Flood detection and mapping of the Thailand Central plain using ...

28 downloads 21874 Views 2MB Size Report
approach consisting of a data retrieval service, flood sensor observation ... A scenario of a RADARSAT and MODIS sensor web data service for flood detection.
(This is a sample cover image for this issue. The actual cover is not yet available at this time.)

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright

Author's personal copy International Journal of Applied Earth Observation and Geoinformation 14 (2012) 245–255

Contents lists available at SciVerse ScienceDirect

International Journal of Applied Earth Observation and Geoinformation journal homepage: www.elsevier.com/locate/jag

Flood detection and mapping of the Thailand Central plain using RADARSAT and MODIS under a sensor web environment Kridsakron Auynirundronkool a , Nengcheng Chen a,∗ , Caihua Peng a , Chao Yang a , Jianya Gong a , Chaowalit Silapathong b a b

State Key Lab for Information Engineering in Surveying, Mapping and Remote Sensing (LIESMARS), Wuhan University, 129 Luoyu Road, Wuhan 430079, China Geo-Informatics and Space Technology Development Agency, 120 Chaeng Wattana Road, Lak Si, Bangkok 10210, Thailand

a r t i c l e

i n f o

Article history: Received 12 February 2011 Accepted 21 September 2011 Keywords: Sensor web Flood detection Web processing service Sensor Observation Service RADARSAT MODIS

a b s t r a c t Flooding in general is insignificant event worldwide and also in Thailand. The Central plain, the Northern plain and the northeast of Thailand are frequently flooded areas, caused by yearly monsoons. The Thai government has extra expenditure to provide disaster relief and for the restoration of flood affected structures, persons, livestock, etc. Current flood detection in real time or near real time has become a challenge in the flood emergency response. In this paper, an automatic instant time flood detection approach consisting of a data retrieval service, flood sensor observation service (SOS), flood detection web processing service (WPS) under a sensor web environment, is presented to generate dynamically real-time flood maps. A scenario of a RADARSAT and MODIS sensor web data service for flood detection cover of the Thailand Central plain is used to test the feasibility of the proposed framework. MODIS data are used to overview the wide area, while RADARSAT data are used to classify the flood area. The proposed framework using the transactional web coverage service (WCS-T) for instant flood detection processes dynamic real-time remote sensing observations and generates instant flood maps. The results show that the proposed approach is feasible for automatic instant flood detection. © 2011 Elsevier B.V. All rights reserved.

1. Introduction A flood is an overflow of an expanse of water that submerges land. Flood directives define a flood as a temporary covering by water of land not normally covered by water. Floods can occur in rivers, when flow exceeds the capacity of the river channel, particularly at bends or meanders. Flood areas can be classified from synthetic aperture radar (SAR) images. It was found (Flood-prone areas, 2010) that SAR image analysis could be used to detect flooded areas reasonably well and the results were well matched in the areas near rivers and in cities where tall trees and buildings dominate the landscapes. In light of the distinct backscatter responses in the SAR data, the inundated area can be delineated using SAR images acquired during the flood period. Flooded areas have a relatively smooth surface with respect to the wavelength of SAR. Thus, they act as specular reflectors, directing the microwave energy away from the satellite and create dark tones on the SAR images. Other areas such as grassland, residential land, paddy field, bare soil, and crop land show a relatively rough surface compared with the wavelength. They reflect more energy back to the satellite, creating the bright tones in the images. So they have been classified

