OEFS Abstracts - Fall 2007

Mila Dunbar-Irwin - Deforestation in the Palembang Area, Indonesia

Deforestation has become an ever increasing problem in Southeast Asia, particularly in Indonesia. This is a threat to biodiversity, carbon emissions, and local livelihoods. Remote sensing is an extremely useful tool in measuring the extent of the problem and composing threat maps to predict where further deforestation will take place. This project aimed to quantify the deforestation occurring in an area north of the city of Palembang, Indonesia using two Landsat TM images from 1986 and 1992 and a Landsat ETM image from 2001. It was found that 289020ha of land in the subsetted image had undergone a drastic reduction in NDVI values, implying areas of deforestation.

 

Sam Price - A proof-of-concept forest stand type map for Armenia from Landsat TM imagery

Multi-spectral satellite remote sensing imagery of north-central Armenia from the LANDSAT TM and ETM+ sensors was classified into forest cover types using September 2007 reflective bands and a time-series of Normalized Difference Vegetation Indices (NDVI) from August 2000, October 2006 and September 2007. Plot-level forest inventory data, collected by a team of researchers over the summer of 2007, were randomly assigned to function as either components of forest cover type training regions for supervised classification or as validation regions to test the accuracy of the products of the classifications. A supervised maximum likelihood algorithm applied to the NDVI time-series was found to be the most accurate method of classification. Randomly selected validation plots (n=56) indicated an overall classification accuracy of 30%. Results showed that Pinus-dominated plots were easily differentiated from deciduous plots. Differentiation of deciduous species was more difficult, with 0% user’s accuracy for the Carpinus, Acer and Mix classes. The classification achieved 44% accuracy for Quercus and Fagus classes despite their relatively low dominance of the training set.

 

Carlos Chiriboga - Identification of archaeological site locations in the northwestern area of the Petén, Guatemala

The use of remote sensing has a long tradition in Archaeology of the Middle East, Europe and North America. In contrast, Maya archaeology has, up until recently, remained somewhat distanced from the uses of remote sensing due to the dense forest vegetation characteristic of the lowland regions of northern Guatemala and southeastern Mexico. Recent applications of remote sensing have changed this scenario due to new applications utilizing data that has become increasingly easier to obtain. The present report will analyze these new methods for archaeological site location used in the region and try to evaluate their applicability in a broader, regional perspective.

 

Will Gardner - Remote sensing survey of prehistoric land cultivation in the Yampa and White River drainage basins

The Yampa and White River drainage basin of northwest Colorado is a vast area consisting of approximately 2.6 million acres of mixed public, private, state, and federal land. From approximately 2100 to 400 B.P. the area was also home to the formative post-Archaic Native American group, known as the Fremont. The Fremont were essentially a horticultural Native American group that relied heavily of wild plant resources, cultivating a variety of plants including Zea mays (maize), and hunting. To date, no distinct archaeological record of landscapes utilized in maize agriculture has been discovered in Northwest Colorado. This lack of information in the archaeological record is due to the dynamic nature of a landscape affected by erosive processes and the use of a low impact form of cultivation which leaves a light footprint. The goal of this project is to formulate a process to identify micro-ecologies that have a higher probability of having once facilitated maize cultivation by Fremont Native Americans. The identification of these micro-ecologies will be based on the analysis of Normalized Difference Vegetation Index (NDVI) signature of the Yampa and White River drainage basin of northwest Colorado. Analysis of NDVI signature was chosen because of the ability of the imagery to function as a proxy for general environmental conditions. Because NDVI signature relates information about the type, total, and health of vegetation communities of an area we can safely extrapolate out certain assumptions about the general environmental conditions present. This holds especially true in arid conditions, such as the project area, were the presence and amount of vegetation is extremely dependent on the availability of water.

 

Chris Milan - Examine regional differences within the Lurin Valley, Peru

The central coast of Peru consists of a series of river valleys that run from the Andes Mountains. Between the Andes and the Pacific Ocean, the landscape is separated into a series of climates and microclimates primarily based on elevation. This project tests the variability of climates along the coast as well as their relation to the morphological features of two river valleys, the Rimac and Lurin. Using a combination of MODIS and ASTER imagery, coastal features relating to human subsistence were identified.

