Integration of Landsat Imagery and an Inundation Model in Flood Assessment and Predictions: A Case Study in Cook Inlet, Alaska

Transcription

1 Integration of Landsat Imagery and an Inundation Model in Flood Assessment and Predictions: A Case Study in Cook Inlet, Alaska Hua Liu Tal Ezer Department of Political Science and Geography Old Dominion University Norfolk, Virginia, USA hxliu@odu.edu Center for Coastal Physical Oceanography Old Dominion University Norfolk, Virginia, USA tezer@odu.edu Abstract High-temporal and spatial resolution coastal topography data is important in assessing and predicting floods. This study demonstrates the capability of remote sensing technology in deriving topographic information of flood areas. Cook Inlet, Alaska, with its large (8-10m) tidal range and extensive mudflat regions is selected as a study area. The shorelines at different tidal stages are detected from analysis of water coverage in Landsat satellite images. All the shoreline data from different times are next integrated with water level data from observations and the inundation model to produce a new topography maps. The method indicates a new way to evaluate the flood prediction of the existing Cook Inlet inundation model, and the potential of using remote sensing data to improve the accuracy of flood perditions by obtaining a high-resolution topography data in shallow regions and flood zones where landbase data are not available. Keywords-remote sensing; flood; prediction I. INTRODUCTION High resolution numerical flood models combined with near shore topographic information are very important tools to assess and predict floods associated for example with storm surge or tsunamis [1]. Inundation models have been developed to monitor and predict the flooding in different regions, such as the simulation models for the tidal flows and inundation processes in Cook Inlet, Alaska [2]. An example of the landscape in the upper Inlet is shown in Fig. 1. However, a potential constraint to the development of such numerical models is the lack of suitable validation data sources [3]. With the rapid development of remote sensing and Geographic Information System (GIS) technologies, it is now possible to develop high spatial and temporal resolution inundation profiles including topographic information as a support for flood monitoring. For example, European Remote Sensing Satellite (ERS-1) images were used to map overbank flooding in the Upper Misssissppi Valley during the Great Flood in summer 1993 [4]. LiDAR dataset also shows the potential for flooding assessment [5], but high cost and limited availability of LiDAR data can not sufficiently support the flooding analysis on the entire earth, and data may be missing in many places. A numerical model has been developed to monitor the dynamics of flooding processes in Cook Inlet, Alaska [2]; the same model was even used to study the movements of the endangered Beluga Whale species in the Inlet [6]. However, it is difficult to accurately simulate the tidal flood or evaluate the inundation model due to the lack of data on the morphology and topography of the mudflat regions. The topography data used by the model were based only on a subjective assessment from various maps and other references. The model resolution (~0.5 km in the upper inlet) is insufficient to describe the various narrow channels in the inlet, but increasing the model resolution would not be practical without having high resolution topography data. The objective of this study was to evaluate the capability of satellite remote sensing data acquired by Landsat platform in improving the model topography and assessing flood predictions. The results will provide very important supplementary information for the existing numerical flood model and further demonstrate the applicability of remote sensing data to flooding assessment and prediction. Figure 1. Photo of flood zone in upper Cook Inlet, Alaska, during low tide. During high tide the exposed land up to the location where the picture was taken will be covered by water.

