OEFS Abstracts - Spring 2015
Simeon Max - Mapping Forest in SE Asia using Data from the Google Earth Engine
Extensive and consistent cloud cover severely impedes monitoring of forest change using remotely sensed imagery. In eminently cloudy areas of the world, such as the Southeast Asian tropics, concerning deforestation trends urgently call for accurate, easy, and broadly applicable methods to resolve the cloud problem in order to reliably inform policy and conservation. In this study, the Landsat Annual Greenest-Pixel TOA Reflectance Composite product was evaluated on its potential and constraints in monitoring forest cover change. The data are available on a global scale and free of charge, and can easily be processed through cloud computing or desktop software. The Landsat Greenest-Pixel product theoretically excludes atmospheric disturbances such as clouds by combining the pixels with highest NDVI values over the course of one year into one single image. While classification results turned out to be relatively accurate, this study shows that there is still noisy data present which impairs the validity of change detection statistics. In conclusion, the product was found to be useful when assessing forest change in locations where no cloud-free Landsat images are available and the product can be used to detect general inter-annual trends.
Maya Midzik - Factors Affecting Vegetative Growth in the Gunnison National Forest
Substantial shifts in the timing of snowmelt and snowmelt runoff toward earlier in the year have been correlated with changing climate in recent years. Understanding the potential response of vegetation growth to these variable conditions is fundamental to modeling the potential effects of climate change on high-elevation ecosystems. Phenological studies using small-scale field observations and common garden experiments have suggested that earlier spring melt and a longer snow-free growing season will significantly affect vegetation growth, resulting in greater overall production at high altitudes. This study employs remote sensing technology to examine large-scale ecosystem dynamics in response to changing climates. Analysis of MODIS NDVI time series over a range of topography and snowmelt conditions suggests that snowmelt date does not necessarily result in greater vegetation growth at high altitude, but rather significantly effects only the dynamics of early spring, having negligible effects on the overall productivity and dynamics of the growing season.
Alex Todorovic-Jones - Forest Mapping of Uttarakhand, India
The Hindu Kush Region is extremely important to the global stability of environmental processes and therefore studying the environmental changes in the Himalayas is critical in order to understand global environmental changes. In this study, Landsat TM images from a section of Central Himalaya were studied in order to look at land cover change between 1998 and 2011. Using corrections, unsupervised and supervised classifications, the images were analyzed for change detection. The main sources of change signaled a decrease in water cover, an increase in urban land, and an increase in forest land cover.
Katie Weber - Crops and Conflict: Irrigation Along the Euphrates River in Syria and Iraq
As droughts limit a shared resource of water such as the Euphrates River, international conflict may increase as one country’s efforts to store water limit availability to downstream users. Do these droughts affect agricultural yields along the Euphrates, and do upstream and downstream countries show differing levels of resiliency to drought? This study looked at three sites along the Euphrates River, in Turkey, Syria, and Iraq, in a drought year (1999) and a year with normal precipitation (2002). All sites were irrigated in part with Euphrates River water stored in nearby reservoirs. The height of the water in each of these reservoirs was remotely sensed by the HydroWeb data network, which could be combined with the surface area of the reservoirs to produce an estimate of volume. Reservoir volume data was considered alongside average cropland NDVI for each country in each year, to gain an understanding of the physical and behavioral response to drought. No differences in cropland NDVI were seen between or among countries, but the volume of each of the reservoirs was rapidly depleted during times of drought.
Eric Fein - Libya Desert Irrigation and Water Consumption
The Green the Desert agriculture project was commissioned by former Libyan dictator Muammar Gadhafi. Although the project serves no economic or agriculturally significant purpose it has been kept in place as a matter of national pride. The Libyan government believes the Green the Desert project demonstrates that Libya could be self-sufficient if it was absolutely necessary. However, in order to cultivate crops in the middle of the Sahara desert a massive amount of water is required. This has negative economic and environmental effects. The Libyan government has released statistics on the levels of water used; however, these figures are not universally accepted as accurate. In this study, the quantity of water consumed by the Green the Desert project was calculated by analyzing remote sensing data from Landsat-8 images. The moisture levels of the soil were calculated in order to estimate water use per unit area. This estimate was then multiplied by the total area of land cultivated in order to calculate an estimate of total water consumed. The estimate calculated by this study was then compared to the figure reported by the Libyan government. The Libyan government’s water consumption estimate is inaccurate. The Green the Desert agricultural project consumes between 2.5-3 times more water than the Libyan government claims. The actual level of water consumption could potentially have an impact on the price of water in Libya however, further econometric analysis would be required in order to produce a quantitative estimate of the effect.