 

Madeline Blount - Climate change in the Tibetan Plateau

Reports of climate change in the Tibetan Plateau region since the 1950’s describe many different phenomena: extreme seasonal temperatures, desertification, melting glaciers, drying rivers, and shrinking lakes. This study focused on Lake Qinghai (37° E, 100° N) and its surrounding environs. This largest lake in China has been reportedly shrinking by a rate of 0.1m per year, and the change has been tied to land use practices around the lake. To experiment with satellite imagery as a potential tool for classifying this area, two MODIS 500m spatial resolution daily scenes (August 1st, 2001 and August 2nd, 2007) were downloaded, subset, georeferenced, and subjected to unsupervised and supervised classification. These classifications were not strong enough to show climate or land use change at the macro level. An NDVI composite image also showed relatively small amounts of vegetation change in the region, and very little change around Lake Qinghai. DEM data and imagery further clarified the topography of the region and the particular distribution of vegetation, but classification and climate change in this study proved difficult, possibly due to the resolution of the imagery. The 500m spatial resolution was not commensurate with change in the Tibetan Plateau on a micro level in the past six years. Further study of change in this particular area will require finer resolution or data from a longer period of time. While quantitative analysis and successful land classifications were limited, exploration of the satellite imagery made clear its role as a powerful tool in gaining familiarity with a remote region before further study.

 

Kyle Meister - Detecting cover change and areas of vegetated refugia in the border region between Brazil and Paraguay

Exploitation of the South American Atlantic Coastal Forest is happening at an alarmingly fast rate, creating fragmented landscape ecosystems. In order to address conservation issues, it is necessary to identify where changes have occurred most intensely on the landscape. Different countries have different policies and levels of enforcement regarding the exploitation of natural resources. Indeed, when one country promotes forest and watershed conservation in a region, another may promote development. Or one country may value land in a different way. For example, in the case of Paraguay and Brazil, Paraguay classifies the same floodplain of the Paraná River as fertile land, while Brazil considers it poor for agriculture. This is despite the fact that both sides of the river have similar soils, mainly ultisols and oxisols (Selvaradjou et al. 2005). These differences in on-the-ground classification could lead both countries to develop the same region in different ways. Where have changes in cover and land use occurred in the border region between Brazil and Paraguay? Where has cover not changed so drastically and are there possibilities of restoring forest cover to connect isolated forest regions within and across borders? In this project, it was hoped that remote sensing could be used as a tool to detect changes in land cover and identify common forested areas between images taken from two different years.

 

Yong Zhao - Land use change of Lhalu Wetland National Nature Reserve from 1980’s to 2006

In the past 30 years since the adoption of reform and opening-up policy, China had been rapidly developing with profoundly and widely changes in material and spiritual life. Lhalu Wetland National Nature Reserve (91”03’N/29”39’E) , a typical alpine meadow and marsh wetland, located in the northwest of Lhasa, Tibet, China, also had been change dramatically by the arbitrary city planning projects. As the highest urban wetland in the world, it is 3645 meter from sea level on average and 6.2 square kilometers, 11.7% area of Lhasa city, playing a critical role in the sustainable development of Lhasa City, which has been living in harmony with nature for more than one thousand years. In this paper, the effects of several significant city planning projects and policy relating to the land use change and protection of the wetland are addressed by assessing the wetland area and quality change using Landsat MSS, TM, ETM and Aster satellite image from 1976, 1991, 2000, 2005.

 

Sara Enders - Changes in vegetation and hydrology of subwatersheds of the Mississippi River Basin

A bimonthly time series of Normalized Difference Vegetation Index (NDVI) reveals increased mean vegetative productivity in two nested watersheds in southeast Indiana between 1982 and 2003. The relationship between NDVI and an index for the balance between precipitation and discharge (P-D)n is explored in an effort to explain increases in precipitation not matched by increases in discharge. The potential of differences in land use to account for differing trends in NDVI and (P-D)n is also explored.

 

Noel Aloysius - Land cover classification based on enhanced vegetation index time series for the Northern Great Plains

The use of Enhanced Vegetation Index to classify land cover types is demonstrated. A phenology based land cover for the Northern Great Plains is presented. An 8-year time series Enhanced Vegetation Index data is used to develop 17 land cover classes. Out of the 17, eight classes are identified as cultivated crops. The classification is largely in agreement with the National Land Cover Classification of 2001, but further helps to identify additional land cover types within the cultivated crops. The accumulation of green biomass during the growing season and its spatial and temporal variation are also presented. This analysis highlights that the regions showing high accumulation of green biomass also exhibits high variation temporally.

