Possible Site Locations
Possible Site Locations
See the presentation here: TomlinThesisProposal2018
Riparian zones are transitional areas of trees and shrubs that run along the banks of a river or stream (Cohen 2014). The riparian area serves as a boundary between the aquatic and the upland environments. The composition of the soil, plant, and animal communities are specific to these wet environments. The riparian zone may include wetland, aquatic, and upland areas and have influence on grasslands, woodlands, and non-vegetative biomes (Fu et al. 2015). Riparian zones provide several ecological services to wildlife and humans. The soil and vegetation buffer against nutrient flow and flooding (Cohen 2014). The riparian areas also serve as a habitat for many species important to both aquatic and terrestrial systems.
Riparian buffers can take the form of a natural forested buffer or human made vegetation strips. The definition of a riparian area varies depending on the study or stakeholder addressing it. The definition may include soil moisture, plant communities adapted to moist soil, soil pH, hydrology, and landform features. The United States Department of Agriculture (USDA) identifies riparian area traits as being adjacent to water, linear, and lacking clearly defined boundaries and a transitional boundary between aquatic and upland environments (Palone and Todd 1997). Natural forest buffers are the preferred buffer type for ecological and hydrological benefits but are often not feasible (Klemas 2014). Such an environment promotes soil health, which is important for buffering excess nutrients, sediment, and water (Fu et al. 2015).
The USDA has developed a flexible three zone system for the planning of forested riparian buffers. Zone 1, closest to the water, consists of mature trees with roots that act to stabilize the stream banks and a canopy that shades the water surface (Palone and Todd 1997). This is particularly important in headwater streams where canopy shade impacts stream temperature (Hawes and Smith 2005). Zone 2, upland of Zone 1, is a managed forested area where nutrients are filtered through natural processes of plant uptake and microbial processes in the soil, including denitrification. Sediment is trapped by trees and foliage, and nutrients are infiltrated from surface runoff in Zone 2. Furthest from the water is Zone 3, where grass buffer strips are recommended to slow surface runoff and trap sediments, nutrients, and pollutants.
The widths of the riparian zone may vary between 5m and 60m, depending on slope and functional goal of the buffer (Fischer and Fischenich 2000). The buffers are more effective when the corridors are long, contiguous, and generally wide. This prevents a channel to form and increase runoff into the stream. Structural diversity and ecological corridors also add to the benefit of a well-managed riparian buffer by adding niche micro environments for habitat and accessibility to the river from upland fauna.
Special focus should be paid to headwater streams as they interact with the surface processes with the greatest area and the greatest length of stream. There is a larger impact to water quality by increasing the riparian buffer in headwater streams than by increasing the width of higher order stream buffer widths (Hawes and Smith 2005). It is in the headwater streams where temperature can be the most affected by tree canopy and have a greater influence on low order and ephemeral streams. For this reason, fixed width buffer throughout the watershed is generally recommended for management and legislation purposes (Fischer and Fischenich 2000).
Chemical, structural, and biological soil properties in the riparian zone are important factors in the development and health of the area (Ye et al. 2017). Riparian soils act as both a sink and a source for phosphorus in streams and rivers. The bioavailable form of phosphorus is a key component in the water quality and eutrophication. Human activities such as agricultural development and physical disturbance from tourism can lead to changes in the accumulation and transport of phosphorus in riparian soils. Riparian plant community composition is also an important determinate of riparian soil phosphorus, nitrogen, pH, and organic matter through litter fall (Medina-Villar et al. 2015). Leaf litter dynamics, soil moisture, and permeability are all key determinants on the soil’s ability to hold and infiltrate water (Klemas 2014). High clay soils may cause more runoff and sandy soils increase infiltration beyond usefulness (Hawes and Smith 2005). Infiltration of water into riparian soils charges the water table and buffers against flooding when combined with riparian vegetation.
There is a lack of understanding of riparian zones globally and locally (Fischer and Fischenich 2000). They are dynamic systems that cover a large geographic area (Sothe et al. 2017). Human activities have exacerbated the changing conditions of riparian area through climate change, pollution, and urban sprawl. Riparian zones are commonly disturbed and are often hotspots for invasive plant invasion, which can negatively impact leaf litter dynamics (Medina-Villar et al. 2015) and soil properties. When utilizing riparian zones, the value of the land is determined based on its market value, rather than including the ecological value, which increases the exploitation of these important habitats (Fu et al. 2015). Efforts are constantly being made to make vegetation and riparian analysis more accessible to stakeholders for land use planning and regulatory strategies for the ecological security of riparian zones (Kollár et al. 2015).
Geographic information systems (GIS) and remote sensing have quickly become indispensable tools in environmental biology. Computer modeling and machine learning are less expensive and arduous than traditional field mapping techniques due to the large scale (Sothe et al. 2017). Open access public data are available through government portals as well as in Google Earth Engine (GEE), a cloud based planetary analysis tool that has been growing in popularity, as it can process complicated algorithms on very large datasets (Goldblatt et al. 2016). The data stored in the images from satellites are important for management of natural resources, including agriculture and forests, and other land uses and disturbances (Madhura 2015). GIS and remote sensing are commonly used in vegetation analysis because much larger areas can be covered than by ground based studies.