2 II. METHODOLOGY A. Study Area Cook Inlet, Alaska was selected as the area of study in this project. It stretches 290 km from the Gulf of Alaska to the City of Anchorage (Fig. 2). It receives water from its tributaries such as the Susitna River and the Knik River. Cook Inlet contains active volcanoes which are closely related to tsunamis. The large semi-diurnal tides (8-10 m range) in the inlet flood and expose extensive mudflats regions (tens of square kilometers) twice daily. B. Remote Sensing Data Processing Remote sensing and GIS have been used extensively for monitoring water resources. Olmanson et al. [7] assessed water clarity of lakes in Minnesota during the period based on water information retrieved from Landsat TM and Landsat ETM+ imagery. Moderate Resolution Imaging Spectroradiometer (MODIS) data were applied to monitor red tide in southwestern Florida coastal waters in 2004 [8]. Here, sample of Landsat data will be evaluated for their usefulness to distinguish between water and land masses and thus map a moving tidal-driven shoreline. The idea has been originally mentioned in Oey et al. [2] and preliminarily tests were conducted with MODIS data analyzed by the Institute for Marine Remote Sensing at the University of South Florida ( but here a higher resolution Landsat data is added and more quantitative analysis is conducted. It is important to collect high-quality remote sensing images with different acquisition dates so that the water coverage derived can represent various tidal stages. One Landsat Thematic Mapper (TM) and five Landsat Enhanced Thematic Mapper Plus (ETM+) images were used to identify the water topography in the study. Table 1 record the acquisition date and time of each dataset. In the Table, the image acquired in 1989 was the only Landsat TM dataset, the rest were from Landsat ETM+. They were downloaded from SwathViewer website hosted by University of Alaska. Landsat TM data include seven spectral bands, among which bands 1-5 & 7 (thermal infrared) have a spatial resolution of 30 meters and band meters. Landsat ETM+ also has seven spectral bands with 30-meter resolution for bands 1-5 & 7, but its band 6 has a spatial resolution of 60 meters. Both TM and ETM+ data have a temporal resolution of 16 days. Only ice-free summer images were used in the study because Cook Inlet, Alaska is a high latitude region with a long winter season. All the satellite data were processed by using ERDAS IMAGINE, a major remote sensing program provided by ERDAS, Inc. We used USGS digital raster graphics (DRGs) as reference to geocorrect all the images to 1984 Universal Transverse Mercator zone 6N, at an accuracy of less than half a pixel from their original projection, Albers Conical Equal Area projection. We next chose unsupervised classification method and Gaussian Maximum Likelihood Classifier to classify the images into three categories: water, wetlands, and others. The color aerial photo of Anchorage obtained from Geographic Information Network of Alaska and was used as reference in image classification. Figure 2. The study area: Cook Inlet, Alaska. The focus of the study is the easternmost part of the Inlet, which includes two narrow branches, Knick Arm, north of Anchorage and Turnagain Arm, south of Anchorage. In order to derive the water coverage in each image, classified images were recoded to remove all the terrestrial features to generate water-only images. Longitude/latitudes of water pixels in each water-only image were stored in an ASCII table. All the water-only images were then converted to vector data using ArcGIS, a GIS software package offered by Environmental Sciences Research Institute (ESRI). All the shorelines derived were stored as polygons in shape files. Based on the observed at Anchorage, each image at different time period was assigned a stage as indicated in Table 1. During flood it may take the high water (including the tidal bore observed in the Turnagain Arm) 1-2 hours to travel from the Anchorage area and reach all the way to the farthermost ends of the inlet [2]. This factor will be taken into account only in a follow up study when fully incorporating the model dynamics with the remote sensing data. In the preliminary study here, this delay is not crucial. TABLE 1. Acquisition Data/Time (Greenwich Mean Time) , 20:47:19 pm , 20:59:56 pm , 21:04:53 pm , 20:57:20 pm , 20:56:06 pm , 21:01:49 pm REMOTE SENSING DATA USED IN THE STUDY Sea Level (Anchorage) (meters) Tidal Stage One hour after maximum Two hours before maximum Three hours after minimum One hour after minimum One hour before maximum One hour after maximum

3 III. RESULTS Fig. 3 is an example of an enhanced Landsat ETM+ image. This image was acquired on May 20, 2002, 20:56:06pm (Greenwich Mean Time), which was about one hour before the maximum water level is observed; during this time only a small portion of the mudflats is exposed (pink area at the edge of the water-covered blue area). It demonstrates the difficulty of discriminating between water and land in this complextopography region and implies the capability of Landsat dataset in deriving water coverage as a support for flood assessment. Fig. 4 shows water coverage during almost a full tidal cycle from one hour after minimum (dark blue) to one hour before maximum (red). The change of water topography across different tidal stages can be clearly observed from the map. Sea level may change along each shoreline, as it may take 1-2 hours for the flood tide to reach the farthermost end of the two arms [2, 6]. In the figure, the shoreline extends and the water coverage expands during the tidal flood when water level is rising. Near Anchorage we assumed that spatial variations of at the time of each image are small, so we matched all the shoreline data with their Anchorage water level to produce the topography. Two topography sections at o W and o W (their location indicated by vertical red lines in Fig. 4) are shown in Fig. 5. Note for example, that water coverage across the Knik Arm shrinks from over 10km during high tide to less than 3km during low tide (Fig, 5a), and farther west the two Arms are connected only when water level is above 1.25m (Fig. 5b). There are no direct observations of the mudflats topography to compare with, but the data provided by the satellite remote sensing analysis seem more accurate than the limited topographic data that was previously available when the model was first constructed. Figure 4. Water coverage during different tidal stages based on satellite data. X/Y axis represents longitude/latitude. The red shoreline is not easy shown on the map because its water level is very close to the green shoreline. Figure 3. An enhanced Landsat ETM+ images of the upper Cook Inlet. Blue represents water covered areas; purple indicates wet mudflats; green, pink, and some other colors in the surrounding areas represent exposed land. Figure 5. Water topography based on the shoreline derived from Landsat TM and ETM+ images. The x axis represents the latitude in degrees and the y axis shows the at Anchorage. Connected blue crosses + indicate the channel profiles at two longitudes, (a) o W and (b) o W (see Fig. 4 for location)

