OEFS Abstracts - Spring 2013

Hanna Mershman - Forest Succession & the 1994 Northwest Forest Plan, Willamette Nat. Forest, OR

This study addressed land cover change on the Willamette National Forest as a result of the 1994 Northwest Forest Plan. The plan was developed to greatly restrict harvesting of late successional forests on public lands in Oregon, California and Washington. The objective was to preserve and prevent habitat loss critical to the survival of the threatened Northern Spotted Owl. The study explored both unsupervised and supervised classification of the region using Landsat 5 TM images from 1985 and 2011 in an effort to more accurately capture the impacts of this controversial policy event. Results show a 17.6% increase in Late Successional Forest Cover since 1985 and a substantial decrease in bare and cleared ground (59.1%) which is indicative of decreased harvest activity. As expected, harvested areas from 1985 showed evidence of successional recovery by 2011. While the results suggest the policy’s habitat conservation objectives were ultimately met, the validity of these findings is questionable due to the limitations of the study.

Jin Yin - Change Detection of Landuse in Leishan County, China

This course project is to examine land use in Leishan County via Landsat satellite imagery to analyze and evaluate Chinese Forest Policy. I worked with three Landsat images (1996, 2000, and 2009), applied unsupervised classification and supervised classification with change detection information. By comparing among different methods, the maximum likelihood classification presents the best results. According to the change detection information, the increasing of forest area indicates that the Chinese forest policies are effectively implemented in Leishan County.

Abigail Eurich - Analysis of Mountain Pine Beetle Damage Extent in Colorado and Wyoming

The Rocky Mountain Pine Beetle (Dendroctonus ponderosae) is a native species to the Rocky Mountain National Forest, but in the last 15 years, climate change and an increasing mean annual temperature of the forest has expanded its habitat and enabled it to attack new tree species in higher elevations that cannot cope with the increased rate of attack. Since the 1960s, remote sensing has been used to view the extent of the pine beetle effects on the forest but only recently has the technology allowed high quality image resolution and accurate assessment of the pine beetle extent in the area. This study analyses three Landsat TM and ETM+ images from an anniversary date of August/September, the season of highest RMPB activity, in 1997, 2003, and 2010 in order to track the effects of the RMPB in the study area. Qualitative and minimal quantitative assessment leads to a definite conclusion that the RMPB has had a large impact on the forest studied; however, further research is needed to more accurately determine a detailed quantitative assessment of the area affected. While mitigation and remediation strategies continue to be tested and implemented, remote sensing offers a method of assessing the situation from space and has played a large role in the current analysis efforts.

Carlos Gould - A Study of Deforestation in Rondônia, Brazil

In the past 30 years, Rondônia, Brazil has experienced massive amounts of deforestation. Now one of the most deforested areas of the Amazon, this west Brazilian state is the focus of this project. Forest clearing is done in a fishbone pattern visible on satellite images. As more forest is removed, further fishbone patterns are added throughout the years in attempts to penetrate and reach uncut forest. By analyzing these patterns, this project quantifies the vegetation losses between July 13, 1985 (the time of the first image) and September 1, 2012 (the time of the last image) and creates a map of the deforestation. Over this period of time, over 14 billion square meters of land have been deforested, representing a loss of 44 percent of the area of the image analyzed. The trend observed in this project is observable on massive scales across the Amazon and in other major tropical forests around the world.

Ambika Khadka - Forest cover changes in Xinjiang River sub-watershed in Jiangxi Province, China

This investigation focuses on studying the decreasing water levels in the Dead Sea over the 26-year span between 1984 and 2010. Using Landsat 5 images, the study finds increased agricultural activity during this time, particularly in the southeastern Syria, has been a major factor in the decreasing water budget of the region. Methods of analysis included NDVI and albedo statistics to classify areas exhibiting the most change. This project aims to shed light on the scarce and interconnected water dynamics in the politically volatile region.