The Chesapeake Bay watershed is about 166,534 km2 and spans six states (Palone and Todd 1997). The main tributary to the Chesapeake Bay is the Potomac River. A large tributary to the Potomac is the Shenandoah River. The Shenandoah runs North from Virginia into West Virginia and enters the Potomac at Harpers Ferry, WV. The surrounding area is socially and economically important as it encompasses agricultural, residential, recreational, and historical land uses (Thomas et al. 2013). Harpers Ferry National Park lies at the confluence of the Potomac and Shenandoah Rivers. Efforts to establish and protect the riparian habitats there have been recommended to mitigate nitrogen and phosphorus loads (Thomas et al. 2013).
The confluence of the Potomac and Shenandoah Rivers lies in the Great Valley at the eastern edge of the Northern Ridge and Valley region of the Chesapeake Bay. The Northern Ridge and Valley region is characterized by northeast to southwest parallel ridges that span parts of Virginia, West Virginia, and Pennsylvania. Known for its karst environment, small to medium sized streams, and low mountains, the Great Valley extends from the Blue Ridge to the Allegheny Front. The soils in the Blue Ridge are generally Inceptisols and Utisols with significant rock outcrops. The Blue Ridge’s growing season is 220 days with warmer slopes facing southeast.
Forested riparian tree species found in the area include silver maple (Acer saccharinum), sycamore (Platanus occidentalis), and black walnut (Juglans nigra) (Thomas et al. 2013). Important salmonid fish species inhabit the streams in this area as well. Salmonids are particularly sensitive to warmer temperatures and tree cover, which along with geomorphological inputs, regulate stream temperatures (Johnson et al. 2017). The rich biological diversity of native fauna includes foxes, squirrels, river otters, great blue herons, bald eagles, and salamanders.
The objective of this study is to develop a tool that can be used to identify unhealthy, moderately healthy, and mature riparian forest zones along the Shenandoah and Potomac Rivers and their tributaries. Vegetation metrics, soil properties, spectral signature and environmental factors will be used as inputs for a machine learning algorithm that outputs a classification of the scene. The tool will leverage the benefits of cloud computing, remote sensing, and ground truth labeled data (Goldblatt et al. 2016) in a way that will allow for landscape scale analysis at a low cost. This study aims to develop a GIS tool that generates products drawn from existing sources of public data.
MATERIALS AND METHODS
The study will focus on an area of Shenandoah River and Potomac River watershed spanning from about 20 km north and south and 10 km east and west of Harpers Ferry, West Virginia, where the confluence of the Potomac and Shenandoah Rivers meet. A square drawn around the total study area will include about 1100 km2 and include areas in the states of West Virginia, Maryland, and Virginia (Figure 1). At the confluence, the Shenandoah is a 4th order stream and the Potomac is a 6th order stream.
Figure 1 – A map of the proposed study area around Harpers Ferry, WV.
While this study focuses on the two main rivers, the collection and analysis of ground truth labeled data will include a selection of 1st – 3rd order tributaries as well. This will allow for the use of the resulting tool in studies that focus on headwaters. Study site locations will be generated using a random point generator applied to a map of the study area that has been clipped to focus on the waterways. Further investigation will be required before actual sites are finalized, since many of the sites along the streams are not accessible to the public. The goal would be to locate at least 50 sites along each river with a mix of dense and sparse vegetation. Adjustments to the study area size and number of ground truth labeled data will be adjusted depending on the performance of the model’s accuracy.
The number of ground truth labeled data needed to classify an area can vary greatly. Some studies rely solely on training points that are created from visual inspection of the imagery and prior knowledge of the study area (Klemas 2014), others use thousands of training points collected from field observation (Colditz 2015). An optimal number of training points (1 pixel) for the entire study area would be 200-240 (Li et al. 2014) . The classification described in this proposal will be only calculated on the area determined in the V-BET tool, which should clip the area of study up to approximately 10% of the total area (Gilbert et al. 2016).
The study plots will be a 10 m fixed radius plot, allowing for each site to train 4 pixels in the classification. With 50 sites for soil and canopy data collection, and 50 non-forested locations selected for roads, rails, and buildings, a total of 250 pixels can be trained for classification. The algorithm will be trained using 80% of the training data and 20% of the training data will be used for validation (Forkuor et al. 2017). The ground truth labeled training data collection will focus on about a quarter of the total study area with validation ground truth labeled data collected throughout the rest of the study area (Forkuor et al. 2017).
Satellites Imagery and Data Acquisition
Spectral studies on riparian zone detection using Landsat at the landscape level can detect cover types, but they are often not fine-grained enough to detect riparian areas (Tormos et al. 2014). Very high resolution can be obtained for riparian analysis with the use of aerial photography (Kollár et al. 2015) but often requires significant investment to make riparian analysis an effective tool for planners. While the low spatial resolution of Landsat is not optimal, the high temporal resolution is useful in analysis of riparian zone dynamics (Klemas 2014). Sentinel – 2 refers to two modern satellites owned and operated by the European Space Agency (ESA) which offer similar spectral bands to Landsat at a higher resolution. Classification of Landsat and processed Sentinel – 2 images (Vuolo et al. 2016) can be performed with similar results (Sothe et al. 2017).