∗ Corresponding author. Tel.: +86 27 68778586; fax: +86 27 68778229. E-mail address: [email protected] (N. Chen). 0303-2434/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.jag.2011.09.017

by commercial programs such as Envi, PCI Geomaticas, and Erdas by manual work and this takes time to classify flood area. Flood detection and monitoring can provide warning of a possible flood before the damage is done or monitoring a direction of flood and flood damage. Flood detectors work by detecting water or moisture that should not normally be present in an area. These detectors work especially well in areas prone to leaks or floods. In this way, SAR imagery can provide real-time flood observations and can estimate the water surface fraction over two control sites. So that microwave remote sensing can provide such information using JERS-1 SAR and LandsatTM imagery data, the actual open flooded area, flood affected area and flood movement could be successfully extracted (Honda, AIT). They used 1 day for a preliminary report after obtaining RADARSAT data, 2 days for a detailed report, and 4 days for a detailed report after acquiring airborne SAR data (Wang et al., 2002). Flood maps, often referred to as flood risk or hazard maps, present in a graphical format the areas of land or property that have historically been flooded, or that are considered to be at risk of flooding. The maps can display a range of parameters and different types of information, such as flows, water levels, depths, etc. Flood maps can provide useful information to planning authorities, a number of other bodies or organizations, and indeed the public, in the performance of a range of functions or for information purposes. However, the fundamental characteristic recorded on a radar image

Author's personal copy 246

K. Auynirundronkool et al. / International Journal of Applied Earth Observation and Geoinformation 14 (2012) 245–255

is the backscattering coefficient, which may vary from surface to surface. The strength of the signal returned from the surface is influenced by the combined system and ground parameters, including the average surface roughness and soil dielectric properties. Horizontal smooth surfaces, such as water bodies, reflect nearly all incident radiation away from the sensor and the weak return signal is represented by dark tonality on radar images with the result that standing water areas are easily recognizable. This specular refection can be decreased by bad weather conditions and/or the presence of vegetation, roughening the surface and making the detection of flooded areas more difficult (Laugier et al., 1997). However, SAR data are subject to speckle, a multiplicative random noise that considerably reduces the interpretability of the images and limits classification techniques, and SAR images have to be filtered in order to increase the signal-to-noise ratio. The enhanced Lee adaptive filter is an efficient tool for reducing SAR image speckle by removing high-frequency noise while preserving edges or sharp features (Bricio et al., 2008). RADARSAT-1 can use object-oriented analysis data in comparison with a pixel-based approach. Additionally, the flexible fuzzy rule base for data and information fusion helps to build up an efficient strategy to derive synergy from SAR and optical data. The hierarchical network provides object features, such as neighborhood relations, together with geometric features and can distinguish between small rivers and flooded areas (Kuehn et al., 2002). In order to correctly highlight the flooded areas by mean values of optical imagery, the best combination of features includes the coherence difference between the acquisitions before and during the flooding, the backscatter intensity in the reference period and the coherence computed during the flood. The twice daily MODIS satellite data (Terra and Aqua) can support this. Therefore, we can use RADARSAT and optical data for flood monitoring and mapping. MODIS data were used for detecting water related surfaces employing the normalized difference vegetation index (NDVI) (Rogers and Kearney, 2004). The main reason for using NDWI is that short-wave infrared (SWIR) is highly sensitive to moisture content in the soil and the vegetation canopy. A number of studies have been conducted using the spectroscopic characterization of SWIR to detect water content (Gao, 1996; Jackson et al., 2004; McFeeters, 1996; Rogers and Kearney, 2004; Tong et al., 2004). Sensor networks are an important means for capturing real-time information about a nearly unlimited number of environmental phenomena. Properties observed by sensor networks range from weather data, hydrological information such as water level measurements or air quality data, to positioning information of objects or persons. In order to make use of such a large range of very heterogeneously structured information, their integration into application systems becomes very important. As spatial data infrastructures (SDI) are well established for building distributed applications in the geospatial domain, it is of special importance to link them with sensors and sensor data. This is the goal of the OGC sensor web enablement (SWE) architecture, which provides an interoperable framework of standards for accessing and encoding sensor data, describing sensor metadata, and issuing alerts based on measured sensor data and controlling sensors. Comparison of research from another technology architecture among standalone system, web GIS based system and sensor web based system a solution for a flooded detection intelligence system three different system will be developed for timely and appropriated. Standalone system. A standalone system refers to an application running on a desktop environment such as Windows or Linux platforms. For this system standalone its old system have a complicated with late, need the time, used human resource, and along with trying for dissemination to the public This is not an easy task and it takes a lot of time and will be only work with local data.

