OEFS Abstracts - Spring 2006

Wendy Francesconi - Reforestation in Esparza, Costa Rica. Landscape changes as a consequence of improved agricultural practices

Costa Rica is a country that has experience dramatic landcover changes in the past fifty years. Since the 1950’s the country began transforming its territory by deforesting more than half of the forest covered land. The major cause of this deforestation was the clearing of land for cattle grazing pastures. Recent changes in agricultural practices and environmental policy have not only halted these high rates of deforestation, but also reverse this process by allowing the forest to regenerate. This remote sensing analysis estimates the amount of forest recovery that has taken place in Esparza, the Central Pacific Region of Costa Rica. By comparing the area of forest cover from a 1986 Landsat TM image with a 1997 Landsat TM image, I found a positive relation between the incorporation of ecological environmental policy and forest recovery.

 

Jennifer Adler - Mapping Blue Gum Eucalyptus in coastal California

Blue gum eucalyptus trees are an exotic invasive species in coastal California. Controversy has arisen over their removal in the name of native habitat restoration and fire safety. No clear information exists on how much eucalyptus is present in California, or the potential costs of removal. This project investigates remote sensing as a tool for identifying forest cover to the species level. Visible, near-infrared, and shortwave bands of San Francisco Bay Area ASTER images were used from spring of 2004 and fall of 2005. Certain bands were found to successfully display eucalyptus species as different from other common forest types such as redwoods and oak forests. Classification of forest species was not successful, but the application of thresholds to ten different bands across two seasons accurately depicted pixels that were likely to contain eucalyptus and exotic conifers.

 

Perrine Punwani - Deforestation around refugee camps in southern Guinea

Deforestation, urbanization, and land conversion were analyzed in a 10,000 km2 area surrounding the Nimba Massif bordering southern Guinea and Côte d’Ivoire. Refugees from Liberia and Côte d’Ivoire flooded into this region in the early 1990s and throughout the decade doubling the populations of the towns and surrounding areas. A number of sources have indicated that such large movements are responsible for serious land degradation, which later jeopardizes the human and environmental security in host countries. Landsat TM and ETM images from 1986 and 2002 respectively were subsetted and analyzed using four qualitative and two quantitative change detection techniques. It was determined that land degradation had indeed occurred to a great extent and urbanization had increased by over 100 percent. Country specific patterns of land conversion were also noted. No definite conclusions could be drawn regarding the causes for the degradation; it was unclear as to whether they could be attributed to the influx of refugees or to natural population growth and urbanization. Recommendations are made for further study and the utility of remote sensing for change detection around refugee camps is assessed.

 

Jessica Albietz - Tracking trends in deforestation in the Antongil Bay watersheds, 1996 - 2001

Agricultural and urban expansion have strained hydrologic and ecological resources around the world. The area around the Masoala Peninsula in Madagascar experiences some of the highest population growth rates in the country. Savoka, or mixed hillside cultivation, is a common land use characteristic in the region for the production of a mosaic of subsistence and cash crops including bananas, papaya, cassava, manioc, potatoes, vanilla, pineapple and cloves. This study uses remote sensing techniques to estimate the land use change over a recent 5-year period in the Andranomena River basin, part of the Antaninambalana River watershed. The Antaninambalana bisects the Makira Forest area, carrying water from the highland forested areas to Antongil Bay. Land use change in the Antongil Bay watershed threatens the quality and flow of water, in addition to the loss of forest resources and habitat with serious consequences for biodiversity and livelihoods.

 

Erin Cushing - Deforestation and land cover change in SE Madagascar

The southeastern region of Madagascar experiences significant deforestation and land degradation due to heavy pressure on forests for wood and charcoal, soil erosion from deforestation and overgrazing, tavy agriculture and desertification. The recent introduction of mining operations to the region poses additional pressures on the land change. Satellite imagery analysis has previously demonstrated success in land cover classification and change detection in Madagascar. Analog image interpretation applied to Landsat images at 0.6-0.7 µm (visible red light) can distinguish rainforests from surrounding savannah and secondary vegetation in regions of eastern Madagascar. This project aims to qualitatively and quantitatively assess the extent of deforestation in SE Madagascar. Landsat images from 1991 and 2001 are applied to assess overall vegetation change over the ten year period, using NDVI as a proxy for vegetation cover.

 

Caroline Simmonds - Comparing spatial deforestation patterns over time

In this project remote sensing image classification and interpretation were used to study miombo forest cover change over a thirteen-year time period in the border region of Malawi and Mozambique in southern Africa. Subsets of two Landsat images from July 1989 and May 2002 were evaluated with ERMapper® software using the following methods:

  • Comparative NDVI analysis
  • Supervised classification
  • Unsupervised classification
  • Calculating error matrices to compare classification accuracy
  • Visual inspection and quantitative comparison of unsupervised images

The aim of this project was to determine how forest cover has changed over time in addition to comparing how different classification methods quantify this change. Results indicate that primary miombo forest has increased while secondary miombo and lowland mixed shrub classes decreased over this time period. Supervised and unsupervised classification methods yielded qualitatively similar but quantitatively dissimilar results.