Sentinel-2 data available in the Google Earth Engine repositories is top-of-atmosphere reflectance. Landsat data are available in surface reflectance and can be a useful tool in determining if the Sentinel-2 data need to be further processed with the European Space Agency’s Sen-2-Cor atmospheric correction tool (Sothe et al. 2017). Sentinel-2’s spectral bands were designed in part to be used in this way to leverage the breadth of work historically conducted with Landsat.
Georeferenced soil data will be provided from the USDA Web Soil Survey (WSS) and include soil physical, chemical, and health properties (Soil Survey Staff 2017). The NASA Global Land Data Assimilation System (GLDAS 2.1) is a model derived from both satellite and ground observations. GLDAS generates surface states and fluxes, such as soil moisture, soil temperature (Rodell et al. 2004).
Other spatial data listed in Appendix I will also be integrated into the classification algorithm. This allows for the inclusion of biological and environmental factors to be considered for the classification. The World Wildlife Fund (WWF) manages a database, HydroSHEDS which includes a suite of flow and drainage related products useful in studies involving river networks, watershed boundaries, drainage directions, and flow accumulations (Lehner et al. 2008). The USGS National Elevation dataset provides a Digital Elevation Model (DEM) at 1/3 arc-seconds (10 m) that will be used for the Valley Bottom Extraction (V-BET) tool in ArcMap. The Joint Research Centre (JRC) Global Surface Water dataset is Landsat derived and provides surface water products for permanent water, seasonal water, percent occurrence, and change in water (Pekel et al. 2016) and will be used to help understand stream characteristics and wetland inundation.
The WWF HydroSheds database is based off the same DEM being used for the V-BET tool, but it provides a different, richer suite of products. Similarly, the JRC dataset and the GLDAS model are based off Landsat data. The JRC provides watershed specific products that can give insight to water seasonality and the GLDAS model provides soil specific information. A surface corrected Landsat scene can be used to visually inspect anomalies in the JRC or GLDAS data as well as the Sentinel-2 output.
The goal of this study is to create an online tool that can be used by researchers as well as expanded upon using publicly available data with public access through a web browser. Classification of landscapes with GIS and ground truth labeled data has been shown to yield useful results (Sothe et al. 2017). This study proposes to use accepted Landsat classification algorithms and industry standard spatial analysis software, such as ArcGIS, TerrSet, R, and ENVI, to validate that the results provided in Google Earth Engine are accurate at the scale provided by Sentinel-2. See Appendix I for a summary of data sets that will be used in this study.
In situ Data
To conduct a supervised classification of the study area, ground data must be collected at random sites along the streams and will include a mix of disturbed and canopied locations along both rivers. Georeferenced canopy composition, canopy cover, and soil information will be collected at each site. The results of the field survey will be entered into a database for use in data analysis.
At each site, a GPS receiver will be used to mark the location of the centroid of the site. Fixed radius plots (Petit et al. 1995) will be measured out from the centroid and in a circle with a 10 m radius (Packard and Radtke 2007). This will allow for the ground truth labeled data to include four 10-meter square pixels inside the plot to match the resolution of the DEM and Sentinel – 2 data. The ground truth labeled data can be resampled to match the Landsat derived products as necessary. Multiple plots can be taken in adjacent areas if spacing permits. This will allow for more data to be collected without changing the resolution.
Canopy composition and cover will be recorded at each site. A count of each tree species will be taken for the entire plot. Additional plots will be recorded for the purposes of classifying areas that are easily recognizable by the classification. Classes include buildings, roads, rails, and fields.
Canopy coverage will be determined using a canopy densitometer (Fiala et al. 2006). The instrument is engraved with 36 squares on a mirrored surface that is viewable through a side aperture. The percent canopy density will then be calculated using the proportion of the crosshairs that were intersected by canopy cover.
Canopy composition will be recorded at each study site by counting all the trees that fall within the fixed radius plot that make up the canopy with at least a 4-inch diameter (Grebner et al. 2013). Tree species being recorded will be mature trees that reach the canopy of the riparian zone. A percent cover for each species will then be calculated (Stumpf 1993).
While it would not be necessary to make a broad classification of forest cover using an in-depth analysis of the forest makeup (Klemas 2014), future researchers could build upon the data to supplement their own work. Possible areas of application include forestry, resource management, policy, hazards and disaster management, cloud computing, and remote sensing (Cohen 2014; Soulard et al. 2016).
A soil survey will be conducted at each site. Important soil characteristics will be recorded into the database. Standardized soil survey tests will include soil structure, resistance to penetration, infiltration rate and pH. The soil survey will take place at the centroid of the plot (Sylvia et al. 2005).