Web GIS based system. Web GIS based system technology this application technology recently was occurred because capability of network were develop and its applications have been accepted about immediately to access to spatial data and processing as well as advance mapping and analysis over the network is turning to common, can access from any location, accessible by multiple users. Sensor web based service system. Sensor web based service is a newest last generation its can defined as a web of interconnected heterogeneous sensors that are interoperable, dynamic, intelligent, flexible and scalable. Its can be accessibility of dynamic geo-information, acquired almost real-time from some kind of sensor. Its method parallels the projects where various peoples collaborate to create and improve and update maps on the web which real-time rather. Decision makers in an emergency response situation (e.g., floods and droughts) need to have rapid access to the existing data sets, the ability to request and process data including the specifics of the emergency, and tools to rapidly integrate the various information sources into a basis for decisions. In this paper, we consider the problem of flood monitoring using satellite remote sensing data, in situ data and results of simulations. The problem of flood monitoring itself consumes data from many heterogeneous data sources such as remote sensing satellites (we are using data from the ASAR, MODIS and MERIS sensors), and in situ observations (water levels, temperature, humidity, etc.). Flood prediction is adding to the complexity of the physical simulation of the task. A sensor web (Botts and Robin, 2007a,b) is a type of sensor network or geographic information system (GIS) that is especially well suited for environmental monitoring. Sensor webs (Delin and Small, 2009) have been set up in harsh environments (including deserts, mountain snow packs, and Antarctica) for the purposes of environmental science and have also proved valuable in urban search and rescue and infrastructure protection. SWE is the technology to enable the realization of sensor webs, consisting of a suite of standards from OGC, which provides interoperability interfaces and metadata encodings that enable real-time integration of heterogeneous sensor webs into the information infrastructure. The SWE standards suite has two major subgroups: the SWE information model dealing with data formats and the service model dealing with the interfaces of web services. The SWE information model comprises four different standards: SWE common (while SWE common is currently still a part of the SensorML standard. In Version 1.0, SWE Common will be defined as a separate specification in later versions) (Botts and Robin, 2007a); observations and measurements (O&M) (Cox, 2007a,b); sensor model language (SensorML) (Botts and Robin, 2007b); transducer markup language (TML) (Havens, 2007). SWE provides four major web services: the sensor planning service (SPS) (Simonis, 2007); the web notification service (WNS) (Simonis and Wytzisk, 2003); the sensor alert service (SAS); the sensor observation service (SOS) (Na and Priest, 2007). When these four information models and four services work together, it is possible to make all types of sensors, transducers, and sensor data repositories discoverable, accessible, and useable in real time or near real time. However, Thailand bump against usually by monsoon season every years that is have influence with agriculture cause damage which flooding is a one of the issues as important we have realized to flood detection management. So that sensor web environment is an alternative to management and processing analysis on timely. Currently, flood detection is still suffering from a long-term information delay, and this also happens in the Chao Phraya basin flood area detection, which is a season-repeat flood plain in central Thailand. In traditional flood detection, it takes a long time to import the data and process data by personnel; data acquisition from satellites through ground receiving stations and the transfer of imagery data by media (e.g., CD-Rom, DVD-Rom or Media

