OEFS Abstracts - Spring 2014

Celine Lim - Change of Peat Swamp Forests and Palm Oil in Kalimantan

Indonesia has one of the highest deforestation rates in the world, led primarily by the conversion of forest for the expansion of oil palm plantations. This has occurred across many states, including Kalimantan, which houses high conservation value forest and critical habitat for the endangered orangutan. This research sought to analyze the change in land use cover change to oil palm and to map peat swamp forest in an area in Central Kalimantan, using basic remote sensing techniques. Landsat TM 7 and 5 scenes from 2003 and 2010 were used for a change detection analysis across the 7 years. The Tasseled Cap Transformation and elevation data were used to detect peat forest. The results showed a significant expansion of oil palm plantations, with high forest cover loss. Deforestation for the oil palm commodity is occurring at a rapid rate in Kalimantan, and remote sensing can serve as a key tool to monitor the dynamic land use changes in the region.

 

Kristin Dreiling - Rocky Mountain Pine Beetle: Small Insects Making a Big Impact Colorado Forests

The Rocky Mountain pine beetle (Dendroctonus ponderosae) (a.k.a. mountain pine beetle, black hills beetle, bark beetle) has helped maintain the health of Rocky Mountain forests in North America, from Canada to Mexico, for thousands of years. Since the mid-1990s, this region has experienced an unprecedented epidemic as Rocky Mountain pine beetle outbreaks have decimated millions of acres of forest. Warmer winters, regional drought, and a change in stand dynamics have created a ‘perfect storm’ of ecological conditions within which the beetles are thriving (U.S. Forest Service 2014). This report will examine the epidemic by employing remote sensing techniques to quantify land cover change and vegetation health along the mountains of northern Colorado. Landsat satellite images will be used to measure the degree of change between images from 1996 and 2010.

 

Sarah Tolbert - Measure Forest Changes Over Time in the Burhiniyi Community Forest

The Itombwe Nature Reserve (INR), located in the Eastern Democratic Republic of the Congo (DRC), is home to a host of endemic wildlife, most notably the eastern lowland gorilla. To preserve this unique ecosystem, international non-governmental organizations successful pushed the government of the DRC to designate the INR as a protected area by law in 2006. Now local conservation organizations suspect that the protected status might not be the solution after all as continued conflict and pressures on the land continue to drive deforestation in the INR.

However, next to little research has been conducted in this portion of the DRC and before a move is made to change protected area policies, it is worthwhile to explore what change is actually occurring. Although researchers use Landsat Thematic Mapper (TM) to map changes in forest cover in the DRC at the country level, regional studies are lacking. The objectives of this study were to examine if free Landsat TM images could be used to classify land cover change on a regional level, and to explore how to best separate primary, secondary, and mixed forests through various classification techniques. Two images were selected, April 2nd, 1990 and April 8th, 2014 from row 174, path 63, LandsatTM 4 and 8 respectively. Unsupervised NDVI classification proved the most useful to tease out the different forest types, although change detection from supervised NDVI images were useful to better understand the general change in forest cover in the area. Ground truthing to develop more accurate ROI’s would greatly enhance the ability of remote sensing to detect local land use change in this area and develop a method to distinguish different types of forest cover.

 

Wan Nurul Naszeerah - Eco-epidemiology of malaria at Bhutan-India Border

Malaria, which is mostly caused by Plasmodium vivax in Bhutan, is a vector-borne disease involving certain Anopheles species. Female Anopheline mosquitoes acquire the parasite Plasmodium when they feed on an infected human. Transmission occurs when these infected mosquitoes subsequently feed on uninfected humans. In 2009 Bhutan experienced a malaria outbreak, with the highest number of incidence in Sarpang district. Although the loss of insecticidal efficacy has been suspected, various environmental factors, including elevation and temperature, made the region highly susceptible to the outbreak. Therefore, remotely-sensed changes in four Landsat 5 TM datasets between January 2007 and March 2009 are assessed to test the following hypothesis: a shift in agricultural and environmental conditions prior to March 2009 provided a ‘window of opportunity’ for malaria vector population expansion or increased contact rate between uninfected malaria vectors and infected human hosts. A supervised land cover classification suggested an increase in agricultural vegetation in both Sarpang, Bhutan and its adjacent Indian neighbor, Assam state. During dry season of late 2008, agricultural areas in Assam experienced dramatic increase in “wet soil” class compared to during the dry season of late 2007. At the inception of the outbreak, however, vegetation in India and Bhutan were replaced by “wet soil” and “water” classes. The classification results correlated with comparative NDVI but only partially correlated with the wetness index of tasseled cap transformation. A higher spatial resolution image and an improved NDWI index may be able to improve the eco-epidemiological investigation.