Jonathan Sullivan - Estimating aboveground tree biomass at Yale-Myers Forest

Abstract Unavailable

Lexi Tuddenham - Analysis of Haiti reforestation

Historical and contemporary deforestation in Haiti has contributed to a widespread problem of erosion and topsoil loss in the steep topography of the mountain regions. The dramatic elevation changes paired with the extremes of the wet and dry season. In addition to endangering an already vulnerable population with natural disasters like landslides, topsoil loss also indirectly exacerbates malnutrition and other diseases of poverty that are widespread across these regions, as inhabitants find it harder to grow food or harvest tree products for income. Similarly, loss of tree cover contributes to the drying up of already sparse water sources in the dry season, and makes it difficult to raise livestock who need shade to survive in the hot temperatures. The Haiti Timber Reintroduction Project has worked in the Artibonite Valley of central Haiti since 2006, training farmers in soil conservation and tree cultivation that has resulted in the planting of well over 600,000 trees to date. This study investigates the possibility of using Landsat Satellite imagery to detect land cover change from HTRIP’s reforestation efforts. Though Haiti is famous for its deforestation, relatively little attention has been paid to mapping land cover change through satellite imagery. This study demonstrates both the utility and the limitations of using Landsat imagery to detect land cover change.

Peter Umunay - Detecting Land Cover Change in the Ituri Forest, Democratic Republic of Congo

The objective of this project is to detect, classify, and examine the change in land cover and land use in the Ituri Landscape using Landsat Thematic Mapper (TM) 30 meters spatial resolution images, 1998 and 2010. The analysis of spectral signature and NDVI values in the scenes was used to identify different forest types in the region; afterward classification techniques supervised) were performed to classify land cover. Band combination, visual interpretation and comparative Normalized Difference Vegetation Index (NDVI) and comparative biomass index were used to detect the changes. The results of this project can be applied to evaluate the impacts of human activities in forest cover, assess the nature of future changes in the areas, and to develop a land use planning system that helps managing forest for human livelihoods, carbon sequestration and biodiversity conservation.

Kimberly Lay - Classification of Salvador, Bahia, Brazil: Implications for Urban Slum Health

Background: Salvador is the third largest city in Brazil with more than 60% of its population living in slums where overcrowding and lack of infrastructure are contributing risk factors for health and health access. Land cover classification would be beneficial to identify areas of formal versus informal land use and predict areas at high risk for disease transmission.
Objective: This project aims to determine the land cover classification of the city of Salvador, Brazil by categorizing the landscape and determine if land cover by slums or, favelas, differ compared to more formally developed areas.
Methods: Land cover was first explored and analyzed using RGB composites, NDVI and Tassle Cap Transformation and preliminary unsupervised and supervised classification to identify areas of interest and any mixed spectral signals. Unsupervised and supervised classifications by ROIs were then used to create classes and describe the urban landscape. Texture analysis with Sobel filter was also used to compare these different areas of the city.
Results: Minimum distance provided the best results for supervised classification and Isodata provided the best results of land cover classification in supervised techniques. In addition, unsupervised methods were also able to differentiate two different urban land cover based on tile or corrugated roofs.
Conclusions: Land cover classification in describing formal and informal urbanization is possible with Landsat imagery provided it is supplemented with texture analysis and demographic information. Advanced techniques in multi-spectral analysis and principal component analysis may improve the classification process. The generalizability of this type of analysis requires research into the region to understand the evolution of slums..

Nicole West - Mapping malaria vector habitat in Sarpang province of Bhutan

Abstract Unavailable

Matt Bare - Measure land use change in the Colombian Amazon

Abstract Unavailable

Emanuel Feld - Land cover change in the Rufiji River and Delta region

This study examines land cover change in the Rufiji River and Delta region between 2000 and 2011, a decade of frequent and severe drought. Beyond rainfall stressors, the vegetation in the region is affected by growing demand for paddy cultivation in the Delta, which entails cutting mangroves, tree felling for logging and agricultural activities, grazing cattle herds, and bush and forest fires. Using Landsat TM and ETM images of Path 166, Rows 65 and 66, the project examines the extent and location of these impacts through supervised and unsupervised classification. It also seeks to determine whether classification becomes more accurate with the inclusion of bands representing the Normalized Difference Vegetation Index, the Normalized Difference Moisture Index, a Normalized Burn Index, and elevation information. We conclude that including these indices generally allows for more accurate change detection, however difficulties remain in distinguishing agricultural areas from soil and grasslands.