Author's personal copy K. Auynirundronkool et al. / International Journal of Applied Earth Observation and Geoinformation 14 (2012) 245–255

resourced) are large tasks. The current process uses a commercial program (e.g., PCI-Geometrical, Envi, Erdas Imagine and ArcGIS) to process the image data and analyze the result, then deploys in picture format (e.g., JPG, BMP and TIF) post online for visualization or overlay in other clients. Apparently, this method requires large amounts of manual work in the processing flow and its timeliness does not fit the disaster relief demand. The aim of this paper is to achieve dynamical flood monitoring and mapping for decision support under a sensor web environment. How to organize the SOS, WCS, WPS and WFS as the framework to achieve real-time or near real-time flood detection has become a challenge in flood emergency response under a sensor web environment. This paper tries to explain how to design, implement and use the proposed framework to process dynamic real-time remote sensing observations and generate an instant flood map. In the remainder of this paper, we present a coupling of MODIS with RADARSAT using the SWE and OGC transactional sensor data services to detect flood areas and generate a live flood map. The architecture of the flood sensor web is presented in Section 2. The design and implementation of the flood classification based on WPS specification are described in Section 2.5. The implementation of the proposed framework under sensor web design, instantiation, and execution is discussed in Section 3. Section 4 discusses experiments to evaluate the feasibility of the Thailand flood detection for MODIS and RADARSAT sensor data, and then the benefits of the proposed method are discussed. Finally, Section 5 summarizes the conclusions and discusses the next steps that are needed. 2. Methods 2.1. Architecture and components As shown in Fig. 1, the architecture of the automatic flood detection system under the sensor web environment consists of the following three components: data retrieval service based on flood SOS, WPS-based flood detection and mapping service, and flood portal. 2.1.1. Flood SOS components Standard service interface for requesting, filtering, and retrieving observations and sensor system information. This is the intermediary between a client and an observation repository; the flood SOS operations section is intended to outline a typical operational context in which an SOS is expected to be used with a flood area. However, in this paper we used RADARSAT and MODIS data for flood monitoring. The flood SOS is the real-time or near real-time data that are encoded using the O&M standard. 2.1.2. Flood detection and mapping service components WPS is a generic interface with no specific processes; the operation standards define a generic mechanism as one “that can be used

247

to describe and web-enable any sort of geospatial process”, including a general mechanism for describing input data and outputs. Such mechanisms support direct input data or an indirect reference of a data source, enabling collaboration with other OGC standards dedicated to data delivery. The Open Geospatial Consortium Web Coverage Service Interface Standard (WCS) provides interfaces allowing requests for raw data values to GIS clients via the web. WCS describes discovery, query, or data transformation operations. The client generates the request and posts it to a web feature server using HTTP. WCS provides access to potentially detailed and rich information of flood monitoring and mapping, in forms that are useful for decision support. While accessing a WCS server, which is accomplished by a service description file, the URL of coverage server, and the name of the coverage. It is important that there should be no spaces or other content before the element. The WCS interfaces and operations are connected to the outside by the WPS-based flood detection service: the “WCS-Transaction” operation in the WCS service. Built-in WCS service is deployed in the middleware and many data service conversions are implemented in the middleware. 2.1.3. Flood portal A Geo-Sensor portal via the client, achieving compliance with OGC specifications, provides an architecture that is flexible and extendable enough so that all kinds of OGC web services (OWS) can be accessed and the queried data can be visualized under sensor web, such as SOS providing a variety of data. Geo-Sensor is a framework based on SOA prototype Sensor Observation Services, which addresses the needs of application developers. It offers developers a customizable and extendable system, which supplies a reusable design. 2.2. Interaction model SOS is the access to obtain both RADARSAT and MODIS satellite data and metadata. It allows the definition of filters and criteria (image of specified flood area) to select the data of interest but also operations for inserting sensors and observations. Usually SOS instances return sensor data encoded to O&M and sensor metadata encoded in SensorML. Web-based geo-processing to analyze realtime sensor data will use O&M to encode data to provide to web processing service (flood detection and flood mapping service) and then use web coverage service (WCS) for providing coverage data (vector and/or raster plus attribute information) to represent in public the result in the GeoSensor. Flood SOS components will use O&M to encode data and provide for Web Processing Service (flood detection and flood mapping service) and then use web coverage service (WCS) to publish the result on the GeoSensor.