 

Carlin Rosengarten - Observing Land Cover Transformations in Vietnam’s Coffee-Growing Highlands

Abstract Unavailable

 

Colin Brown - Change Detection and Land Cover of Almond Orchards in California’s San Joaquin Valley

Abstract Unavailable

 

Gabriel Chait - Shifting Agriculture and Forest Land Uses in the Southern Yucatan Peninsula

This study analyzes shifting agriculture and forest land uses in the southern Yucatan peninsula. It was conducted as part of an ongoing project in coordination with The Nature Conservancy – Merida, Mexico. The focus of the study is to assess improved logging and agricultural techniques to promote sustainable land use and reduced impacts. Regeneration methods for commercially valuable long lived species like mahogany (Swietenia macrophylla) were analyzed. Background research suggests that large disturbances mimicking natural fires and hurricane events are necessary for sustained regeneration of mahogany forests. In this case large shifting agricultural plots can be an effective method of sustainably using land while allowing for natural regeneration to occur after a few years of cultivation. This would allow for mahogany regeneration, contributing to the commercial viability of the forest, and its conservation as forest land. Composite Landsat ETM+ images were used to classify and calculate change between land use in the region between 2000 and 2012. A site visit earlier in the year was used to observe land use and gather qualitative information which was applied to the classification analysis. Results show a loss of forest lands with increases in agricultural plots as well as semiurban areas. This creates a scenario of increased pressure on the forest as well as on opportunity for concerted regeneration efforts.

 

Julia Anderson - Detecting Greenhouse Agricultural Expansion in China using Landsat Imagery

The practice of plasticulture, in which plastic covering is used to protect crops and intensify agricultural production, is increasing on a global scale. Kunming, the capital of Yunnan Province, is well known for its application of plasticulture and serves as one of the major agricultural regions in China. Despite a significant increase in greenhouse land cover, few remote sensing projects have attempted to isolate greenhouses in order to quantify this type of land use change.

This study uses Landsat 8 imagery to isolate pixels containing greenhouses surrounding Lake Dianchi in Yunnan Province. Results indicate that it is possible to identify greenhouses, especially in urban regions, using Landsat imagery. This project is an initial attempt to establish a method for classifying greenhouses so that future research might be conducted to quantify plasticulture land use over time.

 

Sophie Young - Intensification of Cotton Cultivation in Xinjiang Province, China since 1990

China surpassed the United States by 2008 as the top global exporter and importer of cotton, the majority of which is produced in the northwestern province of Xinjiang Uyghur Autonomous Region. Within Xinjiang Province, the arid Aksu-Tarim Region of the Tarim River Basin has been a major base of cotton production since the 1950s, and cotton development entered a second period of intensification between 1989 and 2011. Recent studies focus on ecological degradation that resulted from intensive agricultural development – irrigation and mechanization - which include degradation of riparian vegetation, soil moisture, and groundwater tables. This study focuses on a subset of the region of Bayangol Mongol Autonomous Prefecture in Xinjiang Province, where agricultural land reclamation and employment in the agriculture sector doubled between 1989 and 2011. Land cover change analysis and comparative normalized difference vegetation index (NDVI) are used in this study to detect change in agricultural land cover as a result of China’s cotton production policies, and to lay the groundwork for future studies in this region on urban development, desertification, riparian vegetation, and water diversion.