 

Mercedes Bravo - Drought/precipitation pattern assessment in the Gila Wilderness Area of New Mexico

The goal of this project is to examine vegetation and landscape response to recent weather patterns in the Gila National Forest of New Mexico, a remote and undeveloped forest in the southwestern U.S. Although parts of the Forest have recently received substantial rainfall, like much of the western United States, the southwestern region of New Mexico has been experiencing drought conditions, particularly over the last decade. Four MODIS-Terra images were obtained in total, at two different resolutions (250m and 1000m) and on two separate dates (June 2002 and June 2004). Analysis focused on the MODIS 16-day composite Normalized Difference Vegetation Index (NDVI) product. The NDVI is a simple, well-known, and reasonable proxy for evaluating on-the-ground conditions through change in vegetation type and/or robustness. Elevation and land cover classification data were obtained from the U.S. Geological Survey (USGS). Long-term normals and annual averages for precipitation, temperature, and streamflow data were obtained for weather and stream discharge stations proximate to the Gila National Forest. The percent change in NDVI between 2002 and 2004 was positive for all land classes, indicating that vegetation fared better in 2004 than 2002. Precipitation records showed that 2002 was a very dry year, while 2004 was anomalously wet, particularly early in the year. A transect generated to evaluate the relationship between change and elevation demonstrated that below 2000m a greater NDVI value was consistently recorded in 2004 than in 2002. The combination of satellite imagery and on-the-ground weather data supported the assessment of the immediate and longer-term effects of temperature and precipitation patterns on the natural resources of the Gila National Forest.

 

Abby Fraeman - Mineral identification in the Nili Fossae Trough region of Mars

The Nili Fossae Trough region on Mars is at the top of the list of perspective landing sites for the 2009 Mars Science Laboratory (MSL) rover. The site is considered interesting due to the detection of mafic minerals and hydrated phyllosilicates in the area. These minerals are formed under conditions very different from present day Mars, and their presence could reveal a wealth of information about the planet’s past. The purpose of this project was to examine data from the CRISM hyperspectral imager to confirm the presence of these minerals in Nili Fossae Trough. The first attempt to find these minerals was an unsupervised classification of low-resolution CRISM images. The results of procedure highlighted the difficulty in performing an unsupervised classification and interpreting the results, and it was unsuccessful at unambiguously detecting the phyllosilicates and mafic minerals. The second method used to find these minerals was the use of summary products. These are similar to indices whose values indicate the presence or absence of certain minerals. These summary products, which are defined in the scientific literature, have been proven to be fairly robust at extracting information about specific minerals by the CRISM team. Two high-resolution images were analyzed using these summary parameters, and the analysis confirmed the previously reported detection of mafic minerals and phyllosilicates in the area. The results of this analysis were further used to determine the number of pixels containing positive indicators, although they could not be used to determine the relative abundance of minerals due to the complex relationship between the strength of the summary parameter value and abundance.

 

Maung Moe Myint - Analyze spectral changes during leaf desiccation

In this study, the spectral reflectance data of Norway maple leaves were measured in the laboratory with a spectrometer. Weight of leave was also measured using the precision scale. The measurements were carried out until leaves were dried naturally. Ratio based leaf water content indices were derived. Relationship with leaf spectral reflectance with relative leaf water content; relationship with indices and relative water content were also derived. We found that there is apparent relationship between indices and relative water content. Then the indices were applied to the Moderate-Resolution Imagine Spectroradiometer (MODIS) image to observe and visualize the various level of drought in the land cover.

 

Ryan Carney - Using remotely-sensed satellite data to predict the Peridomestic West Nile Virus avian host populations in Sacramento County, California

West Nile virus (WNV) is a mosquito-borne flavivirus that is maintained exclusively in wild avian populations, most importantly the peridomestic American Crow and American Robin. Elucidation of the key environmental determinants of the distribution and abundance of these two species would greatly contribute to the prediction and assessment of the ecological and epidemiologic impacts of WNV on both bird and human populations. As a first step toward this goal, four remotely-sensed satellite datasets were obtained from the Sacramento County, California area for exploratory data analyses. These included Shuttle Radar Topography Mission (SRTM), National Elevation Data (NED), and LANDFIRE datasets, as well as Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) data acquired at a time (April 30, 2005) that immediate preceded the occurrence in this region of the most severe WNV epidemic/epizootic documented anywhere in the world for that year. The satellite imagery was processed in ER Mapper and ENVI to produce classifications based on temperature, Normalized Difference Vegetation Index (NDVI), and vegetation height, and then subsequently analyzed and integrated into a geographic information systems (GIS) framework for comparison with avian abundance data. Ultimately, the resolution and quality of some of the satellite data was found to be insufficient. However, the ASTER/NDVI dataset proved optimal and was useful for small-scale discrimination of urban/suburban vegetation. The author recommends that future work be focused on more in-depth analyses of the ASTER/NDVI dataset and its statistical correlations with the distributions of the avian hosts of WNV.