The soil sampling methods will be a combination of methods established by Cornell University (Moebius-Clune et al. 2016) and the USDA National Resources Conservation Services (Soil Survey Staff 1999). Methods were selected to ensure standardized field and lab methods for the soil properties important to this study. Composite soil samples will be collected at three locations by removing surface debris, digging an 8” hole, taking a 6” vertical slice 2” thick and storing the slices in labeled Ziploc bags on ice packs in a cooler (Moebius-Clune et al. 2016). These samples will be used for the soil organic matter, aggregate stability, and pH testing in the lab. Each subsample will be divided into three replicates of each test.
Soil pH will be tested with a Lignin pH electrode probe by suspending 1-part soil in 2-parts water (Moebius-Clune et al. 2016). Soil organic matter will be tested by drying samples at 105 degrees C, recording pre-ashed weight, ashing at 500 degrees C in a Lindberg Blue M drying oven and weighing again. Percent loss (%LOI) is calculated by the difference between the two measurements. Percent organic matter (%OM) is then calculated with the equation (Moebius-Clune et al. 2016):
%OM= (%LOI * 0.7)-0.23.
Aggregate stability will be determined by sieving the sample with a 2mm sieve. The sieved sample will then be weighed. The sample will then be slowly wet in a sieve and then wet-sieved. The aggregates will then be dried and weighed. Next, the aggregates will be dispersed in a Calgon solution. Finally, the sand will be dried and weighed. Water stable aggregates can then be calculated with the following equation (Soil Survey Staff 1999):
% of soil > 0.25mm =
((weigh of dry agg. – sand) / (weight of dry soil – sand)) X 100.
Soil properties tested in the field will include infiltration, surface and subsurface hardness, and soil texture by feel. Infiltration will be tested using a 6” diameter ring pressed into the soil and filled with 44 mL of water. The time it takes for the water to drop 1” is recorded and the processed repeated (Soil Survey Staff 1999). Texture by feel will be determined by kneading 25 g of soil in the palm, adding water, and then following the NRCS soil by feel flow chart to determine the sand / silt / clay content (Soil Survey Staff 1999). The results will be compared to the data included in the Web Soil Survey. Surface hardness will be tested using a penetrometer to drive into the soil and recording at which depth 300 psi is passed and then dropped (Moebius-Clune et al. 2016). Maximum hardness will also be recorded for the depths of 0-6” and 6-18”.
The classification will be conducted in GEE. Google stores all its data directly on servers and is available via the GEE Application Programming Interface (API). Any data that are not available on the GEE servers will need to be uploaded to Google Assets along with the database table from the field study.
Analysis and data manipulation will also be conducted locally using ArcMap. The Valley Bottom Extraction Tool (V-BET) will be used to determine the maximum riparian corridor extent (Gilbert et al. 2016). Data downloaded from the WSS will also be converted using the Polygon to Raster tool in ArcMap, where each band in the resulting raster will represent a specific soil characteristic. Finally, sample data can be downloaded from GEE and analyzed locally to see how GEE performs compared to established technology.
As part of the model, important vegetation metrics will be classified in the area indicated as valley bottom from V-BET. The classes will be determined by thresholds of vegetation canopy cover, community composition, and soil properties along with environmental factors such as slope, aspect, and spectral signature. Thresholds will be established to classify the health of each forested riparian area: unhealthy, moderately healthy, and mature riparian forest, based on the spectral signature, soil survey, and canopy analysis.
Google Earth Engine (GEE) supports the Random Forest (Breiman 2001) algorithm for supervised classification. Random Forest is a commonly used non-parametric classifier for vegetation and wetland studies (Dubeau et al. 2017) as it maintains the ability to classify based on discrete or continuous datasets and handle noise (Sothe et al. 2017). The accuracy of the classifier will then be produced using a confusion matrix algorithm, which reports a percent value of model accuracy for the pixel classification of the scene (Stehman 1997).
Riparian vegetation is an important aspect of the overall health of terrestrial and aquatic systems. Special attention has been given to understand this transitionary zone in streams as it affects flooding, nutrient load, and temperature buffering characteristics of that system (Klemas 2014). Analysis of riparian vegetation presents difficulties due to the cost and effort necessary to conduct landscape level studies on riparian vegetation.
This study proposes to develop a software tool designed to classify forested riparian zones using satellite and in situ data. The model inputs include multi – spectral imaging, elevation, stream flow, forest canopy and soil survey data. The final output will include four main classes: Non-riparian, Unhealthy Forest, Moderately Healthy Forest, Healthy Forest for the study area.
The resulting methodology and software would serve as a baseline for a novel tool capable of analyzing large areas of riparian vegetation. The tool would include a richer set of inputs, allowing for a more in-depth analysis and cross pollination of technologies into forest and soil studies. This tool can be used by researchers in studies pertaining to surface and groundwater fluxes, brook trout habit, and land use planning (Lowrance et al. 1997), flood management (Cohen 2014), or fire management (Soulard et al. 2016). Access to free, up to date, high resolution, and preprocessed data via a web browser would save time and money in these studies, where financial or personnel capacity for spatial analysis is limited.
