Mapping Coastal Change Using LiDAR and Multispectral Imagery Contributor: Patrick Collins, Technical Solutions Engineer Presented by
TABLE OF CONTENTS Introduction... 1 Coastal Change... 1 Mapping Coastal Change with Low Resolution Imagery... 2 Low Resolution Imagery Benefits... 5 Assessment of Coastal Change using High Resolution Satellite Imagery... 6 Assessment of Coastal Change using LiDAR... 7 Low-Resolution Data vs High-Resolution Data... 9 LiDAR Data... 10 Conclusion... 10
INTRODUCTION Advances in remote sensing and GIS technology have improved our ability to accurately measure the causes and effects of costal change. The ever-increasing resolution of satellite and aerial imagery, an increased availability of LiDAR coverage, combined with software packages that make it easier to exploit the information contained in that data has allowed the scientific community to better understand the forces behind coastal morphology. In this report, learn about how a team from Exelis assessed different data types for their usefulness in tracking and understanding coastal change. Data types examined include Landsat data, high-resolution satellite data, and LiDAR data. COASTAL CHANGE The world s coastlines are in a constant state of flux due to weather patterns, human interaction, and other phenomenon that are constantly changing the makeup and distribution of coastal dunes and shorelines. The beaches and sand dunes that comprise the shoreline serve as protective barriers that keep the ocean from inundating coastal communities. When large weather systems hit, massive amounts of sediment can be removed or deposited in an area, causing major disruptions. Changes in the dune and beach composition can result in increased flooding potential. Floods cause large-scale disruption of key services, and can paralyze entire cities for days while they wait for the water to recede. The economic cost of mitigating flooded areas is immense, and is responsible for millions of dollars in cost annually. Coastal change can also adversely affect transportation infrastructure, destroying roads, rail-lines, shipping facilities, and other important aspects of commerce. In areas where there is a disaster, this can prevent emergency crews from reaching victims and providing timely assistance. Coastal change can cause major property loss, both from flooding events and by assets being destroyed as coastlines recede or move. 1
Since coastal areas provide both economic and environmental benefits, the ability to analyze and quantify natural and man-made changes to the coastline is very important for a number of industries. Being aware of how the shoreline could change under certain circumstances promotes pre-planning for change, and results in reduced impact from coastal change. MAPPING COASTAL CHANGE WITH LOW RESOLUTION IMAGERY The first case study looked at the effectiveness of low-resolution satellite data from the Landsat constellation for mapping coastal change. The test area was along the northern coast of Alaska, and the study evaluated the coastal change over a period of 30 years from 1978 2008. For this project, hand-digitization was not viable. Using this process to track Arctic Ocean coastline changes as they occurred over several decades would be time-consuming and resource intensive. On top of that, the resulting dataset would likely contain significant analyst-introduced errors or differences in image-feature interpretation among different analysts. In short, it would be extremely ineffective for the stated purposes. Instead, ENVI image analysis software was used for classification and thematic change detection. Image analysis software increases the accuracy of coastal change results with reduced time and effort. As a result, extremely large areas of coastline can be reliably assessed in a much shorter time frame and using fewer resources, all while reducing the potential for human error. For this exercise two images from the Landsat Archive were chosen. These images were selected for two criteria. First, they were separated by a significant amount of time to ensure quantifiable change between the images. Second, the land cover in both images was free of snow and ice to ensure quality results. The 2 scenes were chosen from 1978 MSS data and 2008 ETM+ data in order to determine changes in erosion and sedimentation. Figure 1 below shows 1978 Landsat image loaded into ENVI and being readied for processing. 2
Figure 1: Landsat data over the study area courtesy of USGS The analysis was performed by running a classification on each of the images to determine areas of change. The Classification workflow in ENVI was used to identify areas within the scene that corresponded to known feature types. These known features gave the software examples of the features to be identified. With examples of land, ice, and water identified by the user, the software could automatically pull similar features from the image. Once the classification was run, the program presented the image as broken into its classifications of land in red, water in blue, and ice in green. The same steps for classification can be used on each Landsat scene available, as long as it has clear coverage of the study area. Figure 2 shows the final classification for the 2008 dataset. This particular image includes ice in the bay, but the analysis can be configured to remove this from the computation to focus on changes to the coastline. 3
Figure 2: Landsat data classified in ENVI Next, the derived datasets were used to perform change detection between two images. There were two goals. The process would isolate areas that had been classified as land in the 1978 data that had later changed to water (also known as areas of erosion). It would also isolate areas that had been classified as water that had later changed to land (also known as areas of accretion). Figure 3 shows the results. Dark green and olive areas show where erosion has occurred, and light green shows areas where sediment has been deposited. 4