 

Ross Geredien - Mountaintop removal coal mining in southern West Virginia

Mountaintop removal mining is a form of coal strip mining that completely removes up to 300 meters of bedrock, timber, and soil from the summits of mountain ranges in the Allegheny Plateau region of the southern Appalachian Mountains. The environmental devastation caused by this extreme form of mining has warranted studies to document the landscape-scale effects of mining in this region. I used ERMapper® and ArcGIS® to analyze LANDSAT imagery from 1976, 1987, and 2002 to detect land use change and mining trends in a six-county study area of southern West Virginia. Total mining area increased by 47,183 acres between 1987 and 2002. Supervised classification was the best of three methods to classify land cover types and to differentiate between fresh mining activity and reclaimed mine sites. This analysis can serve as a heuristic tool to develop improved methodology for larger-scale studies of the entire coal mining region in Appalachia.

 

Ikuko Matsumoto - Mining impacts on vegetation in the Philippines

The project aims to explore the land use changes and clarification of southern Cordillera Mountain region, northern Luzon Island, Philippines. This is the area where the Ibaloi indigenous peoples’ dwellings are. Southern Cordillera Mountain region has been exploited since the beginning of twenty century for mining, dams and logging. There are four mining sites and three dam projects in this region, and indigenous peoples’ main dwelling are along the Agno River. Unfortunately, it was difficult to classify the small difference of land use along the river and to figure out the condition of the forest in this mountain region. However, I could recognize some significant changes of the area around industrial mining site, dam construction site, and upstream of dam reservoir especially on the NDVI comparison image.

 

Jessica Darling - Identifying marginal agricultural lands in Mato Grosso, Brazil

Agriculture has been the base of the Brazilian economy for centuries. Over the past thirty years, however, the nature of agriculture in this country has changed dramatically. Soybean cultivation is rapidly becoming the dominant crop and land use in the area and is expanding at an extremely rapid rate. Mato Grosso is now the largest soybean producing state with 5.4 million hectares in production. Agriculture in this region has many potential impacts on ecosystems including: loss of habitat, introduction of exotic species, decreased water quality, and infrastructure development such as roads and shipping.

The primary goal of my analysis is to conduct a land cover classification to determine areas under agricultural cultivation in Mato Grosso, Brazil. In general, land cover in Mato Grosso includes forest, savanna (“Cerrado”), cropland and pastureland. The goal of my analysis is to distinguish cropland from these other land cover types. Another goal of my analysis is to look at the differences in the NDVI and EVI products from MODIS.

 

Kanvaly Bamba - Comparing the recent drought to the climate of prior years

The goal of this project was to look into the recent drought that has hit East Africa this past winter. By using NDVI data to track vegetation cover, the aim was to show how badly the drought affected the region and highlight this change both numerically and graphically. To approach the problem classification of both the supervised and unsupervised type were used as well as thresholding. Numbers gotten through each method varied. However, they showed the same degree of change especially among the best classified land cover types. The conclusion was obvious: the affect of the drought was obvious, significant and devastating on densely vegetated regions such as commercial agriculture.

 

Anthony Didlake - Examining the influence of terrain and precipitation on the seasonal variability of vegetation in northern CA

The unique terrain of Northern California plays an important role in the local climate. The Coastal and Sierra Nevada mountain ranges tower over and surround the Central (Sacramento) valley. The steeply rising slopes act to enhance and localize rainfall by disrupting and influencing the flow of air. This resulting orographic precipitation occurs most often during the winter when the air flow is especially moist, having a significant impact on the large population. Localized heavy rainfall can combine with rapid snowmelt to produce flooding in many towns and large cities.

The seasonality of precipitation and its orographic enhancement influence the variation of the land cover in some way. The goal of this project is to determine this seasonal land cover variation and examine the influence of orographic precipitation through the use of satellite remote sensing. The spatial patterns of vegetation and snow cover can be derived from satellites and serve as a useful proxy for precipitation. Therefore, the relationship between land cover variation and terrain can provide a greater understanding to the effect of orography on precipitation patterns. A characterization of the regional orographic effect can lead to improved predictions of localized precipitation. An examination of its impact on land cover can also lead to more effective land usage and better preparedness for potential ground hazards, such as flooding.

 

Shawn Walker - Exploring diurnal temperature fluctuation in and around 2 metropolitan regions

Inherent in the definition of an urban heat island (UHI) is the notion that characteristics such as heat-absorbing road and building materials, wind-speed reduction by buildings, the drainage, by sewers, of the greater part of any precipitation, prevention of water penetration into soil by impervious material, the reduction of evapotranspiration due to reduced plant populations, and city smog reducing long wave radiation result in the temperatures in urban areas being significantly higher than surrounding rural areas during the hottest times of the year. Recent studies have built upon this paradigm to explore the difference between diurnal fluctuations in urban and rural temperature. There is an overall pattern that has been shown to consistently occur, where the presence of a tree canopy keeps temperature fluctuations moderate in rural areas and the lack of significant canopy in urban areas allows long-wave radiation stored over the course of the day to be released, allowing for cooling.