Additional possible goals for this study could include the integration of specific details about the model inputs, such as canopy composition, in an additional output layer. This is a current topic of research in remotely sensed vegetative studies (Sothe et al. 2017). Another value-added product that could be created in this study is an expansion of the study area to include the entire Shenandoah or Potomac watersheds. Once the algorithm has been shown to work, it can be applied over a much larger area, fully harnessing the computing power of cloud services.
Cloud based computing of remotely sensed data in riparian studies is an emerging technology. Google has been at the leading edge in cloud based geospatial analysis with their Google Earth Engine (Soulard et al. 2016). The repository of GEE is limited to the products that Google has established are most important, and it does not include soil information from the Web Soil Survey. The only soil data available are from the GLDAS-2 model. Layers created in ArcMap for soil and valley bottom must be uploaded to an Asset in GEE.
Limitations exist in the spatial and temporal resolution that the data are available. Landsat scenes are spatially (30 m) and temporally (12 day) coarse for riparian studies, but offer a long historical (over 40 years) dataset. There is a treasure of research in vegetation analysis using Landsat data, so methods are well established and accepted. Sentinel – 2 has a higher resolution (10 m every 5 days) but with only a couple of years of data collection. Sentinel – 2 data are only available in top-of-atmosphere reflectance, which may need to be corrected before analysis (Sothe et al. 2017).
Riparian zones are complex systems that require consideration (Hawes and Smith 2005). This study proposes to approach the classification of the riparian forest along the Potomac and Shenandoah Rivers using a mix of established methods along with the addition of canopy and soil inputs. This method brings realism to the model; further inputs, such as water table dynamics, could be influential enough to require consideration in the model inputs.
Future studies could build off this platform in many different study areas. One such area could be to create a composite image for the mean riparian zones over the course of many scenes. In this way, a historical analysis of the riparian zone could be conducted. Compared with flood gauge data, the historical analysis could be used to identify forest canopy and flood dynamics.
The scenes can be combined by taking the median value for each pixel in each scene. By choosing the median value, high values caused by cloud cover and low values caused by shadows are removed. The result is a single scene with a value representative of the land area over time. This false image can then be used for classification of that land area over time instead of only using information from one scene (Wingate et al. 2016).
Temporal differences could also be observed by classification of an early scene, from 1984 for example, and the classification of a recent scene. Differencing the values of each pixel of the two scenes would result in a riparian zone change analysis. A combination of the two above techniques could lead to decadal land use change analysis of the study area (Wingate et al. 2016).
Studies that integrate vegetation analysis into stream water temperature could use output from this model to better understand the dynamics of canopy cover, groundwater (Wawrzyniak et al. 2017) and precipitation on salmonid fish (Johnson et al. 2017) temperature thresholds. These are areas where GIS and remote sensing are currently being used. LiDAR and thermal infrared imaging are used to detect river surface and stream temperature. These additional data could be integrated into the proposed tool.
Above are only a handful of important possible integrations of the technology being proposed herein. Key components of these studies include the manipulation classification of very large datasets over large areas, traditional ecological field methods, and complex systems that interface with the riparian zone, which are ecologically significant regions. The solution proposed integrates cloud computing, machine learning algorithms, ground truth labeled data, and publicly available Earth observation, ground observation, and modeled data to address the technical complexities of riparian systems while maintaining a measured level of realism to the model.
If successful, the proposed tool will assess riparian zone health, based on the most important inputs to the riparian zone: stream dynamics, vegetation, and soil. A value-added product of this study will be a publicly available software that is accessible through any internet browser for stakeholders, legislators, and researchers to assist their understanding of the study area. Further, this tool can serve as a development platform for future studies by either expanding it to other areas or by integrating it into a larger study. There is currently no such software solution available to the public that offers these capabilities.
Breiman L. 2001. Random forests. Mach. Learn. 45:5–32.
Cohen RA. 2014. Fact Sheet# 1: Functions of Riparian Areas for Flood Control. [accessed 2017 Jul 9]. http://masslib-dspace.longsight.com/bitstream/handle/2452/617362/ocn971134562.pdf?sequence=1&isAllowed=y.
Colditz R. 2015. An Evaluation of Different Training Sample Allocation Schemes for Discrete and Continuous Land Cover Classification Using Decision Tree-Based Algorithms. Remote Sens. 7:9655–9681. doi:10.3390/rs70809655.
Dubeau P, King D, Unbushe D, Rebelo L-M. 2017. Mapping the Dabus Wetlands, Ethiopia, Using Random Forest Classification of Landsat, PALSAR and Topographic Data. Remote Sens. 9:1056. doi:10.3390/rs9101056.
Fiala ACS, Garman SL, Gray AN. 2006. Comparison of five canopy cover estimation techniques in the western Oregon Cascades. For. Ecol. Manag. 232:188–197. doi:10.1016/j.foreco.2006.05.069.
