Spring 2021 - Project Abstracts
Sebastian Baez - Land Use Change in Nassau County, NY, 1988-2016
Nassau County is an older suburb of New York City and primarily has suburban, beach, wetland, open vegetation, and some urban land use for its 1.36 million inhabitants (in 2016). Between 1990 and 2016, the county’s population increased 5%; as the county absorbed this population, how did land use change? I analyze land cover change by performing supervised classification on Landsat 5 and 8 true color images of both the county and multiple counties around the metropolitan core. In both images, there is a gradual progression of land use change, with the dominant changes being open vegetation to suburban and suburban to urban land cover classes. Using classification on the multiple county image reveals greater suburbanization than the single county image, and both images misclassify urban and wetland pixels in the 1988 image. Nonetheless, the suburbanization and urbanization trends for the county are similar in both images, demonstrating the validity of this simple classification procedure while acknowledging the limitations in the representation of land use in the given land cover categories.
Jacqueline Buonfiglio - Investigating Changes in Human Impact Outside Protected Area in Kwazulu-Natal, South Africa
African wild dogs (Lycaon pictus) are an extremely endangered species, and are highly affected by human impacts. The KwaZulu-Natal (KZN) province of South Africa contains an essential population of wild dogs, and my Master’s thesis focuses on investigating how these animals move in a landscape characterized by abundant anthropogenic activity. In this project Landsat 4-5 TM and Landsat 8 OLI-TIRS images are obtained for the study region in June 1998 and June 2020 respectively. These images are clipped to the extent of the study region to exclude protected areas, and classified into five land use categories: agriculture, urban, water, forest, and bare ground. Class and change detection statistics are calculated to investigate how land cover has changed between 1998 and 2020. Class statistics showed similar shapes and reflectance for the urban, forest, and bare ground classes, with decreases in the percentage of pixels in the urban, agriculture, and water classes from 1998 to 2020. Change detection statistics show that water, urban, agriculture, and forest mostly became bare ground, while bare ground mostly became forest. It is likely that urban pixels were often misclassified as bare ground due to the similarities in reflectance patterns between the two classes. Furthermore, the intense drought occurring in South Africa during the period in which the 2020 images were obtained is the most likely cause for the high level of turnover to bare ground. Future studies should investigate if similar trends hold true during the rainy season.
Michela Catena - Evaluation Land Cover Change in Central Vietnam to Inform Flood Hazard Analysis
Loss of green vegetation, especially trees, has adverse impacts on flood risk, especially when combined with climate change. Increased urbanization around the world is leading to more severe flood events. Central Vietnam experienced extreme storms and flooding for weeks, which caused major damage, and many people blame climate change, urbanization, hydropower, and mining. This study examines land cover change in this region using a Landsat 5 image from 1989 and a Landsat 8 image from 2021. Atmospheric correction, supervised classification, and change detection analyses were conducted. Overall, there are been significant conversion of natural vegetation, agriculture, and sand in the region.
Jeamme Chia - Understanding Land Cover Change in Sepaku District, East Kalimantan Province, Indonesia (2000-2009)
Remote sensing is a powerful tool to assist policymakers and environmental advocates in understanding environmental degradation. This is especially important in regions where field studies are challenging due to geographical and political complexities, as well as when the environment is facing significant threats. A significant threat facing the environment in Sepaku District in East Kalimantan, Indonesia, is the construction of Indonesia’s new capital. To understand the impact of the development on the environment, a land cover change analysis was performed on the region. The findings will be compared to later imagery in 2023.
Melissa Halstead - Fracking Pipeline Landscape Change
New technologies and legal framework have contributed to the rapid increase of hydraulic fracturing in the northeastern United States within the last 10 years. Studies investigating the specific impacts such as pollution have been completed. However relatively few studies have investigated landscape change as an impact of the increase in hydraulic fracturing. Those that do investigate landscape change investigate the construction of well pads and do not consider the impact of pipeline construction. This study aims to investigate the landscape impacts of pipeline construction through classification using regions of interest and comparative NDVI in the region of Beaver County Pennsylvania. Results from the comparative NDVI analysis indicate that the process of clearing forest for pipelines has significant habitat fragmentation effects in the region. This could have implications for small terrestrial migratory species found within the area.
Erin Lippitt - Land Use Change Detection in the Southeastern Adirondacks
The Adirondack State Park has seen increasing visitor rates in the past two decades due to efforts to improve park accessibility and has a growing population of seasonal residents. While accessibility to all visitors is important, there are adverse environmental effects to wilderness regions that are experiencing overuse. Particularly, there is land use change causing natural vegetation to be converted into land for infrastructure or bare soil regions. I have used Landsat imagery to try and measure these effects in the highly trafficked southeastern region of the Adirondacks through change detection analysis. While some trends showing the environmental impacts of overuse can be seen, the continued use of a diverse array of land change analysis methods should be encouraged. This will allow for accurate monitoring of the land and inform policy on protecting the Adirondacks from human overuse, while still maintaining features of accessibility.