Molly Roske - Quantify land cover change over time in the southeastern Azuero Peninsula of Panama

Using Landsat TM-5 images from 2000 and 2009 to detect and quantify land cover type change in the eastern part of the Azuero peninsula of southwestern Panama faces a number of challenges. Though a seasonally dry region, interference from clouds and cloud shadows is high even at the end of seasonal rains, posing difficulties in capturing images of the deciduous vegetation before leaves drop with the onset of the dry season. Additionally, small-scale changes in vegetation type resulting from cattle ranching practices, shifting grazing and/or irrigation, and narrow riparian buffers are not necessarily captured by the Landsat pixels at 30m resolution. This analysis explores methods of removing cloud (and cloud shadow) interference from anniversary images, to evaluate the degree of successful classification and change detection. To deal with remaining cloud edges’ interference even after the greatest extent possible of clouds has been removed from images, the integrity of all methods of classification was best maintained by allowing cloud edge as its own class; thus, total vegetation change is difficult to assess in this area, because calculated class statistics reflect change in cloud edge and thus may not give a reliable estimate of forest cover change..

Sumana Serchan - Detecting land use land cover in the northern region of Lake Champlain Basin

Abstract Unavailable

David Tan - Tracking Changes in Vegetation Cover in Tropical Lowland Singapore

Abstract Unavailable

Lucia Woo - Change detection analysis in the western coastal areas of Aceh, Indonesia

Assessment of damage and monitoring of recovery from a natural disaster is a long and expensive process. The technology of remote sensing, however, provides freedom and flexibility to conduct such assessments and monitoring with relative ease and minimal cost for an area of any size for as long or short of duration. This project sought to identify the type and extent of damage and recovery on the western coast of Aceh, Indonesia, from the 2004 Indian Ocean earthquake and tsunami. The project used strategically spatially subsets of Landsat 7 ETM+ satellite images from December 2004, January 2005, and January 2011 to conduct change detection analysis of landcover. It employed a variety of change detection methods such as unclassified classification and additive color to compare pre- and post-disaster scenes. Such analysis revealed the 2004 disaster caused subsidence, loss of vegetation, and creation of new water bodies inland, and these damages were limited to shorelines and low-elevated ground. The recovery is slow but evident: regrowth of vegetation, shrinking of new water bodies, and retreat of sand and mud landcover Given the recent events such as Hurricane Sandy and earthquakes in China, the importance of remote sensing will rise as we seek to better prepare for and predict the outcomes of natural disasters.

Henry Glick - Modeling Cougar Habitat in the Northeast

Abstract Unavailable

Beth Tellman - Land Cover/Land Use Change for Rainfall-Run-off Modeling in El Salvador

Abstract Unavailable

Meredith Azevedo - Effects of Urban Expansion on Agricultural Lands: A Delhi, India Case Study

Urban expansion is proceeding at an unprecedented rate (Seto et al. 2011). As urban centers extend their reach, surrounding suburban and rural land is overtaken. This study focuses on land lost to urban expansion around Delhi, India from 1998 to 2011.
Specifically, this study attempts to quantify the amount of land lost to urban expansion during this thirteen-year period and characterizes this land loss by land use type. A special focus will be placed on the loss agricultural lands.

Ben Friedman - Observing Population Growth in Douglas County, Colorado

The 2010 US census revealed that Douglas County was one of the fastest growing counties in the country, experiencing a 62% increase in population between 2000 and 2010 (census.gov). I acquired two Landsat 5 TM images (Path 33 Row 33) from July 4, 2000 and August 17, 2010 to study land-use change in Douglas County in between the 2000 and 2010 censuses.

Particular emphasis is placed on the conversion from shortgrass prairie to urban development. Both unsupervised and supervised classification methods are used, although minimum distance supervised classification is the dominant analytical tool. Following previous studies in land cover classification and urbanization, land use change statistics are calculated (Yuan et. al, 2005, Campbell and Wynne 2011). This study specifically analyzes land cover change surrounding three State Parks that are within Douglas County. Qualitative and quantitative analysis of the results provide data about fast-changing Douglas County, and may be used to suggest potential future conservation designations.