The primary objective of this project is to explore the capacity of remote sensing technology for playing a role in the analysis of the UHI phenomenon. Although remote sensing inherently causes one to operate on a broader landscape scale, the development of finer resolution technologies suggest that more and more detailed analyses should be possible. In this project, it was hoped that ASTER images with its fine 15m resolution in the VNIR could be used to classify apart an urban-suburban-rural landscape with a degree of reliability that would allow for the detection of UHI temperature patterns specific to those classified regions. For temperatures, land surface temperature data acquired from MODIS images would be used.

 

Jayoung Koo - Detecting landcover change influenced by development in Cheju Island, 1990 - 2000

The paper looked into measuring urban growth in Jeju Island, South Korea between the period of 1990 and 2000. As a popular tourism destination within the country and from the adjacent East and Southeast Asian countries, the number of visitors has grown by 37.4% during 1990 and 2000, while the island population only had a 5.4% growth. The increase of both populations has lead to land use pattern change. The urban areas have grown and there increased number of built structures along the southern coast could be detected.

Through three different classification methods, the project has detected urban land cover changes. However, the slight seasonal difference has influenced the land cover detection. The relatively bareness of the 2000 image affected the detection of urban results for the unsupervised classification. Nonetheless, supervised classification results reflected increased urban areas and the NDVI detection revealed locations of prominent vegetation loss that correspond with expanded urban areas and development.

 

Joel Creswell - Classifying impervious surface cover in New Haven County watersheds

In this project, I estimated impervious surface cover in four watersheds in New Haven County, CT using ASTER imagery. I used the visible and near infra-red bands of the images, dated January 24, 2001, April 30, 2001, February 27, 2004, and August 28, 2004, ensuring a 15 meter pixel resolution in all bands. I used two methods to estimate impervious surface cover: supervised classification and indexing. Supervised classification was the more accurate of the two, but indexing provided more information about each pixel. I determined that winter images were not suitable for impervious surface cover analysis, as snow and ice are too easily classified as impervious by both methods. At the small scale of the watersheds used in this study, inter-annual change detection proved difficult.

 

Shani Harmon - Remote sensing productivity in the Chesapeake Bay

Remote sensing may become a useful tool in investigating the eutrophication in marine and coastal environments. Satellite images are a good tool for identifying the occurrence of algal blooms. Scientists remain unsure whether estimates of the plant pigment chlorophyll-a derived from satellite data can support the evaluation of eutrophication in marine and coastal waters. Satellite images overestimate true chlorophyll-a concentrations in open sea areas by 60-70% and even more in coastal waters.

The goal of my project is to understand how different climate conditions, temperature and precipitation, affect productivity and contribute to eutrophication in the Chesapeake Bay. I focus primarily on primary productivity, since primary productivity is directly linked to eutrophication.

 

Rhead Enion - Measuring sedimentation rates in the Chesapeake Bay through satellite imaging

Estuaries around the world, including Chesapeake Bay, have severe water quality problems. Scientific monitoring of these large water bodies can be time consuming and expensive. By utilizing freely available satellite imagery to build upon existing monitoring techniques, the temporal and spatial resolution of water quality data can be increased with relatively little additional time or effort. This project concerns methods for isolating and quantifying sediment-impaired waters in MODIS satellite images. I have defined empirical relationships between specific MODIS satellite bands and secchi disk transparency by regressing the satellite data with existing station monitoring data in Chesapeake Bay. These regression equations are then used to create pseudocolor images to illustrate the range of secchi disk transparency values throughout Chesapeake Bay.

 

David Griffith - The impacts of land use on total suspended solid and chloryphyll a concentrations

Watersheds serve important ecological, economic, and social functions. The rivers and estuaries that drain them reflect underlying geology, climatic regimes, and are strongly influenced by upstream land use. This study attempts to characterize land use in the Hudson River Watershed (HRW) based on phenology and unsupervised classifications. Two approaches are compared:

  1. Seasonal MODIS NDVI and
  2. 7-band monthly MODIS spectral data.

Agricultural areas, deciduous and coniferous forests, and urban areas in the HRW and major tributary watersheds (Upper Hudson River and the Mohawk River) are characterized for each classification technique. Landsat TM images and USGS land cover data are used to interpret classes and check the accuracy of unsupervised MODIS classifications. The classifications are coarse but suggest that the HRW is 66% forest, 24% agriculture, and 8% urban, and that the Mohawk River sub-basin has a higher percentage of agriculture (39%) than the Upper Hudson sub-basin (15%). This has important implications for the ecological characteristics of the downstream river, estuary and ocean margin.