Fig. 1. The architecture of the automatic flood detection service.

Author's personal copy 248

K. Auynirundronkool et al. / International Journal of Applied Earth Observation and Geoinformation 14 (2012) 245–255

2.2.1. InsertObservation InsertObservation is an operation that requests an SOS instance to perform an “insert” operation. Submitted data is in the form of an O&M document as well as the “InsertID” obtained when the RegisterSensor operation took place. The O&M document used to submit the information must follow the template specified when the sensor registration occurred. Ideally, it should be possible to insert data from more than one sensor at a time, or all sensors, but this may not be possible since disparate sensors could take readings at different intervals. An effective solution is to have the capability to insert all sensor readings as a default case, and a subset as the special case. 2.2.2. GetObservation This operation returns an O&M document containing data values, which correspond to sensors involved in a particular “observation offering”. The O&M document contains the observed data values encoded along with metadata describing the meaning of these values (units of measure, standards, url references, etc.). The observation can be of a single sensor, or multiple, depending on the user’s request. Also, the identifier parameters used to specify which sensor(s) to observe are obtained from the capabilities document returned by the initial call to GetCapabilities. The user can specify the time as well as the sensors involved in the request. In the event that the SOS server does not have the requested data, the GetObservation request will be forwarded to the provider in the form of an SOS-generated RequestObservation invocation. Data returned is of the form of one or more O&M documents, depending on the nature of the data requested. The nature of the request may be for earlier data that has already been broadcast or a collection of data that is different to one previously specified for observation. 2.2.3. WCS-T The extension of the WCS standard specifies an additional transaction operation (WCS-T) that optionally provides a transparent, standard mechanism for the transaction of geospatial data in the web environment. The transaction operation allows clients to add, modify, and delete grid coverages that are available from a WCS server. The transaction operation request references the new or modified coverage data, including all needed coverage description metadata. The transaction operation has two modes of operation, synchronous and asynchronous. Chen et al. (2009a,b) found that a service-oriented framework could integrate and assimilate sensor observations and measurements under a multi-purpose SOS in combination with other standard services-CSW, the WFS-T and the WCS-T. In this paper, WCS-T is used to connect the flood SOS and the WPS-based flood detection service. 2.3. Information model 2.3.1. O&M model An O&M is an action with a result that has a value describing some phenomenon. The observation is modeled as a feature within the context of the general feature model. An observation comprises the following key properties: featureOfInterest, observedProperty, procedure, result. An observation feature binds a result to a feature of interest, upon which the observation was made. The observed property is a property of the feature of interest. An observation uses a procedure to determine the value of the result, which may involve a sensor or observer, analytical procedure, simulation or other numerical process. The observation pattern and feature are primarily useful for capturing metadata associated with the estimation of feature properties, which is important particularly when error in this estimate is of interest. An observation results in an estimate of the value of a property of the feature of interest. Observation values may have many data types, including primitive types like category or measure, but also more complex types such as time,