Esther Rojas-Garcia - Observation of land-use change of the top megacity worldwide: Delhi, India

This study examines land-use change in Delhi across eleven years. The objectives of this study are to (i) characterize urban land cover and (ii) quantify land use change. The images used for this study are Landsat TM 5 images with summer dates of May 4, 1998 and May 18, 2009. To examine land-use change, this study uses two types of unsupervised classification (K-Means and IsoData) and a minimum distance supervised classification. Overall, the three types of classification produced different results but the trends were similar. Specifically, urban land cover increased and vegetation and rural land cover decreased. An interesting finding of this study was the change in shape of the Yamuna River and a reduction of water cover.

Kellie Stokes - Quantifying Urban Extent and Urban Greenspace Changes in Tianjin China

Abstract Unavailable

Thomas Rokholt - Exploration of pollution & human activity on vegetation in Sognefjord region of Norway

Satellite data can be an important tool for understanding and detecting landscape changes over a broad area. The purpose of this project was to examine forest cover trends in Sognefjord, one of the fjord networks located in western Norway, and the second longest fjord in the world. The center of focus was the town of Flåm, located at the terminus of Aurlandsfjord, a popular cruise ship docking port. The geography of the region prevents the escape of pollutants, and allowed for a focused study of pollutants if other variables could be controlled. Change in forest cover, while limited by spatial resolution, could indicate both anthropogenic pressure as well as other broader climatic pressures.

Anna Wade - Environmental changes caused by Mountaintop Mining

This paper will illustrate the environmental changes caused by increased mountaintop removal in southern Appalachia from 1987 to 2009. Landsat images were acquired from 1987, 2000, and 2009 for analysis. The techniques of unsupervised and supervised classification were used to map active mining and reclamation sites over the 22-year period. Both classification techniques, while able to distinguish between forested and non-forested land, had difficulties separating mining sites, reclaimed grassland, and urban centers. Hyperspectral remote sensing has potential to monitor environmental risks from mountaintop mining, such as the spread of grassland and the rise of slurry ponds.

Madeline Yozwiak - Quantify the impact of shale gas infrastructure on the Pennsylvania landscape

The growth of shale gas has moved production to new regions of the country, primarily Pennsylvania and rich Marcellus shale. This paper seeks to develop a method to quantify vegetation loss as a result of drilling, in order to better understand the extent of deforestation and land clearing for well pads. Focusing on Bradford County, which has the highest concentration of permitted unconventional wells, the results of the analysis indicate that the relative vegetation loss within areas immediately permitted rigs was greater than that of the county as a whole.

Sarah Ditchek- Remote Sensing of SST, Chlorophyll A, and NDVI off the Coast of Peru

Using remote sensing data in ENVI from Landsat TM, an exploration was undertaken of ocean and land areas near the Peruvian coast during ENSO, El Niño and La Niña, years. Ocean analyses included visualizing SST and Chlorophyll A fluctuations. Land analysis included the study of vegetation fluctuations. Results gathered show that it is possible to detect changes between ENSO events through remote sensing. SST increased in the El Niño image and decreased in the La Niña image as expected. Additionally, a non-uniform spatial distribution of ChlA and an increase in vegetation can be found in the La Niña image while a uniform to semi-uniform spatial distribution of ChlA and a lack of vegetation can be found in the El Niño image.

Allegra Gordon- Seasonal change in Arctic Sea Ice

Sea Ice is important as it helps cool the global climate because of ice’s high albedo. This study looks at the change in land cover, albedo and sea surface temperature of an Arctic scene from path 28, row 8 in Baffin Bay, between Greenland and North America. Two Landsat-5 TM images were used to identify changes between two seasons: May 26, 1990 when ice fully covers the ocean and June 12, 1993 during ice melt. Four land classifications were interpreted as pure ocean water, mixed pixels with a low fraction of ice (such as ice chunks that are smaller than 30 meters floating in open ocean), mixed pixels with a high fraction of ice (such as melt ponds or slush), and pure ice. These four classes were confirmed using K-Means and ISODATA unsupervised classifications, as well as by comparing the values from albedo and sea surface temperature calculations. The area of water increased from May to June by 1,077%, coupled with a decrease of ice by 38%. The melting of ice, and in consequence, the increased coverage by water may be due to an increase in sea surface temperature of 13.5 K in June. In addition, the albedo of the sea ice scene decreased by 21%, from 0.593 in May to 0.365 in June. With a rise in temperature, a decrease of albedo and a land cover change favoring more water and less ice during the melt season, an ice-albedo effect may be at play. Although this study shows only the change between two seasons, there are implications for an ice-albedo effect that may trigger enhanced warming in the northern latitudes with worldwide consequences.