 

Kate Woodruff - The impact of water diversion projects on land use in the Lower Mekong River region

One of the most important natural resources is water. An accessible and reliable water source encourages settlement as well as development that also depend on water. Ubon Ratchathani is one of Thailand’s many provinces located in the far Northeast region of the nation. Most of the province consists of the Khorat Plateau, an area characterized by dry climate and somewhat infertile soil. The plateau is bordered in the North by forested land, in the East by the Mekong River which distinguishes the border between Laos PDR and Thailand, and in the Southeast and South by the Phanom Dongrak Mountain range. As it is, the primary land use in this area has become agriculture. The increase in agricultural development can be witnessed along with the increase in reservoir construction to ensure irrigation supplies in a region known for unpredictable rainfall trends and persistent droughts.

Using satellite imagery and remote sensing tools, a study has been conducted to try and characterize the agricultural development and surface water change in the region. This study, while only preliminary, will analyze several remote sensing techniques using three Landsat images, in order to greater understand, and quantify where possible, the land use change in the particular district of Nam Yuen in the Province of Ubon Ratchathani. To narrow the scope of the project, subsets will be focused on, namely a cluster of farm plots to the northwest of the district and the reservoir development in the center of the image. Applications of imagery analysis discussed hereafter will pertain to agricultural identification and analysis as well as surface water alteration. Finally, the analyzed region will be classified and processed in a variety of ways in an attempt to quantify the area of land use change as a whole. This paper will conclude that some techniques were more appropriate in this particular case study as well as the further implications of the project and possibility for expanded research.

 

Steven Brady - Predicting Marbled Salamander distribution based on surface ice cover

The focus of this project was to use high spatial resolution imagery to classify wetland attributes. In particular, the attributes of interest pertained to the known winter habitat requirements of the marbled salamander (e.g. frozen wetland vegetation, ice coverage, open-water). Ponds of interest were generally small, often ranging 250-500 m2 with some occupying less than 100 m2. IKONOS imagery was selected for analysis because of its high spatial resolution (4 m multi-spectral and 1 m panchromatic) and its availability. This project focused on conducting numerous supervised and unsupervised classifications. Expert review and ground truthing provided feedback to guide this process. Class type from the final classified image was quantified by pixel for each pond. It was predicted that marbled salamander presence would decrease with increasing ice coverage. A logistic regression relying on five years of presence/absence data was used to evaluate this prediction.

 

Fox Kral - Detecting glacial change on Mt. Kilamanjaro

Geological and climatic evidence indicate that the tropical glaciers on Mt. Kilimanjaro have been decreasing in mass as a result of local climate change. The use of Landsat TM images provides a rapid and accurate analysis and quantification of the decrease in area of the tropical glaciers on the summit of Mt. Kilimanjaro. The results support previous studies which indicate rapid deterioration of the glacier.

 

Denise Levitan - Assessing change in the extent of glaciers in Glacier N.P.

Montana’s Glacier National Park is currently home to thirty-seven glaciers. The shrinking of these glaciers has been observed in the area for the past one hundred and fifty years. Currently, this is believed to be the result of global climate change, an issue of great interest to many in scientific and political communities alike.

This study examines several techniques to quantify the land area covered by glaciers in the park and surrounding areas and to assess its change over the period of 1988 to 2004 using Landsat TM and ETM+ images. The use of three indices and two combinations of indices yielded differing results. In general, the overall trends showed an increase in glacial cover, though individual data points were somewhat erratic. The data obtained from this study suggest the possibility of other factors, such as local climate, influencing the remote sensing of glaciers in this region.

 

Critter Thompson - Changing snowpack & changing climate in the Puget Sound - Georgia Basin region

The climate of the Puget Sound-Georgia Basin region of the Pacific Northwest has been undergoing significant changes over past half century. Temperature is rising and snow levels are falling. In the years to come, this trend is predicted to continue. The region will likely experience wetter winters, with more precipitation falling as rain than snow, and drier summers. Such changes should not be taken lightly. The Puget Sound-Georgia Basin region relies heavily on flows and storage of water to maintain both human and natural ecosystems. This paper attempts to look at changing snowpack conditions throughout a portion of the region through the use of satellite remote sensing. Understanding and monitoring changes in regional snow cover will be a critical component of climate science in the coming years.

 

Kasia Wegrzyn - Seasonal ice pack and change over time of the size of the Ross Ice Shelf

In this study, I used seven 8-day composite images of a MODIS surface reflectance package to study the change in sea ice extent in the North Atlantic over the past seven years (2000 - 2006). One image per year was analyzed from March 6 - March 21, with an attempt to take all images from the same period. The study area is centered at 55.48° N and 61.32° W off the southeast tip of Labrador, in the Labrador Sea. The methods used to analyze the maximum presence and extent of sea ice in the North Atlantic included five indices-the Normalized Difference Snow Index (NDSI), the Normalized Difference Snow/Ice Index (NDSII), NDSI-band 7, Ice Index, and Albedo-and two visual comparisons of true-color RGB-143 and also RGB-367 images to comment on trends in ice distribution and thickness at its maximum extent. This analysis was then used to determine if in fact there is a decrease in the maximum extent in sea ice in the North Atlantic, as much literature suggests.