Matheus Couto - Brazilian Soy Moratorium Impacts on Productive Area and Deforestation
Conversion of forest to soybean field was a significant driver of forest in Southeastern Amazon before 2008. The Soy Moratorium was the first zero deforestation agreement signed in 2008, stating the soybean fields would not occupy lands deforested after 2006. This study assessed land use change in two municipalities in the State of Mato Grosso, the largest soybean producing State in Brazil. The municipalities of Lucas do Rio Verde demonstrated older agricultural transition compared to Tapurah. Neither of the municipalities assessed presented the percentage of natural vegetation cover required by the Brazilian Forest legislation. The Soy Moratorium revealed efficiency in reducing conversion of forests to soy agriculture, but did not eradicate deforestation in both municipalities, nor promoted compliance to the forest code.
Phil Esterman - Using Satellite Night Lights to Monitor Urban Expansion in India
Satellite images of nighttime lights have been increasingly recognized as an objective, systematic, and spatially disaggregated source to study economic development. This study uses DMSP-OLS imagery of nighttime lights in India to study two questions in the aftermath neoliberal reforms passed in the 1990s: 1) did reforms lead to growth, and 2) did growth actually reduce poverty. We study poverty reduction using with a poverty index (population divided by night light radiance) using LandScan population data. Between 2004 and 2010, the analysis shows night lights increased in India by about 55.7%, corresponding to a 157% growth in GDP. It also shows that the poverty index decreased by 48%, corresponding to a 58% reduction in official poverty estimates. The analysis shows that, though growth concentrates in city centers, the poorest seem to share in its benefits.
Paige Breen - The Effects of Overgrazing and Attempts at Restoration in Chacabuco Valley
Overgrazing has become an environmental issue associated with the desertification of large areas of Patagonia in South America. All traditional ranching in the Valley Chacabuco was halted when Conservación Patagónica purchased the land in 2004, after centuries of degradation associated with overgrazing. This study uses remote sensing to evaluate the extent of recovery by comparing Landsat 5 TM and Landsat 8 OLI data from 2001 and 2014, respectively. Indicators of land quality used in this study include the Normalized Difference Vegetation Index (NDVI), a maximum likelihood supervised classification, and the Tasseled Cap Transformation.
David Gonzalez - Land-Use Change and Development in Madre de Dios, Peru
Many social and economic factors drive land–]use change in the Amazon. In order to effectively manage natural and cultural resources in the Amazon it is critical to understand how land–]use is changing, the patterns, pressures, and trends. The expansion of artisanal and small–]scale gold mining (ASGM) in the Peruvian Amazon, the majority of it illegal or informal, presents is a driving factor behind forest loss and degradation in the state of Madre de Dios in southeastern Peru. ASGM in Madre de Dios presents management challenge to organizations and officials concerned about the health of the forests and the human population, who are at risk of exposure to the mercury used and released in the gold mining process. This project uses remote sensing techniques to try to detect land–]use change due to gold mining in the Peruvian Amazon. Using three Landsat imagers spanning 1987 to 2011, I investigated small–]scale land–]use changes associated with gold mining in part of the Madre de Dios watershed. I compared vegetation loss through NDVI analysis images from each of the three years. I used supervised and unsupervised classifications to attempt to automate the process of identifying mining areas. I found that automated classification techniques were not able to differentiate between mining areas and water, as the open mines were inundated with water. Using NDVI and layer stacking was an effective technique to quickly visualize forest degradation and loss between the three time points. With further refinement, such as Gonzalez 2 finding a spectral signature for mines during specific times of the year and applying specialized professional tools, it may be possible to track the development of new mining areas on an ongoing basis as new Landsat images become available.
Vinh Lang - Change Within and Around Sinharaja Man Biosphere Reserve, Sri Lanka
This study sought to gather relevant information to my study area within the vicinity of the Sinharaja Biosphere Reserve. These methods include analysis of topography, temperature, and land use change in the surrounding landscape. The objective was to derive useful imagery relating to biophysical measurements of the region to aid in my summer analysis of forest structure and disturbance over environmental gradients in proximity to the Sinharaja Forest Reserve.
Understanding of topographic, moisture, vegetation, and temperature gradients across the landscape are paramount to the understanding of disturbances effects on forest structure. Analysis of the effect of the political boundaries of the reserve will also aid in my understanding of disturbance regimes (previous logging) in primary forest (within the reserve) and secondary forest (outside the reserve). Using freely available data provided by the USGS and NASA, the methodologies of data acquisition could potentially useful for others in the region given few studies have focused on this area with respect to remote sensing.