 

Alexander Shiarella - Tracking Urban Expansion and Change in Warsaw, Poland

In the quarter century since the Polish Round Table agreement and Poland’s subsequent transition into a free market economy, Warsaw has become an important center of political and economic power in Europe. The corresponding developments in Warsaw’s cityscape occur on an already varied landscape shaped by its long and turbulent history. Despite the opportunities the city therefore provides with respect to satellite imagery analysis, extensive study of Warsaw remains largely absent in the remote sensing literature. This study is aimed to collect and analyze information on urban change in Warsaw’s post-socialist period using Landsat imagery data. Preliminary results show trends of large-scale growth in regional development, decreases in agriculture, and the effective maintenance of forests near Warsaw’s periphery. Hopefully, these finding encourage further investigation into urban expansion in the greater Warsaw metropolitan region and into characterizing urban change in postsocialist states using satellite imagery. Background

 

Avishesh Neupane - Urbanization and Land Use Change in Kathmandu Valley

Kathmandu valley, Nepal is rapidly urbanizing. This study primarily uses Landsat 7 and Landsat 8 images of Kathmandu valley, Nepal to examine land cover change in the valley between 1999 and 2013. The objective of this study is to characterize the current land cover of Kathmandu valley and quantify land cover change over the past 14 years. Analysis showed that between 1999 and 2013 urban and sub urban areas in the valley have increased at the expense of agriculture and open land. Forest cover has slightly increased while areas with water have declined. Although changes are happening all over the valley – higher changes are visible along road networks.

 

Ijeamaka Anyene - Evaluating Urban Expansion and Land Cover Change in Nairobi, Kenya

This study aims to (1) quantify Nairobi’s land cover change and (ii) quantify urban expansion in Nairobi, Kenya. This is in order to evaluate urbanization in Nairobi. The images used for this project are Landsat TM 5 images from the dates of July 3rd, 1987 and August 19th, 2010. To examine land-cover change, a minimum distance supervised classification and an unsupervised classification (K-Means and IsoData) on a masked region were used. Then a change detection analysis was conducted. Overall, all three types of classification produced the results that urban areas increased and agricultural areas decreased in Nairobi. The change detection analysis shows that urban areas, especially urban less dense, became more widespread. Most of the agricultural lands became developed into urban less dense areas.

 

Peter Hirsch - Informal Settlement Mapping of Mumbai, India

An estimated one billion people reside in informal settlements globally. These informal communities are defined by digressions from established urban forms, where the infrastructure is autonomously assembled to yield hyper-dense, programmatically specific environments. With global population set to increase drastically over the coming decades, informal settlements will become pervasive as an urban land typology. Understanding the location and growth of these communities will be vital to understand the energy and material demands of urban centers and to plan for the growing challenges created by climate change.

This analysis seeks to locate informal settlements within Mumbai, India, home to one of the largest informal settlements in the world. Using Landsat 8 and VIIRS satellite imagery, informal settlement locations are determined by utilizing the communities’ defining characteristics to set them apart within urban centers.

 

Vivienne Zhang - Measuring the Accelerated Population Growth and Urbanization of the Tibet Autonomous Region

Rapid economic development has spurred significant land use change in China since the Chinese government initiated its economic reform in 1978. While the land use dynamics in many parts of China has been well-documented and studied, because of the political sensitivity surrounding the Tibetan Plateau, less attention has been paid to understanding the change in the Tibet Autonomous Region. The objective of this study was thus to use remote sensing to understand the land use dynamics in Lhasa, as an entry point to understand the broader trending happening in Tibet. Two Landsat TM images were obtained over Lhasa, Tibet (path 138, row 39). One was from February 2001 and the other from February 2014. Initially, my goal was to study the urbanization trend in relation to elevation in Lhasa. However, because the mountainous areas surrounding the Lhasa valley had a wide range of spectral signatures due to the varying angles of the slopes and degrees of shadows, attempts of supervised classifications failed to distinguish the hills from different land covers of urban settlements. Incorporating thermal infra-red data was ineffective to help classify the higher elevation area that was expected to have lower temperatures, because many of the south-facing slopes were warmer due to high levels of insolation. The images were then displayed on Digital Elevation Model, which shows that most land cover change has been taking place in lower elevation. The mountainous regions were subsequently masked out using the elevation data. Supervised classifications using the spectral data and Tasseled-Cap transformed data were performed on both images. Change analysis shows two major trends from 2001 to 2014:

  1. suburban areas increased by 47% and had become one of the largest land use classes
  2. greenhouses, although being pushed further to the city’s boundaries, had almost tripled in size.