 

Na Xu - Investigating glaciers and landcover change in the Chile - Argentina border

Changes of mountain glaciers are among the most sensitive natural indicators of ongoing climate change. Globally, moutain glaciers generally have been retreating since the later part of the 19th century. This is especially the case in mid-latitude mountain ranges such as the Himalayas, Alps, and the Andes as well as isolated tropical summits such as Mount Kilimanjaro in Africa.

One of the most important consequences of global warming is the eustatic sea-level rise produced by the melting of mountain and other glaciers. Meier estimated that between 1900 and 1961 the contribution of mountain and other types of glaciers (ice caps and ice fields, but excluding the Antarctic and Greenland ice sheets) to global sea-level rise was 0.46mm/a. The contribution of Chilean glaciers to eustatic sea-level rise has been estimated to be approximately 8.2% of the worldwide contribution of small glaciers on Earth during the last 51 years. On the other hand, in areas that are heavily dependent on water runoff from glaciers that melt during the warmer summer months, a continuing glacier retreat will have potential impact on water supplies, such as the Andes of South America and Himalayas in Asia.

In remote mountains, remote sensing techniques are often the only way to monitor and analyze a large number of glaciers. Multi-temporal satellite image analysis is an important tool for monitoring the variations in the total area of glaciers, the position of their front and their general facies. In my study, three methods for glacier mapping with Landsat TM/ETM and SRTM DEM were applied to investigate the glaciers variation in area: to test whether or not glaciers retreat occurred in the study region; and to estimate the severity of the retreat during this relatively long time span. Particularly, the position change of three glacier tongues was examined in terms of elevation and length. In addition, based on the result of supervised classification, the landcover change is also presented.

 

Jenna Bourne - Evolution of Hurricane Katrina

The goal of this study is to examine several properties of Hurricane Katrina over a six-day period when the storm system was classified as a hurricane. Properties examined include eye width, hurricane width, hurricane area, speed, cloud top temperature, cloud top altitude, and ice index. An effort was made to correlate some or all of these properties to the relative strength of the hurricane. This project was executed fairly smoothly and the data obtained from the MODIS images seemed to correlate well with the information that was already known about Hurricane Katrina. The width, area, speed, cloud top altitude and cloud top temperature of Hurricane Katrina did not change predictably with the strength of Hurricane Katrina although the data were logical and self consistent with the other aspects measured in this project. Remote sensing seems to be a useful tool for observing hurricane storm systems for the properties examined in this study.

 

Kara DiFrancesco - Wetland loss after Hurricane Katrina

This project assessed the amount and type of wetland change in Breton Sound, Louisiana caused by Hurricane Katrina, which traveled through the area on August 29, 2005. Change was detected and quantified by comparing NDVI visually, quantifying NDVI differences, and assessing post-classification comparisons. Through this project it was determined that an accurate assessment of the specific types of wetland and species loss would require significant ground truthing, which is not possible at this time. None the less, assessment of the type of loss based on USGS land cover data from 2001 and newly created classifications allowed for some qualitative and quantitative estimates of wetland change.

 

Devin McPhillips - Identification of dust storm source regions using MODIS imagery

The Bodele depression in north central Africa is the leading source of dust in the atmosphere. Atmospheric dust influences Earth’s climate through albedo and fertilizes the oceans. The high temporal resolution of MODIS scenes provides a method for observing the ephemeral dust storms in remote areas. This project aims to develop a straightforward algorithm for enhancing airborne dust. The spectral signature of a March, 2003, dust storm in the Bodele depression is the basis of this algorithm. Ultimately, normalized difference algorithms are assigned to red, green, and blue guns in order to create a useful composite algorithm. The final enhancement product is used to explore dust sources and measure average wind speed. Unsupervised classification suggests that the enhancement is somewhat better than true color for examining diffuse dust.

 

Beth Feingold - Characterize the Augustine Volcano eruption of January 2006

Volcanic clouds can be detected using thermal infrared bands of MODIS and easily deciphered from meteoric clouds. The height of a volcanic plume can be assessed similar to how it is done for meteoric clouds, by comparing brightness temperature values to Radiosonde data. Applying a model to ascertain particle size and mass of ash portions of volcanic plumes is a harder task for a novice to achieve. This paper is my exploration of the use of MODIS to quantify these characteristics on the January 30th, 2006 Augustine Volcano eruption in Alaska.