Nitsan Shakked - Monitoring Water Level Changes of the Aral Sea Between 1985 and 2014
The shrinking of the Aral Sea is an environmental catastrophe mainly caused by the increase in irrigation and the population growth in central Asian countries. As a result of the Sea’s regression, the salt deposit areas expand and the former seabed grows more and more exposed. In addition to the aridification of the area and the increase in dust storm potential, the regression causes the contamination of the new land covers with remnants of fertilizers and induces their high salinity. Frequent dust storms in the region transport such contaminants to nearby populated areas, negatively impacting public health and creating a growing concern. The goals and methods of this study are: monitoring the Aral Sea’s regression using Landsat imagery from 1988 and 2010; analyzing its former seabed formations by remote sensing techniques; and classifying and quantifying the hazardous land covers that it leaves behind employing the ENVI image analysis software.
Chris Bowman - Remote Sensing of Wetland Hydrology in South Florida
The wetlands of south Florida are an incredibly valuable and biologically diverse ecosystem, and recent efforts have been made to restore historic freshwater flow to the greater Everglades region. Previous studies have utilized the Tasseled Cap transformation to predict changes in wetland hydrology across south Florida. The primary objective of this study was to use Tasseled Cap brightness and wetness indices to detect changes in the hydrology of Big Cypress National Preserve between 1998 and 2011. It was found that these indices did not significantly correlate with Everglades Depth Estimation Network water depth measurements across the study region. Furthermore, the Tasseled Cap indices appeared to be highly sensitive to short-term precipitation patterns. The results of this study question the utility of the Tasseled Cap transformation in assessing wetland hydrologic change. Alternate avenues for future wetland hydrology research, both ground-based and remote sensing-based, are proposed.
Ross Bernet - Snowpack and Glaciation in California’s Eastern Sierras
2012 marked the beginning of California’s most severe drought in 1200 years. California farmers that are dependent on irrigation are one of the most dramatically affected populations. Reservoirs are filled from snowmelt originating in the Sierra Nevada Mountain range. Farmer water allocations are based on estimates of winter snowpack. The number of farms receiving no state water allocations has risen in recent years. For this reason, among others, the importance of understanding historical patterns of mountain snowpack is of great importance. This study uses two methods; (1) classification and (2)a snow index to quantify snow cover in California’s Eastern Sierra mountain range around Mono Lake.
Gabriela Baeza-Castaneda - Observing Drought Effects Adjacent to the Sacramento-San Joaquin River Delta
The continuing drought affecting the state of California has led to several environmental and physical changes in its landscape. One of the most prominent features in the region is the Sacramento-San Joaquin Delta located in the San Francisco Bay area, which provides much of the state’s water supply. This study’s goal is to observe the changes that have affected the delta as well as its surrounding landscape which vastly consists of urban, agricultural and marshland regions. Has there been a significant decrease in the region’s NDVI? Would this correspond with an increase in the soil coverage? I have used Landsat 4-5 images for the dry year and Landsat 8 for the wet year. The study involves using methods that include vegetation index, moisture index, burn index, unsupervised and supervised classifications.
Matt Goldklang - Using MODIS NDVI Timeseries to Study Drought
Since 2010, California has experienced an extreme drought. The lack of sufficient precipitation has severe negative consequences on ecosystem and vegetation health. This study uses MODIS time series data from 2000-2014 to determine the response of evergreen forests and temperate shrubland/grassland in Central California to precipitation and drought conditions through spatial NDVI standard deviation analyses, and a statistical comparison of NDVI and SPI for each vegetation type. SPI is the standard precipitation-based drought index. The results indicate potential drought resilience in evergreen forests, but a strong correlation (R2=0.72) between shrubland health (NDVI) and drought conditions.
Ben Hayes - Forest Drought Stress in Northwest Oregon
The ownership of America’s private forests has changed over the past 25 years, driven by a divestment of timberland by vertically integrated forest products companies and a surge in institutional investment. This has resulted in changing forest harvest patterns. These changes were observed and quantified for a study area in Northwest Oregon using Landsat imagery from 1990, 2002, and 2015. NDVI and image change analysis was used to identify areas of increased harvest and areas of regrowth. The study saw a general shift in harvest pressure away from existing mill infrastructure and towards the coastal area. This shift was strongly correlated with the sale of land near sawmills and a surge in log export markets closer to the coast.
XinXin Xu - Mesoproterozoic Geology of the Sinclair Group
Remote sensing to map lithological features has been widely used and studied. With the launch of the new ASTER satellite, remote sensing has become even more useful in creating geologic maps of hard-to-reach regions. In this project, a single ASTER image from the Sinclair region in Southern Namibia was classified using various techniques and compared to an actual geologic map that has been ground-truthed. To determine the accuracy of the remote sensing images, six major geologic formations that outcropped in the area were identified (Aubures, Barby, basement granite, Guperas, Kumbis, and the Kunjas). There were varying degrees of success with each of the methods used. Specifically, creating a composite image via three indices worked the best, and isolating and creating mineral specific images did not work at all. The images produced, though helpful, would have been quite difficult to interpret without comparing to the geologic map. Therefore, at least in this specific case, though remote sensing did provide more evidence attesting to the accuracy of the geologic map, it does not substitute for the actual mapping process.