location and geometry. Complex results are obtained when the observed property requires multiple components for its encoding. In this paper, we use O&M for connecting to the flood SOS database. 2.3.2. Coverage model Web coverage service interface standard (WCS) defines a standard interface and operations that enable interoperable access to geospatial “coverages”. The term “grid coverages” typically refers to content such as satellite images, digital aerial photos, digital elevation data, and other phenomena represented by values at each measurement point. 2.4. Implementation of flood SOS The OGC sensor observation service (SOS) implementation specification defines a web service interface for requesting, filtering, and retrieving observations and sensor system information. Observations may be from in situ sensors or dynamic sensors. SOS has three mandatory “core” operations: GetCapabilities, DescribeSensor, and GetObservation. The GetCapabilities operation provides the means to access SOS service metadata. Several optional, nonmandatory operations have also been defined. The DescribeSensor operation retrieves detailed information about the sensors and processes generating those measurements. The GetObservation operation provides access to sensor observations and measurement data via a spatial-temporal query that can be filtered by phenomena. There are two operations to support transactions, RegisterSensor and InsertObservation, and six enhanced operations, including GetResult, GetFeatureOfInterest, GetFeatureOfInterestTime, DescribeFeatureOfInterest, DescribeObservationType, and DescribeResultModel. Used in conjunction with other OGC specifications, the SOS provides a broad range of interoperable capability for discovering, binding to and interrogating individual sensors, sensor platforms, or networked constellations of sensors in realtime, archived or simulated environments. 2.4.1. MODIS and RADARSAT sensor description in SensorML The MODIS instrument has a viewing swath width of 2330 km and views the entire surface of the Earth every 1–2 days operating on the Terra spacecraft. Its detectors measure 36 spectral bands. The spatial resolution of MODIS (pixel size at nadir) is 250 m for channels 1 and 2 (0.6–0.9 ␮m), 500 m for channels 3–7 (0.4–2.1 ␮m) and 1000 m for channels 8–36 (0.4–14.4 ␮m). MODIS flood SOS is implemented using 250 m MODIS surface reflectance data. MODIS data is the primary data in this monitoring work. For the high temporal resolution, MODIS data are useful in monitoring floods changing day by day. MODIS flood SOS is based on the reflectance characteristics of floods. In this paper, one example of the near-infrared band’s reflectance is used to identify flooding in the Thailand Central plain. We use a SensorML document named TERRA-MODIS-Band2 (NIR).xml to describe the metadata of Modis-Band2, such as inputs referring to Modis-band2 input, outputs referring to Modis-band2 image gray value, represented in Fig. 2. SAR imagery acquired by the RADARSAT satellites can be used for flood monitoring due to the cloud penetrating capability of SAR. Therefore, the use of SAR imagery acquired at appropriate times would greatly help in the compilation and timely updating of flooding maps. The metadata of RADARSAT system encoded by a SensorML document is shown in Fig. 3. MODIS and RADARSAT data are encoded in O&M. MODIS and RADARSAT flood SOS are especially useful. These provide consistent and independently verifiable information, such as flood area, flood perimeter, and the location. It is an important and economical enhancement to flood warning and flood response. The values of MODIS and RADARSAT flood

Author's personal copy K. Auynirundronkool et al. / International Journal of Applied Earth Observation and Geoinformation 14 (2012) 245–255

249

Fig. 2. MODIS Band2 sensorML instance file.

SOS encoded in O&M have six fields, such as MODIS in Fig. 4: the time is between “2009-10-10T12:00:00.000 + 08:00” and “2009-10-10T 12:33:03.000 + 08:00”. The observation has six items, separated by commas. For the first record “200910-10T12:00:00.000 + 08:00, Bangkok, 661554.820, 1513525.481, 281250.00000000, 3121.32034356”, the time of measurement is “2009-10-10T12:00:00.000 + 08:00”, the location of feature is Bangkok, the center coordinates of flood extent is 661554.820 m, 1513525.481 m, the flood area is 281250.00000000, the flood perimeter is 3121.32034356 m. Fig. 4 is the result of the MODIS flood SOS. The RADARSAT flood extent is extracted from the RADARSAT image. It is shown in the records in Fig. 5. For a record “2009-10-10T12:00:01.000 + 08:00, Bangkok, 2 767.71388711, 345292.78466797, 505268.759, 1367875.000”, the time of measurement is “2009-10-10T12:00:01.000 + 08:00”, the location of the feature is Bangkok, the flood perimeter is 2767.71388711 m, the flood area is 345292.78466797, the center coordinates of flood extent are 505268.759 m and 67875.000 m, as represented in Fig. 5.