 

David Davis-Boozer - Using remote sensing to model robin distribution in Connecticut

The natural reservoir for West Nile Virus (WNV) is in songbirds. Bird species vary in their competence for transmitting the virus to uninfected mosquitoes, but the American robin is known to be highly competent. Culex pipiens, the primary vector for WNV in Connecticut, has been shown to feed almost exclusively on birds, demonstrating a preference for robins over other songbirds (Molaei et al. 468-474; Ngo and Kramer 215-222). This fact combined with robins’ high competence as a host makes them a crucial element in the amplification of West Nile Virus in suburban and urban environments.

With the ultimate goal of developing a predictive model of robin abundance across the southwestern portion of Connecticut, my final project sought to develop a means of describing an urbanization gradient in terms relevant to suitable robin habitat. Our model will require a finer understanding of land use within the NLCD’s Commercial land cover class. In exploration of this problem, I ran four variations of an ISOCLASS unsupervised classification on ASTER VNIR images of Central Connecticut in hopes of defining new classes within this broad land cover class.

 

Keita Ebisu - Association between residential landuse and asthma severity

In this project, agreement between supervised and unsupervised classification was studied. Using these classified data, whether specific land use affects asthma severity was investigated using several statistical models. As a result, the agreement between supervised and unsupervised classification is 38.0%. From the statistical model, grass land future protects from asthma severity, while utility/factory land use affect badly on asthma severity based on supervised classification result. Though unsupervised classification result did not reach to statistical significant level on these variables, magnitude of odds ratios are comparable to supervised classification result.

 

Nat Wilson - Using Landsat imagery to characterise snow metamorphosis over alpine topography

Using a pair of Landsat 7 images created over Kluane National Park and Reserve in the Yukon Territory, algorithms have been applied to determine temperature and grain size characteristics of snow in summer and winter. A mask was created to identify pixels representing snow-covered ground, after which the properties derived are snow surface temperature and qualitative differences in grain-size using a proxy for snow albedo. Band saturation poses a significant source of information loss in using ETM+ images over snow surfaces, and so time was spent working to address the problem. Data thus derived suggests that air temperature plays a larger role in snow metamorphosis than direct solar radiation. Finally, an algorithm was developed to distinguish snow from ice, which makes it possible to separate glaciers from other landscape covers, and to separate treatment of ice and snow.

 

Andrew Delman - Sea surface temperatures and precipitation patterns in the Galápagos Islands area, Pacific Ocean

The Galápagos Islands of Ecuador are located in the equatorial Pacific, in the zone where the cold Humboldt current that parallels the coast of Peru leaves the continent and heads west to meet the cool upwelling zone of the equatorial Pacific countercurrent. The convergence of these two currents gives this region a unique climate, which almost no other island groups in the world share. Unlike the typical wet climate that characterizes most equatorial regions, the Galápagos Islands have very arid lowlands, with more precipitation at higher elevations. The difference in climate between elevations is most evident during the so-called “dry” season from June-December, when the Humboldt and equatorial countercurrents dominate, the cold water inhibits convective precipitation and rain rarely falls near sea level. However, because the air is still very humid from nearby warm currents, low stratus clouds tend to form over the ocean, and when these clouds come into contact with higher terrain, a light misty precipitation over the highlands results, known locally as “garúa”.

The head of the biological station in the highlands where I stayed explained that rainy weather tends to precede the full moon. At first it seems unlikely that a significant connection could exist between the lunar cycle and weather patterns, but there are studies that have suggested a link between tidal cycles and sea surface temperatures. Keeling and Whorf have hypothesized a possible link between extreme tides and global sea surface temperatures, and they in turn refer to a study done in the seas surrounding the Indonesian archipelago. This study analyzed sea surface temperatures measured in situ at several different locations as well as by the AVHRR satellite, and picked up strong periodic signatures at periods of 14-16 days and 27-30 days at most of the measuring sites (Ffield and Gordon 1996).