Azusa Takeishi- Inter-annual variability of wintertime precipitation over the Rocky Mountain

Abstract Unavailable

Johanna Press- Comparison of Dust Identification Methods in the Middle East Region

Understanding the sources and transport of desert dust has profound implications for the studies of global environmental change and human health and safety, yet the similar reflective properties of mineral aerosols and desert surface in the Middle East make it difficult to identify dust in unadjusted satellite imagery. This paper compares the effectiveness of four algorithms for dust source and plume identification using MODIS data:

Ackerman’s model of brightness temperature difference (BTD)
Roskovensky and Liou’s dust identification algorithm incorporating short wave reflectance ratio and long wave BTD
The Normalized Difference Dust Index (NDDI), a ratio of visible and infrared reflectances
The Middle East Dust Index (MEDI), a ratio of BTD

It is demonstrated that thermal-based algorithms are more successful than techniques using reflectances alone. Of the models applies and analyzed, MEDI is best able to counteract unique challenges of regional conditions to distinguish both dust source and plumes above the desert surface.

Jenna Hessert- Environmental Change in Etosha, Namibia

The Etosha Salt Pan in Namibia, Africa is a very dynamic system throughout the year with the true name of the “Great White Place of Dry Water”. The pan alternates its environment from one of a water basin during its rainy, summer season to a salt wasteland in its dry, winter season. A Landsat image from May 28, 2010, which is at the end of the rainy season, was analyzed with a variety of different methods including NDVI, NDMI, albedo, and surface temperature. Using masks made from the NDVI and NDMI, everything except the salty lake sediments was eliminated. A spectral analysis was done on several types of salts in the lab and their signatures were compared to the signatures of the sediments in the Pan. Although it was not possible to detect the different types of salt present, the amount of salt was able to be detected. A Normalized Difference Sediment Index (NDSI) was created and could be used to detect the amount of lake sediment (comprised of salts and clay particles) present in a certain location. This has the potential to determine the suitability of a land for agriculture. The 2010 image was then compared to a May 2009 image where there was much more rainfall, and to a May 1998 image where there was light rainfall and little water retention in the Pan. Differences and similarities were studied in the NDVI, NDMI, albedo, surface temperature, and NDSI.

Sonam Choden- Melting Glaciers in the Himalayas: in The Punatsang Chhu River Basin, Bhutan

Bhutan is blessed with enormous fresh water resources and at the same instance challenged with constant threat from natural disasters. The fresh water resources such as the high altitude lakes, glacial lakes and glaciers are scattered on the northern end of the country that feeds into the river system. However, the fragile northern glacial lakes and retreating glaciers are like time bombs that can cause a devastating effect when changing climates and global warming could increase the rate of glacial retreats and cause devastating floods downstream. Such an incident was recorded in October 7, 1994 where: the Punatsang Chhu watershed was flooded and the nation experienced a real time flood event killing dozens of people on its path and destroyed the ancient capital of Bhutan.

This remote sensing project, where two annual Landsat images (Oct 30, 1999 and Oct 1, 2009) that were a decade apart was analyzed to see if there were regional glacial melts. The results from the three methods – Maximum Likelihood Supervised Classification, Thresholding of TM4/TM5 Ratio and Thresholding of Normalised Difference Snow Index – indicates that glacier (snow and ice) is melting in Lunana, Bhutan – the highest mountain relieves found on Earth.

Israel Kositsky- Observing the Decreasing Water Levels in the Dead Sea

Ellen Arnstein- Measuring Water Level Changes in Lake Titicaca

The most commonly cited statistic evidencing climate change in the Andes is that the level of Lake Titicaca, which straddles the border of Bolivia and Peru, has fallen by 2.6 feet, the lowest level since 1949. Using remote sensing techniques, I investigated this change over time using Landsat TM5 images from July 1990 and 2010 to compare the area of vegetation and shallower water directly surrounding the lake.