Fischer RA, Fischenich JC. 2000. Design recommendations for riparian corridors and vegetated buffer strips. ARMY ENGINEER WATERWAYS EXPERIMENT STATION VICKSBURG MS ENGINEER RESEARCH AND DEVELOPMENT CENTER.
Forkuor G, Hounkpatin OKL, Welp G, Thiel M. 2017. High Resolution Mapping of Soil Properties Using Remote Sensing Variables in South-Western Burkina Faso: A Comparison of Machine Learning and Multiple Linear Regression Models. Hui D, editor. PLoS ONE 12:e0170478. doi:10.1371/journal.pone.0170478.
Fu BL, Li Y, Zhang B, Yin BS, Zhu HL, Xing ZF. 2015. Dynamic evaluation of ecosystem service value of the riparian zone based on remote sensing from 1986 to 2012. Solid Earth Discuss. 7:2151–2184. doi:10.5194/sed-7-2151-2015.
Gilbert JT, Macfarlane WW, Wheaton JM. 2016. The Valley Bottom Extraction Tool (V-BET): A GIS tool for delineating valley bottoms across entire drainage networks. Comput. Geosci. 97:1–14.
Goldblatt R, Wei You, Hanson G, Khandelwal AK. 2016. Detecting the Boundaries of Urban Areas in India: A Dataset for Pixel-Based Image Classification in Google Earth Engine. Remote Sens. 8:1–28. doi:10.3390/rs8080634.
Grebner D L, Bettingger P, Siry J P. 2013. Introduction to forestry and natural resources. 1st ed. San Diego, CA: Academic Press.
Hawes E, Smith M. 2005. Riparian buffer zones: Functions and recommended widths. Prep. Eightmile River Wild Scen. Study Comm. [accessed 2017 Jul 9]. http://eightmileriver.org/resources/digital_library/appendicies/09c3_Riparian%20Buffer%20Science_YALE.pdf.
Johnson ZC, Snyder CD, Hitt NP. 2017. Landform features and seasonal precipitation predict shallow groundwater influence on temperature in headwater streams: SPATIOTEMPORAL PREDICTORS OF GW DYNAMICS. Water Resour. Res. 53:5788–5812. doi:10.1002/2017WR020455.
Klemas V. 2014. Remote Sensing of Riparian and Wetland Buffers: An Overview. J. Coast. Res. 297:869–880. doi:10.2112/JCOASTRES-D-14-00013.1.
Kollár S, Vekerdy Z, Markus B. 2015. Remote Sensing Analysis of Riparian Vegetation Habitats. Int. J. Geoinformatics 11:63–67.
Lehner B, Verdin K, Jarvis A. 2008. New global hydrography derived from spaceborne elevation data. Eos Trans. Am. Geophys. Union 89:93–94.
Li C, Wang J, Wang L, Hu L, Gong P. 2014. Comparison of Classification Algorithms and Training Sample Sizes in Urban Land Classification with Landsat Thematic Mapper Imagery. Remote Sens. 6:964–983. doi:10.3390/rs6020964.
Lowrance R, Altier LS, Newbold JD, Schnabel RR, Groffman PM, Denver JM, Correll DL, Gilliam JW, Robinson JL, Brinsfield RB, et al. 1997. Water quality functions of riparian forest buffers in Chesapeake Bay watersheds. Environ. Manage. 21:687–712.
Madhura SV. 2015. Comparison of Supervised Classification Methods On Remote Sensed Satellite Data: An Application In Chennai, South India. Int. J. Sci. Res. IJSR 4.
Medina-Villar S, Castro-Díez P, Alonso A, Cabra-Rivas I, Parker IM, Pérez-Corona E. 2015. Do the invasive trees, Ailanthus altissima and Robinia pseudoacacia, alter litterfall dynamics and soil properties of riparian ecosystems in Central Spain? Plant Soil 396:311–324. doi:10.1007/s11104-015-2592-4.
Moebius-Clune BN, Bianca Nadine DJ, Gugino BK, Idowu OJ, Schindelbeck RR, Ristow AJ, van Es HM, Thies JE, Shayler HA, McBride MB, et al. 2016. Comprehensive assessment of soil health: the Cornell framework manual. Geneva, NY: Cornell University.
Packard KC, Radtke PJ. 2007. Forest sampling combining fixed- and variable-radius sample plots. Can. J. For. Res. 37:1460–1471. doi:10.1139/X06-321.
Palone RS, Todd AH. 1997. Chesapeake Bay Riparian Handbook: A Guide for Establishing and Maintaining Riparian Forest Buffers.
Pekel J-F, Cottam A, Gorelick N, Belward AS. 2016. High-resolution mapping of global surface water and its long-term changes. Nature 540:418–422.
Petit DR, Petit LJ, Saab VA, Martin TE. 1995. Fixed-radius point counts in forests: factors influencing effectiveness and efficiency.
Rodell M, Houser PR, Jambor U, Gottschalck J, Mitchell K, Meng C-J, Arsenault K, Cosgrove B, Radakovich J, Bosilovich M, et al. 2004. The Global Land Data Assimilation System. Bull. Am. Meteorol. Soc. 85:381–394. doi:10.1175/BAMS-85-3-381.