 

Emily Hicks - Indirect effects of the Indian Ocean Tsunami - Assessing changes in inland forest cover

This research examined areas in Sumatra, Indonesia affected by the 2004 Indian Ocean tsunami. For three different study sites, the questions posed were:

  • What is the type and extent of land cover changes that are taking place in Gunung Leuser National Park, following the Indian Ocean tsunami?
  • What are the changes in vegetative cover that are taking place on Nias and Simeulue islands?

MODIS 16-day Normalized Difference Vegetation Index (NDVI) products were used to examine the period 2001 until 2006. Mean annual NDVI (as sampled in early January) was calculated for each of the three areas. Transects were also used to determine spatial patterns in NDVI change on an annual basis and (for the two islands) on a seasonal basis as well. In Gunung Leuser National Park, it was found that variability in NDVI is higher inside the park than along its edges, whereas the opposite was true for the islands. All three study areas showed observable declines in NDVI since 2002. Although this project demonstrated the potential of using NDVI data to highlight trends in vegetation, it is not possible with NDVI data alone to answer the more specific questions about the particular types and causes of landcover changes that are taking place.

 

Catherine Schloegel - Examining fire scar patterns in the Mesa Verde Plateau

Shortened wildland fire periodicity in the southwestern United States may be a direct effect of the 100-year policy of fire suppression which has allowed for the growth of dense forests, or it may be a regular cycle within arid-land ecosystem dynamics. This project will critically explore the use of satellite imagery to identify different-aged fire scars within pinyon-juniper woodlands on the Mesa Verde cuesta. Single image unsupervised classifications identified recent burn scars, but reflected considerable confusion between areas of bare soil, shrubland, and burn scars. The use of a digital elevation model and the National Land Cover Dataset to mask land cover classes outside the primary study area increased the accuracy of the burn scar identification. Image ratios including the normalized vegetation index and the normalized burn ratio were used to compare a pre- and post-fire image of the study area. Isolating individual land cover classes followed by analysis using both the vegetation and burn index served as the best method for understanding how different types of land cover respond to fire. The results of this project suggest that classified remote sensing datasets can aid in the exploration of interactions between pre-fire conditions, and biophysical settings which will aid in the interpretation of fire’s effects. All the methods tested provide an adequate basis for future analysis and application. It is hoped that a clearer understanding of fire frequency and vegetative response will assist policies makers across the Four Corners in designing ecologically-appropriate plans for landscape management.

 

Avery Anderson - Impact of land use change in the Upper Green River Valley, WY on large carnivore populations

Abstract Unavailable

 

Mohamad Chakaki - Landcover classification and change detection in Damascus, Syria

Damascus sits on an oasis, fed by the Barada River, on the edge of the vast Syrian desert-steppe. The city has been continuously inhabited for more than 5000 years. Readily available satellite images of the city, however, extend only as far back as the early 1970’s. While Damascus’ ancient history certainly puts observing change over time in perspective, the city has seen its most intense growth in population and associated urban expansion in the past 30 years. To the extent that this urbanization has been ad hoc and often informal (i.e. illegal squatter settlement), it is encroaching upon the Ghouta. This is the Arabic name given to the farms, fields and orchards that collectively make up the oasis surrounding Damascus. My objectives for this project were threefold, to classify the various land covers in and around Damascus; to identify the nature and extent of change in these land cover classes over times; and to do so using a combination of pre and post-classification change detection techniques. Landsat TM and ETM+ datasets from 1987 and 2000, respectively, were used to run comparative NDVI, unsupervised and supervised classification techniques. Both classification methods failed to adequately distinguish between urban land cover and bare soil or rock at the outskirts of the city. While this precluded any accurate post-classification change detection, the comparative NDVI did reveal a pattern to urbanization that is consistent with the trends I know to be prevalent in the city’s urban development. In sum, while both accurate land cover classification and the identification of a specific pattern in land cover class transition over time remained elusive, results on the overall pattern of urbanization in Damascus were acceptable.

 

Manisha Gangopadhyay - Tracking temporal changes of two rivers in Bangladesh

Understanding the patterns of urbanization and peri-urban growth is of critical importance in subtropical cities today, particularly in relation to flooding and waterlogging. Flooding during the monsoon season is a problem that many Asian countries are accustomed to, however, the extent of flooding and the ability to cope with this seemingly natural phenomena are exacerbated as climate change and ‘development’ is accompanied by the expansion of impermeable surfaces, among other things, continue . Remote sensing is a valuable instrument in understanding the patterns of urbanization and peri-urban growth, particularly in cities where fluctuating and seasonal population movement, undocumented citizenship and sheer numbers make census information difficult to assess. This is true for the city of Kolkata, India as well, where waterlogging and flooding during the monsoon season gives rise to unsanitary conditions leading to sometimes fatal diseases that affect scores of people, hinders mobility, business, and damages the infrastructure of the city. I use remote sensing to quantify spatially and separately, urban growth and peri-urban, and the loss of wetlands and vegetation related to that growth.