2.4.2. Insert sensor description and O&M data into postGreSQLGIS database using the SOS “InsertObservation” operation A Register Sensor operation request contains a SensorML description as parameter and the Insert Observation operation request needs an O&M: observation as parameter. An Observation from the O&M model is an event whose result is an estimate of the value of some property of a feature of interest, obtained using a specified procedure. O&M describes a generic model for metadata associated with property value estimation. In our test, we use the “Register Sensor” operation to register TERRA-MODIS-Band2, inserting flood observations into SOS using the “Insert Observation” operation. SOS communicates with the PostGreSQL database PostGis to insert metadata and observation information by Register Sensor and Insert Observation operations. The Metadata information describing the sensor hardware is encoded in SensorML. If it is registered successfully, it will return the sensorID registered to the SWE client, and will use it to retrieve the sensor description to determine if the SOS service is capable of fulfilling an Observational request. The observation data encoded into the O&M specification

Fig. 3. RADARSAT sensorML instance file.

Author's personal copy 250

K. Auynirundronkool et al. / International Journal of Applied Earth Observation and Geoinformation 14 (2012) 245–255

Fig. 4. MODIS flood SOS encoded in O&M.

Fig. 5. RADARSAT flood SOS encoded in O&M.

Author's personal copy K. Auynirundronkool et al. / International Journal of Applied Earth Observation and Geoinformation 14 (2012) 245–255

251

Fig. 6. Description of flood detection process in Flood Detection WPS.

is inserted to PostGis by an insert Observation operation and observationID will be returned to the SWE client. 2.4.3. DescribeSensor and GetObservation operation of MODIS SOS and RADARSAT SOS We can use the sensorID to obtain the sensor description by a DescribeSensor operation. Observation data can be acquired by a GetObservation operation. The results of MODIS and RADARSAT data are encoded in O&M, which correspond to the data inserted by the insertObservation operation. According to the specification of SOS, we define the WSDL document for our Sensor ObservationService. For our solution, we use O&M to encode flood detection information so that each counterpart in communication can understand the semantics of SOS. Besides, with the help of WSDL based on XML, SOS is described in a standard way, which is convenient for service integration. WSDL is a specification for describing web services and how to access them. We have defined the WSDL for SOS, WCS and WPS, and set up services such as MODIS SOS, RADARSAT SOS, WCS-T and WPS-based flood detection services. 2.5. Implementation of flood WPS The implemented GR flood detection WPS is deployed in http://host:port/wps/ and defines three mandatory operations: GetCapabilities, Describe Process, and Execute. The GetCapabilities operation of GR Flood Detection WPS gives the clients service level metadata. The GR Flood Detection WPS Capabilities response is an XML document containing the elements Service Identification, Service Provider, Operations metadata, and Process Offerings. The Process Offerings contains the brief description of the processes offered by this server

and process Version 2. In this paper, the Identifier of Process is “cn.edu.whu.swe.wps.geosensor.GR Flood Detection” and the process version of flood detection process is 2. Clients can obtain the XML based service metadata documents for the flood detection process and determine the specification version for client server interactions from the results of a GetCapabilities request (http://host:port/ wps/WebProcessingService?Request=GetCapabilities&Service= WPS). The “DescribeProcess” operation of Flood Detection WPS (http://host:port/wps/WebProcessingService?request=Describe Process&Service=WPS&Identifier=cn.edu, whu.swe.wps. geosensor.GR Flood Detection) gives the clients a detailed description of the parameters of the inputs and outputs of the “cn.edu.whu.swe.wps.geosensor. GR Flood Detection” process. It is shown in the records in Fig. 6. The ProcessDescription contains one or more process descriptions for the request processes identifier. Each description includes brief information on the process, metadata and description of the input and output parameters. Two kinds of “Data Inputs” parameters and one kind of “Process Outputs” result are concerned in the description of “GR Flood Detection” process. “Data Inputs” parameters are bands required. The names of the “bands” input parameters are band1, band2. The input values are MODIS terra images, encoded in “image/tiff” mime type. The “Process Outputs” result is the raster layer in “image/tiff” mine type. Just as Fig. 7 shows, the “Execute” operation of flood detection WPS (http://host:port/wps/WebProcessingService? service=WPS&request=execute&identifier=cn.edu, whu.swe.wps. geosensor.GR Flood Detection&Datainputs = band1 = 1.tif; band2 = 2.tif) carries out the “Flood Detection” process triggered by the clients. The returning result is the raster layer whose flood area is red. With Execute, clients can execute the flood detection process