Yiqi Zheng- Examining the Responses of Different Land Types to Drought

Land surface properties respond to drought differently on different land types, which further has a feedback on atmosphere and energy. In this study, MODIS satellite products of Land Surface Temperature (LST), Normalized Difference Vegetation Index (NDVI) and Land Cover Type, combined with GAMO MERRA precipitation reanalysis have been used to examine the impact of 2012 central U.S. drought on grassland, cropland, deciduous broadleaf forest and urban. The drought has the largest impact on grassland, whose NDVI decreased by 40% and LST increased by 4% in 2012 compared to the normal levels in a reference year 2004. A following timeseries analysis on fine spatial resolution and a general correlation analysis on coarse resolution in central U.S. both indicate grassland is the most sensitive to dry conditions, and forest is the least sensitive. All these analyses suggest the role of soil and vegetation type played in surface feedback mechanism: a drier, thinner soil layer is more susceptible to drought due to more reduced evapotranspiration, greater damage to vegetation health and rapid increase of surface temperature. The final detection of land cover change in central U.S. from 2001 to 2011 suggests such a short-term land surface change is not reliable to contribute to the severity and persistence of drought.

Xiangying Shi- What caused the Forest Degradation? Differentiating Droughts and Deforestation

Abstract Unavailable

Luke Cartwright- Examination of the 2012 heatwave and drought in the UK

Abstract Unavailable

Emily Farr- Trends in Agriculture in the Po River Valley

Abstract Unavailable

Frieda Fein- Expansion of irrigated agriculture in the Western Nile Delta

In 2007, the Egyptian government, with financial backing from the World Bank, announced an extensive irrigation project to convert significant acreage of desert to farmland along the edge of the western Nile Delta. Prior to the implementation of this irrigation project, the role of Egypt’s agricultural sector in the country’s economy had begun to decline. This decline was reputedly in part a response to market liberalization and in part due to expansion of other sectors of the economy. How these trends played out in the West Nile Delta, however, is unclear. Satellite remote sensing was used to examine trends in land use in the western Delta region to attempt to better understand the conditions that preceded the Egyptian government’s implementation of such an extensive desert reclamation project.

To determine changes in agricultural patterns in the West Nile Delta, Landsat 5 TM images from the two decades before the new irrigation proposal were analyzed using a variety of change detection techniques. Albedo and Normalized Difference Vegetation, Moisture, and Dust Indexes were calculated and used in a time-lapse comparison to determine overall trends of changes in land use. Supervised classification was then used to quantify the area that had undergone changes in usage over the two decades.

Michael Nestler- Center Pivot Irrigation Agriculture in the Arabian Desert

Abstract Unavailable

Philippa Stoddard- Identification of pre-colonial Andean agricultural terraces

Remote sensing can be a valuable tool for detecting archaeological sites in areas that may be difficult for researchers to survey on the ground. Many archaeological features, however, are too small for satellites to detect reliably. In this study I develop a method to detect agricultural terracing, a common feature of pre-colonial Andean settlements, using ASTER imagery. Because terraces almost always occur in conjunction with other structures like storehouses and roads, the presence of terraced agriculture is a good proxy for archaeological finds that would otherwise be difficult or impossible to detect remotely. Using ASTER and ASTER’s 30 meter Global Digital Elevation Model (GDEM), I was able to combine slope and vegetation thresholds to predict the location of terraces in a section of the Urubamba Valley, Peru, and surrounding agricultural areas. Digital ‘ground truthing’ with Google Earth revealed the reasonable accuracy of these predictions; however, more work is needed to refine the model to increase its predictive power.

Beata Fiszer- Wallow Fire: Burn Scar characteristics

The single post-fire approach and multitemporal approach in detecting fire scars were compared using LANDSAT TM 5 images of Bear Wallow Wilderness which was affected by the 2011 Wallow Fire. The spatial and spectral capacity of LANDSAT TM allowed for the application of three fire detection methods within both approaches which were the burn index, the vegetation index and temperature. The burn index uses LANDSAT TM bands 4 and 7, the vegetation index uses bands 4 and 3 and temperature uses LANDSAT’s thermal band, band 6. Mean statistics, colormapping, transects and stretching techniques were used to identify the effectiveness of each method in fire scar detection. The use of two distinct post-fire images deepened the analysis to include the ability of these methods to distinguish the fire a period of time after the fire.