 

Anil Pokhrel - Land use changes along the Koshi River Basin of E. Nepal

The objective of this research is to compare land cover change within the Koshi River Basin using satellite imagery and to specifically observe if there are any direct impacts to the river regime arising from this change. I am specifically interested to find the change in snow cover and whether it has any relations to the water available in the rivers. Secondly, I am interested to other land cover changes such as forests, built up areas and agricultural fields.

 

Markelle Smith - Changing landuse patterns in the Grand Traverse Bay watershed

My goal in conducting this project was to use post-classification remote sensing techniques to quantify land use change in the GTBW from 1977 to 2001. The primary focus of the analysis was on the conversion of forest area to other land uses within this watershed. In particular I hoped to quantify the decrease in forest area and resultant increase in bare soil, suburban, and urban land uses that result from development and sprawl. I assessed these changes in the GTBW and in each of the three sub-watersheds that comprise the majority of the GTBW: Boardman River watershed, Elk River watershed, and Platte River watershed.

Although this analysis was no doubt influenced by some technical errors, it is possible to conclude that forest loss occurred in the GTBW between 1977 and 1988. In addition, it is likely that the most urbanized and developed of the sub-watersheds, the Boardman, experienced the majority of deforestation when compared with the Platte and Elk. Other more quantitative conclusions in regards to the study area would require more analysis and revised techniques.

 

Elizabeth Ra’cz - Changes in vegetation and water on Pine Ridge Reservation, 1979 - 2000

Vegetation change assessment over time is especially important for effective management in underserved and under-funded areas. The Black Hills, due to its sparse population and lack of highly trained personnel compared to the size of the area to be managed, is a region that could truly benefit even from rudimentary Satellite data analysis and interpretation. NDVI and supervised classification regions did a reasonable job at defining how much mixed coniferous forest was lost to the Jasper fire. In a single decade 1990-2000, 815 square miles of mixed coniferous forest were reduced to 504 square miles with 240 square miles of the reduction due to burned areas. General satellite analyzes can be helpful in perceiving and presenting long-term vegetation trends, specifically in the reduction of mixed coniferous forest and the increase in general vegetation (potentially second growth) in the Black Hills National Forest and surrounding regions. Understanding an areas long-term ecological trends permit more effective land management. It is recommended that a suite of satellite image processing procedures be conducted to properly assess landscape trends over time.

 

Emily Goble - Vegetation difference in the Tugen Hills/Lake Baringo area of Kenya

Faunal material has traditionally been utilized to study paleoecology but many ecological tools can be used to then extrapolate about the past. The main tool I plan to use in ecological analysis with application to paleoecology is remote sensing to analyze differences in vegetation in relation to lake level/area. In order to determine the utility of remote sensing to paleoecology a pilot study with modern data must be undertaken to determine the relationships between lake level, rainfall and vegetation. Once those relationships are established a model can be formed to use in paleontological applications. In this study the manipulation of two Landsat images from path 169 row 60 were used to detect change in water and vegetation.

 

Laura Jeanty - Studying land cover change in MA using MODIS imagery

This project studies the feasibility of using MODIS 16-day NDVI composite data to classify land use in Massachusetts. Three classification schemes are studied: unsupervised, supervised, and the application of thresholds. Each classification scheme is applied to a 3 part time-series of NDVI images (plus a single blue band) for the years 2000 and 2004, and the possibility of studying land use change by comparing the results of these classifications is discussed.

 

Kristen Welsh - Masking clouds in satellite imagery: Effectiveness in land cover classification

In the field of remote sensing, cloud cover can be a problematic issue, particularly in regions of world with high amounts of cloud cover, such as the tropics. While several techniques exist in order to compensate for this obstruction in viewing land cover on satellite imagery, cloud masking is one method to remove these problematic areas. In this project, I masked cloud cover in a 1979 Landsat Multi-Spectral Scanner (MSS) image and in a 2000 Landsat Enhanced Thematic Mapper Plus (ETM+) image and used these masked images to classify land cover for both years. I then compared these results with the same analyses performed with the unmasked images, prior to cloud removal. Through these comparisons, I assessed whether cloud removal through masking is an effective and worthwhile technique for these images and for similar regions where cloud cover might be an impediment to remote sensing analysis.

 

Amany von Oehsen - Comparing MODIS and Landsat scenes in the State of Mato Grosso

The aim of this project was to study how to detect human impacts on the natural vegetation in the area around Campo Novo do Parecis in western Mato Grosso in Brazil with different methods of analysis for two Landsat images (TM and ETM+) from 1990 and 2001. The techniques employed were visual comparison of the unclassified images, red green NDVI overlay of the two dates and supervised as well as unsupervised classification. Interesting observations about burnings could be made in the NDVI overlay. The attempt to quantify the increase in fields was made by using a supervised and an unsupervised classification. Both suffered from confusion between fire scars and certain fields and thus it was tried to quantify only the sum of fire scars and fields as a measure of “anthropogenically impacted area”. The question of where to set the threshold between fire scar and natural savanna was encountered which could not be answered due to lacking ground truth. Results of unsupervised and supervised classifications were compared.