Ethan Kyzivat - Remotely Sensing Soil Moisture: Spectrometry and Heat Budget Analysis
This project was an attempt to find an indicator for soil moisture using a radiation or heat budget analysis. The inputs were a subset of a Landsat TM scene, a digital elevation model, and temperature and humidity atmospheric soundings for a nearby vertical profile. Over 25 layers were computed using ENVI band math and the results were fed into a simple radiation balance equation to obtain a product that shows latent heat of evaporation, a scaled version of evaporation. This product was found to correlate strongly with NDVI, rather than soil moisture.
Pamela Soto - Sea Level Rise and Wetland Loss in Coastal Louisiana
This project analyzes the extent of vegetation destruction in the New Orleans section of the Mississippi River Delta as a result of Hurricane Katrina in 2005. The project uses three images from August 31, 1985, October 4, 2003, and October 9, 2005 respectively, all acquired by Landsat 4-5 TM. Using several classification methods and NDVI stacking, the researcher was able to quantify the amount of vegetation that was displaced by water post-Katrina. Unexpected findings included the amount of urban development that occurred in New Orleans in the two decades preceding Katrina, and a section of the marsh that appeared healthier after Katrina than before.
Jiani Yang - Monitoring Norway Spruce Migration in the Scandinavia Peninsula
The global warming may lead the potential of plants’ migration within decades. When tracing
the record from fossil record, plants migration can be rapid enough to indicate the climate change
by looking at its leaps in a long-distance. In this final project for remote sensing class, the
distribution of Norway Spruce will be identified during 1988 and 2004 in Scandinavia Peninsula.
By using ENVI and GIS, the distribution changes of Norway Spruce were identified within
different years. Supervised classification method is used to create training regions to quantify the
land use change. Results showed Norway Spruce’s geo distribution has been changing within 16
years. The reason to cause that change might be due to climate change. However, climate change
may not be the exclusive reason to explain that phenomenon since human activity and water
volume change can also determine the distribution of plants. The tool of ENVI can help with
tracking plants identify. The finding will be helpful for forestry managers to identify potential
tree line as an environmental indicator to do the further monitoring research on that.
Steven Patriarco - Evaluating Forest Change in the Greater Gunung Palung Region of West Kalimantan, Indonesia from 2007-2014
Gunung Palung National Park (GPNP), spanning an area of 900 km2 in the province of West Kalimantan, Indonesia, is among the most biologically diverse regions on the planet – home to an estimated 178 bird and 71 mammal species, including 17% of the total population of endangered orangutans (Pongo pygmaeusaous). The Park is situated in a broader landscape that has seen dramatic changes in recent years, beginning with the industrial logging boom of the 1970s, and continuing in recent years with rapid oil palm expansion in the province that has contributed greatly to Indonesia’s rise as the #1 global producer of palm oil. In this study, I utilize Landsat 5 & 8 imagery, SRTM 90m global data, and vector data from the World Database on Protected Areas (WDPA) and the RePPProt: Land Systems of Indonesia and Papua New Guinea database to assess two key questions: (1) Whether, and to what extent, industrial-scale oil palm expansion (OPE) is occurring on carbon-rich, protected peatland soils in the greater GPNP region (Row 121 Path 61), and (2) Whether, and to what extent, industrial-scale OPE is occurring in the 2.5km buffer zone surrounding GPNP, an area known to be of great importance for local fauna.
To perform my analysis, I use a combination of NDVI change detection of 2007 and 2014 NDVI indexes, and RGB change detection of forest to non-forest classes. NDVI change detection reveals that 27% and 19% of total protected peatland soils and 2.5 buffer zone area have undergone significant drops in NDVI (>0.15) from 2007-2014, respectively, with paired visual inspection revealing the primary source of these declines to be industrial-scale OPE – as evidenced by straight lines and regular grids characteristic of oil palm monoculture – further confirmed by oil palm concession data and locations from the World Resource Institute’s Global Forest Watch (GFW) tool. RGB change detection of forest change, which was restricted to a spatial subset of the primary region of interest, found a reduction of forest cover from 2007-2014 of approximately 8,000Ha, or 18.7%, together with an increase of bare soil and non-forest vegetation of approximately 4,800Ha and 3,100Ha, respectively. Here again, visual inspection confirmed OPE as a significant, if not primary, source of this change, a conclusion likewise supported by oil palm concession data from GFW. While my analysis makes it clear that significant industrial-scale OPE is occurring on carbon-rich peatsoils and in the 2.5 km buffer zone, due to the challenges of distinguishing palm oil from other vegetation types, it is unclear the exact geographic extent of this expansion.