Major uncertainties involved the classification of suburban classes, as the density of a suburban area is difficult to define. Future work should involve a more quantitative definition for suburban classes. Images of higher resolution such as ASTER might also be used. To ensure that most urban development is limited to lower elevation, spectral angle mapping will be helpful to study the urban areas in relation to elevation.

 

Catherine Kuhn - Climate Change Vulnerability in the Powder River Basin, WY

The goal of this project is to characterize vegetation and snow response to drought in a subwatershed of the Powder River Basin, WY using Landsat TM and Landsat OLI swath products. NOAA climate records were used to identify both wet and dry dates. Spatial image analysis was conducted on images from the wet and dry summers in order to derive drought response in vegetation, land surface temperature and snow cover area. Land surface temperatures were ground confirmed using data from SNOTEL snow monitoring sites. These variables were used to create a simplified drought index (TVX). The images were then classified using multispectral bands and land surface temperature. Results from this research show that Landsat recorded vegetation response to climate change does not correlate to the drought monitor estimates from the NOAA climate records. This study can be used to identify areas most at risk of water stress during drought years and prioritize resiliency interventions for water resource managers.

 

Joseph Calamia - Determining the Growth of the Jharia Coalfield Fire

Using three images from 1989, 2001, and 2014—acquired by Landsat 5 TM, Landsat 7 ETM+, and Landsat 8 OLI & TIRS, respectively—this project determines the spread of a fire, first detected in 1916, in the Jharia coal field in Jharkhand, India. The study evaluates brightness temperature readings in Kelvin. These are first calculated manually from each satellite’s respective thermal band—band 6 (10.40-12.50 micrometers) for Landsat 5 TM and Landsat 7 ETM+, and band 10 (10.60-11.19 micrometers) for Landsat 8 OLI and TIRS—and then verified using ENVI’s built-in Radiometric Calibration tool. Using this brightness temperature data, the project then explores two different procedures for determining change of the fire extent.

The first uses analysis of a layer stack of the three years’ brightness temperature images and then normalized brightness temperature values to determine areas of temperature change, and thus potential fire areas. The second technique requires masking out all but the hottest areas in scene by setting a temperature threshold for each image and eliminating all data representing areas cooler than potential fire spots. An unsupervised ISODATA classification of these fire spot images helps to determine the area in square kilometers of those spots and allows the calculation of change statistics.

The project concludes that change detection is possible from remote sensing techniques and that the fire does appear to be growing in expected areas. Further analysis is required using more images for interstitial periods, a classification of landcover to improve temperature values (given varying emissivity of landcover types), and the overlay of a geology map to help determine fire spread and depth.

 

Marjorie Hirs - Oil Slick Monitoring

Abstract Unavailable

 

Jamie O’Connell - Can salt marsh community vegetation be identified based on spectral signatures?

Salt marshes are valuable coastal ecosystems that are at risk of drowning due to accelerated sea level rise caused by climate change. The landward migration of salt marsh vegetation is an understudied, yet potentially critical, survival mechanism for salt marshes to avoid habitat loss from rising tides.

The goal of this project was to use remote sensing of aerial photographs to:

  1. quantify the amount salt marsh vegetation that migrated into lawn and forested uplands between 1974 and 2010 at Hammonassett State Park in Madison, CT,
      and
  2. identify bands, band ratios, and band indices best suited for supervised classification of salt marsh and upland land cover types.

Due to limitations in identification of training regions for the 2010 and 1974 images, it was not possible to classify salt marsh migration in the course of this project. However, supervised classification of a 1 m pixel resolution aerial photograph from 2012 found that a 7 band layer stack of infrared, red, green, blue bands along with a ratio of the green to blue bands and NDVI successfully classified most regions of the lawn and upland. The primary flaw was that the northwest region classified as mudflat was too extensive. The true land cover class for most of that region was likely wetter or sparsely vegetated S. alterniflora.