 

Yue Wang - Vegetation mapping and spatial interpolation in the southwest of China

Forest cover type mapping is important for forest management and land use planning. Ground-based forest inventory is an important method to monitor and assess forest resources. However, conventional forest inventories for every piece of forest cover on the ground are not economically and timely feasible. Instead, we systemically or randomly sample floristic characteristics in a relative few sample points. To map forest cover types with limited available ground data, we need to use some other ancillary information to interpolate ground-based sample data to non-sampled area. Remote sensing data could be one of the best sources of such kind of information which offers complete spatial coverage for large areas in a consistent and regularly up-datable manner. The primary goal of this study is to compare the different classification schemes using to map forest types for a forest property in the southwest of China. Floristic characteristics gradually change over a landscape. Traditional distance-based classification methods may not able to distinguish these delicate changes. Several statistics techniques such as principle component analysis, discriminant analysis, ordination, and Knn estimation were proposed to analyze the variation of remote sensing data associated with floristic changes. Forest classification using supervised, unsupervised, and discriminant analysis are compared in the present study. Finally, the accuracy and potential usefulness of these classifications are discussed.

 

Kevin Lauterbach - Identifying waterbodies at different spatial resolutions

The purpose of the research was to compare the abilities of LANDSAT, ASTER, and IKONOS satellite imagery to identify small bodies of water. The study area was the Saugutuck Watershed in southwestern Connecticut, which contains numerous lakes and ponds of various sizes. Images were classified using both supervised and unsupervised classification methods in ER Mapper® 7.0. The classified images were then compared to a detailed study area to determine the limits of size detection and the overall precision of the classification. LANDSAT had the lowest limit of detection (LOD), being able to detect a pond with a surface area of 0.5 acres. ASTER was precise in delineating the area of water bodies but its LOD was over twice the size of the LANDSAT image. The IKONOS imagery had similar LOD to that of the LANDSAT image and similar precision as the ASTER image, but misclassified many areas as water that in reality were shadows.

 

Rebecca Sanborn - Developing conservation monitoring techniques using Landsat and ASTER images

Conservation and development are occurring at unprecedented rates in the United States, but technology and policies to manage land have not kept pace. As land trusts and governments acquire more and more land for conservation, they will need new ways of monitoring the properties and enforcing conservation policies. Remote sensing is a promising option, used successfully now to monitor the 762,000 acre Pingree Conservation Easement in northern Maine. While the Landsat-based classifications and multi-tiered monitoring work well for properties the size of Pingree, these techniques are of little use to land trusts with 50-100 acre parcels and potential disturbances on the scale of one to five acres, rather than a thousand times that. This paper examines the potential of remote sensing techniques to work at significantly smaller scales. Landsat has been considered too coarse in resolution to detect changes at the small easement scale, and my findings confirm that. I found that ASTER imagery was satisfactory for identifying changes in certain landcover types and at relatively large scales, but it will not accomplish all of the monitoring goals of most organizations. While there is great potential in remote sensing technology to meet the needs of conservation organizations, a number of other roadblocks will need to be addressed before it is a useful tool—imagery must be widely available, high quality, broad in scope, and relatively inexpensive. Image analysis—or the ability to become trained in image analysis-must also become more available to the conservation community.

 

Jason Nesbitt - Using ASTER to study archaeological landscapes in the Huanuco region of Peru

For the last several decades, archaeologists have shown a great deal of interest in studying archaeological sites from a landscape or regional perspective. In this respect, remote sensing has played an important role, typically used for identifying archaeological sites and features. In more recent years, another way in which archaeological landscapes have been modeled is through the use of Digital Elevation Models (DEMs). Yet one of the difficulties in the use of Digital Elevation Models is their frequent coarse grained nature. In this paper I compare an ASTER DEM with one from SRTM to assess the differences between them. As my case study, I examine the Huánuco Basin of Central Peru. The results showed major incongruities between ASTER and SRTM elevation values. Nevertheless, I argue that ASTER offers significant potential in modeling and visualizing archaeological landscapes.

 

Liza Lutzker - Examining deciduous forest fragmentation at two levels of resolution

Using remotely-sensed data to model the suitable habitat for Ixodes scapularis ticks in the eastern United States would require numerous images of fine spatial resolution, but fewer images of coarser spatial resolution. Given that I. scapularis lives in closed-canopy deciduous forests, a vegetation index might prove successful in identifying its habitat. Here, the Enhanced Vegetation Index (EVI) from the MODIS satellite (with 250m resolution) was examined for its ability to predict deciduous forest content as classified using Landsat data (with 30m resolution) in a study area in southern New England. Three different classifications were used to assess the relationship. For each classification, increasing EVI was found to be positively correlated with increasing deciduous forest content; however, this relationship is not linear in nature. Further studies need to be performed if MODIS EVI is to be used to identify I. scapularis habitat.