4 A comparison of the shoreline during relatively high tide (1.3 m above mean water level at Anchorage), as obtained from Landsat in July in 1989 and in 2002 is demonstrated in Fig. 6. Despite the 13 years difference, the shorelines are almost identical, except small differences near the edges of the inlet where mudflats are found (at the end of the two arms and south of Anchorage). The strong tidal velocities in the inlet (up to 5 m s -1 tidal bores, see [2]) are likely to change the morphology of the inlet by moving the sediment and changing the shoreline over long time. However, the very close proximity of the two shorelines in Fig. 6 indicates that the shoreline identification method is quite reliable. IV. DISCUSSIONS AND CONCLUSION This study shows that satellite remote sensing imagery can support mapping, assessment, and prediction of tidal flooded regions in Cook Inlet, Alaska and in any other region where LiDAR remote sensing imagery or other land-based observations are not available. However, the limitations in spatial and temporal resolutions of Landsat data requires an integrated approach that combines available satellite data with different acquisition dates and times with data from observations and from model simulations. To demonstrate the difficulties of such a task, Fig. 7 shows two Landsat ETM+ images acquired on April 28 th, 2000 (left) and July 30 th, 2002 (right). While both images use the same Bands (Band 6 in red, Band 3 in green and Band 2 in blue) and acquired during similar water level stages, they look quite different in color and water coverage. Reasons for those differences could include colder weather during April, with possibly partly ice coverage which influences reflectance, [9] and other elements like winds, river flows etc. influencing the dynamics [2]. Therefore, careful analysis should combine the remote sensing data with other data such as surface temperatures, water and river flows, shoreline developments and possible climatic changes. Figure 7. Two Landsat ETM+ images acquired on April 28 th, 2000 (left) and July 30 th, 2002 (right) respectively. Band 6 was shown in red, Band 3 was shown in green, and Band 2 in blue in every image. Two intersections were not on the same stage of although their s were expected to be similar according to the simulation model. The lack of reliable topography data in the upper inlet to be put in the model make it difficult to predict the detailed water coverage. The new information provided by the remote sensing data would thus be extremely useful to construct a new high resolution inundation model. In the next stage of the study, more Landsat data will be collected to add additional water coverage information to the current database since more Landsat imagery will be accessed with no charge after USGS decided to release the Landsat archive after October 1, SPOT satellite image with 2.5-meter spatial resolution will be also used to validate the coastlines estimated from the Landsat data. The water topography in the upper arms of inlet can be assessed using model to account for spatially uneven s along the shorelines. ACKNOWLEDGMENT The authors thank ODU s Office of Research for funding us through the Summer Experience Enhancing Collaborative Research (SEECR award). The Cook Inlet inundation model was initially developed with support from the Mineral Management Service. T.E. is also partly funded by NSF s Climate Process Team project and by a NOAA s National Marine Fisheries Service. Figure 6. A comparison between shoreline derived from two Landsat data with similar, but separated by 13 years. REFERENCES [1] Z. Kowalik, W. Knight, T. Logan, and P. Whitmore, The tsunami of 26 December, 2004: Numerical modeling and energy considerations, Pure and Applied Geophysics, vol. 164, pp. 1-15, [2] L. Y. Oey, T. Ezer, C. Hu, and F. Muller-Karger, Baroclinic tidal flows and inundation processes in Cook Inlet, Alaska: Numerical modeling and satellite observations, Ocean Dynamics, vol. 57, pp , [3] P. D. Bates, M. S. Horritt, C. N. Smith, and D. Mason, Integrating remote sensing observations of flood hydrology and hydraulic modeling, Hydrological Processes, Vol. 11, no. 14, pp , [4] G. R. Brakenridge, B. T. Tracy, and J. C. Knox, Orbital SAR remote sensing of a river flood wave, International journal of remote sensing, vol. 19, no. 7, pp , [5] J. Bales, C. R. Wagner, K. C. Tighe, and S. Terziotti, LiDAR-derived flood-inundation maps for real-time flood mapping applications, Tar River, North Carolina, U.S. Geological Survey Scientific Investigations Report , [6] T. Ezer, R. Hobbs, and L. Y. Oey, On the movement of beluga whales in Cook Inlet, Alaska: Simulations of tidal and environmental impacts using a hydrodynamic inundation model, Oceanography, vol. 21, no. 4, pp , 2008.

5 [7] Olmanson, L. G., M. E. Bauer, and P. L. Brezonik. In press. A 20-year Landsat water 215 clarity census of Minnesota s 10,000 Lakes, Remote Sensing of Environment, vol. 112, no. 11, pp , [8] C. Hu, F. E. Muller-Karger, C. Taylor, K. L. Carder, C. Kelble, E. Johns, and C. A. Heil, Red tide detection and tracing using MODIS fluorescence data: A regional example in SW Florida coastal waters, Remote Sensing of Environment, vol. 97, no.3, pp , [9] S. G. Warren, R. E. Brandt, and P. O. Hinton, Effect of surface roughness on bidirectional reflectance of Antarctic snow, Journal of Geophysical Research, vol. 103, pp , 1998.