from 2001-2011
To analyze the rapid rates of deforestation in one of Indonesia’s palm plantation hotspots, Kalimantan, we utilize remote sensing. Two LANDSAT 5 TM images—one from 1988 and the other from 2010—are obtained using the United States Geological Survey’s Earth Explorer for analysis. The images are taken at path 118, row 62. To prepare the data, we conduct extensive cloud correction and manual masking. Afterwards, a comparative NDVI and a maximum likelihood, supervised classification are performed. We conclude that deforestation and the establishment of palm plantations are primarily concentrated around bodies of water. The data indicates 36.16% of native forest loss between 1988-2010, though we discuss why this metric is likely an underestimate of deforestation.
Riley Wadehra - Mapping the Extent of Savanna Ecosystems Over Time in Kruger National Park, South Africa and Surrounding Area
Savannas cover ⅙ of the land surface and account for 30% of primary productivity of terrestrial vegetation (Grace et al., 2006). They are characterized as open, grassy landscapes with interspersed tree cover, and are home to a unique set of fauna as well as many human communities. Their status is of serious concern due to their function as critical habitat, as well as their major impact on the global carbon budget. Therefore, it is imperative that we understand drivers of vegetation change in these systems. In this study, I aim to answer whether vegetation cover in Kruger National Park and the surrounding area has changed over the past twenty years, and whether potential changes can be related to fire activity. I first use Landsat products to evaluate vegetation change from Normalized Difference Vegetation Index (NDVI) and albedo. I then compare these results to burned area maps from the global fire atlas to examine a potential relationship between burned area and vegetation change, where I expect to find a close association between vegetation change and fire activity.Fires are undoubtedly a driver of vegetation change, however, using multiples avenues in this study I could not find a clear connection between fire activity and vegetation change. In both analyses I found vegetation change that was unaccounted for by fire, some of which may be explained by weather conditions. A major drought occurred in the growing season of 2016 in southern Africa, which would account for the decrease in NDVI particularly seen in the second analysis (Di Liberto 2016). The increased NDVI seen over the entire twenty-year time period could be due to several factors, including general climatic changes, variability in land management, or even fire activity that is not represented by this specific dataset. Although I could not make sense of the relationships between fire and vegetation in this study, a more specific, better temporal resolution project that follows the methods of the second analysis may have better luck understanding these relationships.
Shelby Warrington - Assess the Effect of Urban Expansion in Houston on High Flood Risk Area
Over the past several decades, Houston, Texas has experienced significant urban expansion into surrounding rural areas. Additionally, major flood events in the area have increased in frequency, a trend that is expected continue over the next century due to climate change and sea-level rise. The main goal of this project was to assess urban development in Houston, Texas over the last twenty years and explore how this development relates to areas of high flood risk. This goal was further divided into three main objectives: (1) determine how much urban landcover has increased in Harris County from 2002 to 2022 (2) determine what natural landcover types were most commonly converted to urban area and (3) explore how patterns of development relate to areas of high flood risk. The results indicate that urban landcover in Harris County has increased by 52% or around 956 km2 in the past twenty years. Most of this development was low-density. Additionally, around 40% of both wetlands and agriculture landcover in 2002 was converted to urban land by 2022. Lastly, around 50% of the new development fell in either moderate or high flood risk areas according to FEMA flood map delineations.
Blake Weyerhaeuser - Land Use Changes Around Minneapolis, Minnesota from 2001-2020.
I sought to identify patterns and quantify land use changes around Minneapolis, Minnesota using images from August 2001, August 2011, and August 2020. I used supervised classification to create seven classes: Water, Urban/Suburban, Urban/Suburban Light Vegetation, Crops, Bare Soil, Dense Vegetation, and Clouds. I used Landsat 5 and Landsat 8 Analysis Ready Data, along with ENVI and Google Earth Pro, in addition to my 21 years of experience with the area, for ground truthing in order to manually classify the three images and produce RGB change detection results. I then ran classification statistics on the classified images to quantify land use changes. Four conclusions are able to be derived from my work. Firstly, cloud coverage severely impacted the 2020 image, making full image comparisons to the 2020 image challenging. Errors in image metadata meant that a cloud mask was not able to be built. Despite this, the expansion of Minneapolis, Minnesota is visibly detectable and growth is concentrated in North Minneapolis. Conservation efforts around the St. Croix River are working, proving that the partnership between the Minnesota Land Trust and local landowners is a viable method of nature conservation. Lastly, there was a major increase in Minnesotan cropland between 2001 and 2011, which is consistent with outside data sources.
Yiheng Zhou - Monitoring Emperor Penguin Colonies at Cape Washington from 2013-2021 Using Landsat 8 Imagery
Emperor penguins, Aptenodytes forsteri, are key indicators of environmental shifts in Antarctica. However, necessary knowledges about population dynamics are limited due to difficulties of conducting field surveys under extreme weather conditions. Recent studies have shown that satellite imagery can serve as an effective alternative to detect and track emperor penguin colonies, mainly through the unique spectral signature of their guano. The purpose of this project is therefore to explore the potential of remote sensing data to better understand the abundance and distribution of emperor penguins. Five historic Landsat 8 OLI images, including 2013, 2015, 2017, 2019, and 2021, are used to quantify the biennial changes in guano covered areas at Cape Washington (-74.6455, 165.388). NDFI (Normalized Difference Feces Index) and NDSI (Normalized Difference Snow Index) are calculated as an attempt to automate this change detection process, then supervised classification with Maximum likelihood method is performed to create land cover maps in each year. Moreover, another monthly sea ice area dataset is imported to link the changes between emperor penguin colony size and sea ice level.