Soil Survey Staff. 2017. Web Soil Survey. Natural Resources Conservation Service, United States Department of Agriculture. [accessed 2017 Nov 1]. https://websoilsurvey.sc.egov.usda.gov/.
Soil Survey Staff NRCS United States Department of Agriculture. 1999. Soil Quality Test Kit Guide.
Sothe C, Almeida C, Liesenberg V, Schimalski M. 2017. Evaluating Sentinel-2 and Landsat-8 Data to Map Sucessional Forest Stages in a Subtropical Forest in Southern Brazil. Remote Sens. 9:838. doi:10.3390/rs9080838.
Soulard C, Albano C, Villarreal M, Walker J. 2016. Continuous 1985–2012 Landsat Monitoring to Assess Fire Effects on Meadows in Yosemite National Park, California. Remote Sens. 8:371. doi:10.3390/rs8050371.
Stehman SV. 1997. Selecting and interpreting measures of thematic classification accuracy. Remote Sens. Environ. 62:77–89. doi:10.1016/S0034-4257(97)00083-7.
Stumpf KA. 1993. The estimation of forest vegetation cover descriptions using a vertical densitometer. In: Joint Inventory and Biometrics Working Groups session at the SAF National Convention, Indianapolis, IN.
Sylvia D M, Fuhrmann J J, Hartel P G, Zuberer D A. 2005. Principles and applications of soil microbiology. 2nd ed. Pearson Education Inc.
Thomas JE, Costanzo SD, Lehman M, Dennison WC, Nisbit D, Nortrup M, Parsons M. 2013. Harpers Ferry National Historical Park natural resource condition assessment: National Capital Region. Fort Collins, Colorado: National Park Service Report No.: NPS/CATO/NRR—2013/746.
Tormos T, Van Looy K, Villeneuve B, Kosuth P, Souchon Y. 2014. High resolution land cover data improve understanding of mechanistic linkages with stream integrity. Freshw. Biol. 59:1721–1734. doi:10.1111/fwb.12377.
Vuolo F, Żółtak M, Pipitone C, Zappa L, Wenng H, Immitzer M, Weiss M, Baret F, Atzberger C. 2016. Data Service Platform for Sentinel-2 Surface Reflectance and Value-Added Products: System Use and Examples. Remote Sens. 8:938. doi:10.3390/rs8110938.
Wawrzyniak V, Allemand P, Bailly S, Lejot J, Piégay H. 2017. Coupling LiDAR and thermal imagery to model the effects of riparian vegetation shade and groundwater inputs on summer river temperature. Sci. Total Environ. 592:616–626. doi:10.1016/j.scitotenv.2017.03.019.
Wingate VR, Phinn SR, Kuhn N, Bloemertz L, Dhanjal-Adams KL. 2016. Mapping Decadal Land Cover Changes in the Woodlands of North Eastern Namibia from 1975 to 2014 Using the Landsat Satellite Archived Data. Remote Sens. 8:1–20. doi:10.3390/rs8080681.
Ye H, Yuan X, Zhou R, Wan J, Xu J. 2017. Distribution and Environmental Significance of Phosphorus Forms in Riparian Soils and River Sediments of Jianxi Basin, Fujian Province. Pol. J. Environ. Stud. 26:2331–2341.
Zarnoch SJ, Bechtold WA, Stolte KW. 2004. Using crown condition variables as indicators of forest health. Can. J. For. Res. 34:1057–1070.
|Shapefile / Raster||Description of Data||Resolution||Supplier||Retrieval Information|
|Landsat 8||Multispectral||30 m / 12 day||USGS||https://landsat.usgs.gov/landsat-surface-reflectance-data-products|
|Sentinel – 2||Multispectral||10 m / 5 day||ESA||https://sentinel.esa.int/web/sentinel/user-guides/sentinel-2-msi|
|National Elevation Dataset||DEM||1/3 degree||USGS||https://nationalmap.gov/elevation.html|
|Web Soil Survey||Soil properties||Various||USDA||https://websoilsurvey.sc.egov.usda.gov/App/HomePage.htm|
|GLDAS||Soil Moisture / Temp||1/4 degree||NASA||https://ldas.gsfc.nasa.gov/gldas/|
|JRC||Surface Water||30 m / monthly / yearly||European Commission||https://global-surface-water.appspot.com/|
It was an honor to be selected as a Best of Hood for 2017! I love my school!
Copied from Hood College website:
Why did you choose to pursue a graduate degree?
I started college as an adult wanting to change my career path from information technology to a field science related pursuit. I knew that to be competitive in that space I would need a graduate degree. I planned my educational career with that in mind and took extra science and math courses as an undergraduate in ensure my success.
Why did you choose Hood College Graduate School?