Yihong Zhu - Forest Cover Change UnderUrbanization Background from 2000 to 2020 in Fuzhou, Southern China
Since 1978, national policies like Reform and Opening up policy lead to rapid urbanization process in southern China, which contains abundant forest resources. The land-use change caused by urbanization drives losses of natural area, which then threaten biodiversity and affect ecosystem productivity. However, recent research also found that the southern China is turning green with increasing leaf area of vegetation. This raises the question: whether forest cover increased at the city-level and how did it change? To explore this question, historic Landsat satellite images of 2000 and 2020 are used to quantity the change in forest cover and urban land. Land cover maps were developed for each year with five classes: forest, agriculture, water, urban and bare land. Class statistics and change detection statistics were used to quantify the land-use change. Also, NDVI index was calculated to detect vegetation change.
Result shows that forest doesn’t change much in both location and area. 80.24% forest area remains to be forest. Only 16.40% degraded to bare land and 5.2% transformed into agricultural land. Planation program and Grain for Green program may play an important role in maintaining forest. NDVI change detection showed that vegetation loss mainly happened in or around urban areas. Urban area increases significantly from 401.96 km2 to 1616.94 km2, and the expansion is obvious in the map even though urban only occupies a small percentage area. The total area of water, agriculture land and bare land didn’t have significant change. However, their distribution did change through transformation among different land-use. One problem is that water tends to be misclassified into urban, thus requiring manual edit after the auto-classification. Methods that can distinguish water from urban land should be applied in the future to improve the accuracy of classification and class change detection.
Eudora Miao - Characterizing the Change in Forested Area Around Noggang Village from 1990-2020
The highly biodiversity and endemic subtropical forests of South China Karst is spread on a heterogenous and complex landscape. The forest cover change in the area is highly dynamic and local as impacted by human disturbances. In this project, I aim to use remote sensing to attempt to classify the rugged, challenging terrain and quantify the forest area change in a 100 km2 area around Nonggang Village in Guangxi, China with two Landsat images from the November of 1990 and 2020. I utilized indices such as NDMI (normalized difference moisture index) and MSI (moisture stress index) to separate out karst forests from agricultural lands, then performed supervised classification with Maximum Likelihood methods with the Landsat bands and the two indices. The classification was cleaned and then change detection was used to compare the land cover in the two years. A net loss of 2.48km2 (~3.053%) of karst forests was detected, mostly on the periphery of existing agricultural areas.
Annie Polish - Multi-Instrument Observation of a Glacial Feature on Mars
Mars is known to have large water-ice caps at its poles, but much of the rest of its surface may also have been covered by glaciers in the past. The extent and history of martian glaciation is very relevant to the ongoing scientific effort to understand the history of water and potentially life on Mars. In this project, I select a single potentially glacial site on Mars, and examine it using data from the HiRISE and CRISM instruments flying on the Mars Reconnaissance Orbiter. I use qualitative methods and image classification to demonstrate that the site is very likely to have a glacial origin, and comment on the utility of remote sensing methods typically used for Earth data in the new context of Mars. Then, I introduce the analysis method of Spatial Fourier Transforms, and demonstrate its potential for classifying and measuring glacial terrain on Mars.
Carlos Carrillo-Gallegos - Urbanization of Conservation Area in Southern Mexico City
Mexico City is the largest city in North America with just over 20 million residents in the metropolitan area. However, over 50% of the city’s land is environmentally protected (Sanchez), meaning that private residences are not allowed to be constructed. The city’s large population coupled with wealth inequality and high real estate prices have resulted in the establishment of technically illegal but significant communities being built in the environmentally protected land (Sanchez). These form one of the largest clusters of informal settlements in the world (Lopez et al 2016), and they pose a significant humanitarian and environmental concern. My project aims to use image classification to quantify how much land in the south of Mexico City, where most of these communities are, converted to urbanized land from 2000 to 2020. The analysis found that about 15.83% of the Natural Land in the region we focused on was lost, primarily to urbanization, and that a significant portion of that land was within the protected regions.