 

Jill Kelly - Modifiable Areal Unit Problem in Remote Sensing Pixel Resolution

Abstract Unavailable

 

Parker Liautaud - Change Detection Study of Antarctic Ice Shelf Collapse

Abstract Unavailable

 

Colby Tucker - Effect of Drought on Soil Moisture and NDVI in Relation to Water Origin in Northern Ca

Spring-fed rivers are known refugia for cold-water fish. These rivers maintain their temperature and water depth throughout the year and across years with varying amounts of precipitation in comparison to snow-melt dominated rivers. This study compares soil moisture and vegetation cover between spring-fed and snow-melt dominated watershed using remote sensing techniques. Four Landsat TM/OLI images of Northern California from four different years (1986, 1993, 2003, and 2013) were manipulated through Tasseled Cap transformations and NDVI. The wetness band statistics from each watershed were calculated and compared with precipitation data. The results suggest that soil moisture in spring-fed watersheds and snow-melt dominated watersheds respond differently to the decreasing precipitation observed in Northern California.

 

Julie Carson - Flood Impact in Sept 2013 in Boulder & Estes Park Region

With the ultimate goal of quantifying flood extent in Central Colorado after a historic rain event in September 2013, this project largely focused on cloud and cloud shadow removal. Clouds were removed successfully by thresholding the Cirrus band on a Landsat 8 image from July 8, 2013. Cloud shadows were removed successfully by first refining the areas of interest with an unsupervised maximum likelihood ISODATA classification, then performing multiple iterations of a maximum likelihood supervised classification to distinguish cloud shadows from other scene features. The cleaned images were most successfully classified for flood extent using an ISODATA classification, but careful employment of the Tasseled Cap Transformation on cloud free images (unavailable for this project) and calibrated for use with Landsat 8 images is expected to reveal similar results.

Using the original Region of Interest, defined as the overlap between Landsat 8 Path 34, Row 32, and Landsat 8 Path 33, Row 32, the unsupervised ISODATA classification revealed that approximately 61.65 km2 of the study area was flooded by the September storm event. Subsetting the original scene further to refine the area of interest, then performing an unsupervised classification, revealed that approximately 147.93 km2 of the area was flooded. While other techniques, like LiDAR, have been employed to better assess areas where floodwaters are hidden by vegetation or other structures, this project’s method of assessing Landsat 8 images was fairly successful in providing preliminary results. Given that Landsat 8 images are free and have good spatial, spectral, and temporal resolutions, they are an excellent starting point for this type of analysis. Still, any cloud free images would be the best basis for analysis.

 

Simon Gore - Flood Plain Classification for Hydrology Friction Modeling

This project utilized passive remote sensing techniques to analyze patterns of wetland loss near the Mississippi River Gulf Outlet (MRGO). A number of image classification and transformation processes were performed on two images, from 1984 and 2014, to detect quantitative and visual patterns of change. Ultimately, the tassel cap transformation proved the most effective image processing technique providing class statistics illustrating the conversion of wetlands to open water. The eventual goal of this research is to derive an accurate assessment of changing surface friction parameters in the region to better understand the ecosystem services provided by regional wetland types.

 

Uma Bhandaram - Urbanization and Implications on Wetland Areas in Hyderabad, Andhra Pradesh, India

There has been rapid urbanization in Hyderabad, India over the past two decades and it is only expected to increase. This growth, however, has been unchecked and its ecological implications have not been researched extensively. This report looks specifically at the change in urban, lake, and wetland areas from 2000-2009 to see the impacts of urbanization on water resources. A Landsat 7 ETM image from December 2000 and a Landsat 5 image from January 2009 of Hyderabad, India in the winter season were compared. The two images were classified into seven classes using a supervised method. Comparing the classified images showed there was more urban, fewer lakes, and more wetlands in the 2009 image than the 2000 image. The increase in wetlands may be due to misclassification of roads and wetlands in the 2000 image because of the small-scale nature of these features. Because of the misclassification, the results may not be indicative of the impact of urbanization on lakes and wetlands. Further analyses with more robust classifications are recommended.

 

Yiyuan Qin - Exploring RS Applications in Assessing Water Quality in Belgrade Lakes Region, ME

Abstract Unavailable

 

Jan Kolmas- Seasonal Changes in Martian Polar Regions

Abstract Unavailable