Hood College’s offering in a graduate level geospatial information systems (GIS) education was a big start. The program has a very strong focus on the applied aspect of GIS and remote sensing specific to the study of biological systems. As I was finishing my B.S. in environmental science, I could see the power of GIS and remote sensing in the future of landscape level ecological studies. The Hood advisors were responsive and motivated to bring me into the fold. Add that to the growing list of successful Hood students working at NASA and other important organizations sealed the deal. I knew I wanted to get my master’s at a campus location and the Hood campus and facilities are unquestionably pleasant.
What do you value most about your relationships with your professors?
The Hood professors have a strong and close interaction with the students. By the time a semester in any class concludes, I feel like I know my professor and they know me and my strengths.
What was your most rewarding class or academic experience at Hood? Why?
So far, my most rewarding experience at Hood has been my internship at NASA Goddard under the NASA DEVELOP program. I participated in a 10 week, dual capacity building internship where I was able to work with other students from around the country as well as NASA scientists under the NASA Applied Science Directorate at the world-renowned Goddard Space Flight Center in Greenbelt, Md. I was one of three recipients this year of the Science Systems and Applications Incorporated (SSAI) scholarship for problem solving and teamwork, skills I fostered at Hood.
How do you manage your many other obligations such as work and family with the demands of graduate school?
I am a proud husband and father of three: Lucille, Dorothy and William. The support I receive from them is enormous. The main key to success (as well as the hardest to maintain) I have found is to keep consistent on the work at hand and plan as far ahead as possible. Chipping off a little at a time allows for good planning and overall happiness.
Can you offer any tips to prospective students?
You can do it if you put the work in! Read the material before class and ask lots and lots of questions. Most importantly, work to advance your abilities in networking and as a team member. It may seem silly in class but in the science community, it is the many conversations between collaborators that make the real difference.
When you aren’t in class, what is your favorite thing to do in your spare time?
I love the outdoors and rafting. I have been a river guide for six years and get into it as often as I can. Nature is a big stress reliever (sailing, scuba, hiking, etc.) I also play and record guitar and drums, brew beer and paint and draw. I enjoy volunteering for causes such as the US Army Divers, Audible Eggs Easter Eggs for the Visually Impaired and the Harpers Ferry Outdoor Festival.
Text from Hood College Graduate School:
Jared Tomlin, C’16, a Master of Science candidate in environmental biology, was presented with a scholarship award from Science Systems and Applications, Inc. at a recent event held at the NASA headquarters. He worked with NASA this summer on a project focused on ecological forecasting. This video gives an overview of the project.
Jared is back at Hood to start work on his thesis, which will examine the effect of riparian zones on flooding in the Shenandoah Watershed.
AUG 10, 2016
This summer, Science Systems and Applications Inc. (SSAI) selected three participants to receive scholarships: Jared Tomlin from Goddard, Rachel Cabosky from Langley and Daryl Ann Winstead from Marshall. All three embody not only DEVELOP’s core values but those of SSAI! Congratulations to Jared, Rachel and Daryl Ann and thank you for your many contributions to DEVELOP
In celebration of the National Park Service Centennial in 2016, DEVELOP partnered with NPS to use NASA Earth observations to monitor change and threats to America’s national parks and inventory and monitoring networks. On August 9th, DEVELOP presented a poster session at the DOI building in Washington DC. Congratulations NPS on 100 years of excellence!
Text from Hood College Graduate School Blog:
Jared Tomlin, C’16, a Masters of Science in environmental biology candidate, is working with NASA this summer on a project focused on ecological forecasting.
Tomlin is conducting work as a participant in the NASA DEVELOP Program, which is a part of NASA’s Applied Sciences Program and operates at 13 locations throughout the nation. Tomlin’s project team is working at NASA Goddard Space and Flight Center in Greenbelt, Md., and partnering with the National Park Service and the U.S. Geological Survey to monitor and forecast the abundance and distribution of invasive brome grasses in the Northern Plateau.
The brome grasses impair the area’s native grasslands and contribute to a decrease in native species diversity. Understanding the behavior of the invasive species through space and time is key in developing successful management efforts.
“The program functions to give partner organizations, such as the National Park Service, the ability to better understand complex, landscape level environmental questions for decision making by utilizing the constellation of Earth observing NASA satellites, tools and operational support,” he said.
In addition to the years of field data collected by scientists in the area, the job requires the use of Landsat and Terra satellites, both part of NASA’s Earth observations fleet.
Tomlin earned a certificate in geographic information systems from Hood College in May, making him well equipped for the position. The selection process for participants in the DEVELOP program is considered highly competitive.
“Attending the Hood job fair with a résumé in hand to talk to the DEVELOP representative gave me a start, and my adviser was key in helping navigate the process,” he said. “A strong GPA with a background in GIS and Earth sciences, as well as technical ability in programming and design, were key in my acceptance.”
Tomlin learned about many different GIS and remote sensing solutions throughout his GIS course work, and he maintained a focus on environmental biology and climate change.
“The education I received at Hood College was paramount,” he said.
Before pursuing graduate studies at Hood, Tomlin attended Shepherdstown University in West Virginia, where he earned a bachelor’s degree in environmental science and sustainability. He plans to continue his education to earn a doctorate and go on to work at NASA or a similar organization.