Figure 3: Final assessment of coastal change LOW RESOLUTION IMAGERY BENEFITS In general, low-resolution data is very useful for detecting coastal change on a macro scale, including assessment of coastlines at a country or continental-level scale. It is also useful for assessing drastic changes to an area. Low-resolution data is less useful for micro-scale changes or accurate assessment of the area and volume of sediment displaced because the coarse resolution of the data makes it impractical to monitor coastlines on a smaller scale. 5
ASSESSMENT OF COASTAL CHANGE USING HIGH RESOLUTION SATELLITE IMAGERY The second study focused on an area of Fire Island where Hurricane Sandy had carved a new inlet out of the barrier dunes in 2012. For this assessment, high-resolution satellite imagery obtained from DigitalGlobe was used. The goal was to run change detection on the two images to identify areas of major change along the coastline. Figure 4 shows the before and after images of the study area on Fire Island, where the hurricane cut the new inlet through the dunes. The erosion is drastic. This dune had apparently been over washed in previous storm events, which you ll be able to identify when looking at the extracted elevation data. Figure 4: Before and after images of study area courtesy of DigitalGlobe, Inc. The two datasets were imported into ENVI for processing and generated a fairly accurate twodimensional assessment of major changes that occurred in the area. Since the entire vegetative coverage of the dunes was so drastically changed, a relatively simple detection algorithm was used so as to highlight only areas of major change. Figure 5 shows where the software identified these areas of change, effectively mapping the creation of the new inlet and the erosion of the shoreline. 6
Figure 5: Assessment of major change using ENVI ASSESSMENT OF COASTAL CHANGE USING LIDAR Next, LiDAR point clouds from USGS were examined in an effort to assess volumetric loss of sediment from the dunes from the storm. Figure 6 is a 3D view of the point cloud showing the study area after the damage caused by the storm. Figure 6: LiDAR data over the study area courtesy of NOAA Coastal 7
Digital Elevation Models were then automatically extracted from the point clouds using ENVI LiDAR. Next, this data was imported to ENVI to classify the elevation models and run further analysis to determine where sediment had been removed or added to the coastline. Figure 7 shows both the pre- and post-event elevation data as classified by height. Classifying the data by height helps to visualize the different datasets, but also to derive data to use for a volumetric change detection analysis. For this height model lower elevation areas are in green, with the higher areas being colored in yellow and red. Figure 7: Pre and post-event elevation data classified by height The final step in testing the LiDAR data was to run band math between the two elevation datasets to come up with a volumetric change difference assessment of the area. Figure 8 shows the final results of that assessment as exported from ArcGIS. In this image, areas of red and yellow depict areas of sediment loss, or erosion, while areas in green depict areas of sediment gain, or accretion. The polygons surrounding the various categories of change are where ENVI has exported shapefiles to assist in further analysis. Overall, the costal change over this area from Hurricane Sandy resulted in the addition of over 36,500 cubic meters and the removal of almost 87,000 cubic meters of sediment from this area during the storm. Enough sand was removed from this area to fill almost 35 Olympic-sized swimming pools. 8
Figure 8: Final map depicting volumetric movement of sand LOW-RESOLUTION DATA VS HIGH-RESOLUTION DATA This exercise showed how high-resolution data and LiDAR can be used to depict coastal change on a small area of interest. In situations where the resolution of data such as Landsat is too coarse to accurately assess change, high-resolution satellite data offers a way to track two-dimensional change. Using such imagery, areas of major change within the scene were assessed quickly and then the results were exported. 9
LIDAR DATA LiDAR is a very useful data type when assessing coastal change, because the elevation information that can be extracted from the point clouds is extremely accurate and can be used to depict volumetric, or three-dimensional, change. Using this data, elevation data was extracted from the point clouds, the difference in elevation between the two datasets was calculated, and the amount of change that had occurred over the study area was determined. CONCLUSION GIS and remote sensing advancements have gone a long way toward improving our understanding of coastal morphodynamics. Different data types are useful for different types of coastal monitoring, and it is important to understand the goals of your analysis prior to deciding which type of data to use. While low-resolution data is ideal for large-scale mapping projects such as mapping at a country or continental scale, high-resolution imagery is much better suited for small-scale analysis and projects where more accurate results are needed. Where time and data availability permit, LiDAR is the best individual data type for assessing costal change from a volumetric standpoint, as it allows you to create and analyze extremely accurate elevation information to reach your results. However, as with high resolution imagery and other larger datasets, the size of your study area and the computing requirements for LiDAR may potentially make using LiDAR prohibitive on a larger scale. Note: This whitepaper is the product of the transcript of a presentation given at the International LiDAR Mapping Forum 2014, and available online at www.sparpointgroup.com. While the speakers are cited here as contributors, this whitepaper was not written by the contributors or speakers who appeared in the presentation, nor is it endorsed by the contributors or speakers, or any company, organization or entity they represent. For more information on how this whitepaper was produced, send inquiries via email to info@sparpointgroup.com. 10