Raymond Zhao - Utility-Scale Solar Farm Development and Its Impact on Land Use, NDVI, and Land Surface Temperature in California’s Imperial Valley
The Imperial Valley is a historically agricultural region that has seen an increase in utility-scale solar farm (USSF) development since 2012. Although large PV solar power stations play an important role in the global clean energy transition, it still has nontrivial local environmental impacts. Development of USSFs and change in land cover, Normalized Vegetation Index, and land surface temperature are investigated using Landsat 5 (TM) and Landsat 8 (OLI & TIRS) images. Statistical, visual, and time series analyses revealed that about 89% of USSFs were built on agricultural land between 2011 and 2021. Agricultural land decrease by over 140 km2 and USSFs accounted for about 17% of that change. USSFs were shown to have an influence on NDVI and ST in the area, particularly in the southern region. These findings have various regional climate implications. Therefore, future PV Heat Island Effect research coupled with the integration of advanced remote sensing experiments at the appropriate temporal resolution are needed to monitor the impacts that the growth of USSFs will have on the Imperial Valley.
Benjamin W. Stern - Pittsburgh’s Changing Vegetation Cover: Estimates & Inferences Using Satellite Imagery
The city of Pittsburgh has a relatively high proportion of vegetative cover compared to many other United States cities. Recent mayors have prioritized tree plantings and urban forestry to increase vegetative cover. So how have these efforts fared? This project will utilize satellite imagery (Landsat 5 & 8 images) to see how vegetative cover within the city limits have changed from 2001 to 2016, both at the city-wide scale and at the neighborhood-level scale. Using the ENVI program and the NDVI index, a change detection map was created. With the change detection map, pixels were classified into areas of no vegetative change, vegetative increase, and vegetative decrease.
The city overall saw a slight increase in vegetative cover (~3%) from 2001 to 2016. There also existed spatial, racial, and socioeconomic patterns regarding which neighborhoods saw increases (and decreases) to vegetative cover. The neighborhoods experiencing the greatest increases to vegetative cover were poorer and more racially diverse than the city average, while the areas seeing the greatest decreases were whiter and wealthier. More research is needed to understand why this surprising finding is the case.
Carolyn Savoldelli - Quinnipiac River Marsh Vegetation Succession
Monitoring vulnerable salt marshes over time can improve our understanding of these systems and inform management practices. This study uses maximum likelihood classification to geospatially examine salt marsh vegetation changes in a brackish tidal salt marsh on the Quinnipiac River. It identifies native vegetation, non-native vegetation, and loss of vegetation to mudflat in select years from 1974-2016. The vegetation lost to mudflat area peaks in 2000 when roughly a quarter of the analyzed marsh and river area is mudflat, up from less than ten percent in 1974. The native plant evaluated, Typha latifolia, was reduced to a sixth of its original presence while non-native phragmites more than doubled. However, recent years indicate a rebound through the growth of native Spartina alterniflora. Careful understanding of these results will help predict future vegetation succession and identify paths for management.
Rachael Ross - Using Remote Sensing to Analyze Smallholder Farmland Expansion in Loibor Siret, Northern Tanzania
Over the last two centuries, habitat loss driven mainly by the expansion of agriculture has disrupted many ungulate migrations in East Africa. As a result, ungulate populations are in decline. The Tarangire-Manyara Ecosystem including Tarangire National Park, one of East Africa’s most important wildlife habitats with large numbers of migratory ungulates, is experiencing rapid agricultural expansion. I used comparative NDVI and minimum distance supervised classification techniques on Landsat 8 imagery from 2013 and 2019 to visualize and quantify smallholder farm expansion in communal lands bordering Tarangire National Park. Comparing the class statistics of farmland in 2013 and 2019, there has been an increase of 470 hectares of smallholder farms. This represents a 1% increase in farmland each year.
Nora Hardy - Monitoring Vegetation Regrowth at Dam Removal Sites Using Remote Sensing Techniques: Two Approaches for Large-Scale Sites
Dam removals have increased in frequency in the United States, motivated by both ecological and economic factors. However, vegetation response to dam removal is currently understudied. Common management concerns include the potentially slow revegetation of drained impoundments, or colonization by non-native herbaceous species, halting or stalling ecological succession. I tested two classification techniques to quantify vegetation regrowth in two locations: (1) a large dam removal site in Elwha, Washington, USA, and (2) two small dam removal sites in southern Connecticut, USA. Maximum-likelihood classification revealed a modest increase in vegetation cover at the Elwha, WA site during the study period (2013 to 2017). A LiDAR-derived vegetation height model of the two Connecticut sites also indicated clear regrowth of trees over the 20-year chronosequence. While these techniques show promise for use in dam removal monitoring projects, a lack of field validation makes it difficult to draw solid conclusions about their accuracy. Additionally, these techniques do not reveal information about vegetative species composition, which is usually a key consideration in restoration project monitoring.