Kandice Harper- Impact of Fire on Albedo and Land Surface Temperature in the Alaskan Boreal Forest

Three classification techniques were applied to a Landsat TM image in east-central Alaska in an attempt to differentiate a burn scar from the surrounding undisturbed boreal forest. The three classification methods produced different results in terms of area classified as burn scar. The two methods employing the top-of-atmosphere reflectance of bands 4 (near-IR) and 7 (mid-IR) misclassified river-edge pixels as burned area. The 6-band supervised classification method excluded soil pixels from the burn class, while the other two methods grouped soil with the burn scar class. The impact of these differences on the calculated fire-induced changes in albedo and land surface temperature were minimal. In the fire year, elevated land surface temperature and reduced albedo are observed in the burn scar perimeter for all classification methods.

Natalie Schultz- Surface heat budget analysis of deforestation in Indonesia

Dramatic land use change has been taking place in Indonesia over the past few decades and is expected to accelerate in the future, due in large part to the expansion of oil palm plantations. Little research has been conducted in this region on the impacts of deforestation and the growth of oil palm on regional climate. The objectives of this study were to examine how the conversion of tropical rainforest to intensive oil palm plantations affected the surface energy budget. A Landsat TM image was obtained over Kalimantan, Indonesia from January 16, 2010 (path 119, row 62). A 3036km² subset of this original image was used for the analysis. Clouds were removed from the image through a dual classification using brightness and surface temperature. A supervised classification was performed on the image subset, and the seven detailed classes were clumped into “forest” and “palm” for the energy budget analysis. The Landsat image was used to calculate the radiative fluxes, while the turbulent fluxes were estimated using a combination of satellite data, empirical relationships, and parameters sourced from the literature.

The results of this analysis found that deforestation and the conversion to oil palm altered the energy budget components. The surface temperature in the palm regions was warmer by 0.9K compared to the forests. The net radiation was reduced by 21 W/m² as a result of a higher albedo and increased longwave radiation flux. The NDVI was reduced from 0.61 to 0.52. The sensible heat flux increased by 20 W/m² and the latent heat flux decreased by 47 W/m², resulting in a Bowen Ratio increase from 0.24 to 0.33. While these results are preliminary and approximate in nature, they show that land use change in this region can have important feedbacks on regional climate. Future work should incorporate surface measurements for validation, as well as quantify these changes on larger spatial and temporal scales. Additionally, this research could be expanded by using satellite observations to quantify changes in carbon storage and carbon fluxes due to deforestation in this region.

Ryan Laemel- Surface Albedo Measurements of New Haven

In recent years, urban areas have become the focus of many remote sensing efforts because of unprecedented urban growth and warming over the last century. Many analyses of urban areas investigate the impact of surface albedo and NDVI on urban temperature and the formation of urban heat islands. Most of these studies have focused on large cities. As a consequence, few studies have addressed the issue of scale for urban warming and heat islands. This study is an attempt to analyze small-scale urban warming as a result of low surface albedo and NDVI by measuring surface albedo, NDVI and temperature in LANDSAT 5 TM images of a small mid-latitude city, i.e. New Haven, CT. Results indicate New Haven, CT was 2.85°C warmer and saw lower surface albedo and NDVI values than surrounding areas on average over a 15-year period. Accordingly, urban warming in New Haven, CT suggests urban warming and perhaps heat islands occur at a range of scales.

Lindsi Seegmiller- High Resolution vs Low Resolution: A comparative study of satellite imagery analysis

Urban vulnerability to coastal hazards is an important, contemporary issue that has recently seen an increase in academic attention. Measuring exposure to hazards, however, can be difficult given the lack of available data and uncertain broader implications of site-specific measurements. Using remote sensing to measure urban exposure to hazards has the potential to provide a comprehensive, repeatable, and relatively accessible method of tackling this issue. Considering the variation in spectral and spatial resolution of satellite images, the limitations of each satellite must first be assessed in order to select appropriate data. This study begins this assessment by comparing Ikonos and Landsat 5 satellite images of New Haven, Connecticut. Using visual manipulation, NDVI transformation, and unsupervised classification comparison, four distinct differences were identified: shadow prevalence, unwanted class division, differences in land cover classification accuracy, and limited analysis availability due to spectral resolution. Given the findings, informed allocation of Landsat and Ikonos satellites will serve measurement of exposure to coastal hazards. The identified limitations of each satellite will also aid in assessing the accuracy of the resulting exposure measurements.