Fig. 7. The execute request for the extraction of flooding area in Flood Detection WPS.

Author's personal copy 252

K. Auynirundronkool et al. / International Journal of Applied Earth Observation and Geoinformation 14 (2012) 245–255

Fig. 8. Study area in the central part of Thailand (Central Plain).

implemented by the server to retrieve the extraction of the flooded area in Thailand. 3. Thailand flood detection and mapping experiment and results 3.1. Study area As shown in Fig. 8, the central part of Thailand (Central Plain) is the lower part of Ping River, Wang River, Yom River, Nan River, Cho Phraya, Pasak River and the Chin River, which drain into the gulf of Thailand. Geographically, it located between latitude 13◦ 25 and 17◦ 30 N and longitude 99◦ 20 and 101◦ 30 E and covers an area of approximately 92,306 sq. km. For administration, it is divided into 22 Provinces. Its topography is composed of plains and mountains. The rivers lie in the center of the region. The highest mountain is approximately 1500 m above mean sea level. The climate is influenced by southwest monsoons and northeast monsoons; the average temperature is 27.6 ◦ C. Average precipitation during the year is approximately 1208.8 mm. 3.2. Framework This paper we focused on RADARSAT and MODIS data imagery with sensor web environment in the near-real time flood detection to support decisions management as well as helping and find a trend of flood areas at central part of Thailand. 3.3. MODIS flood classification service The MODIS data set is an extensive program using sensors aboard two satellites and each satellite provides complete daily coverage of the Earth. The data are different in resolutions;

spectral, spatial and temporal. The MODIS sensor is carried on both the Terra and Aqua satellites; it is generally possible to obtain images in the morning (Terra) and the afternoon (Aqua) for any particular location. Night time data are also available in the thermal range of the spectrum. The time of day should be considered when ordering a scene for a specific day. The MODIS instrument provides high radiometric sensitivity (12 bit) in 36 spectral bands ranging in wavelength from 0.4 ␮m to 14.4 ␮m. The responses are custom tailored to the individual needs of the user community and provide exceptionally low out-of-band response. Two bands are imaged at a nominal resolution of 250 m at the nadir, with five bands at 500 m, and the remaining 29 bands at 1 km. The raw information can be obtained via the website of the Land Process Distributed Active Archive Center (https://lpdaac.usgs.gov/lpdaac/products/modis overview).

3.4. RADARSAT flood classification service The RADARSAT data products have many modes. The SAR instrument consists of a radar transmitter, a radar receiver and a data downlink transmitter. The radar transmitter and receiver operate through an electrically steerable antenna that directs the transmitted energy in a narrow beam normal to the satellite track. The elevation angle and the elevation profile of the beam (beam positions) can be adjusted so that the beam intercepts the Earth’s surface over the desired range of incidence angles. The ability to choose the beam and position is important since image characteristics vary with the incident angle associated with each beam. In addition, by varying the characteristics of the transmitted pulses and the receiver timing, different resolution and coverage can be achieved. The beam modes are each characterized by a specific beam elevation angle and profile.

Author's personal copy K. Auynirundronkool et al. / International Journal of Applied Earth Observation and Geoinformation 14 (2012) 245–255

253

Table 1 The visualization of data contains 10 data set; each dataset cover different central part of Thailand. Data set

Time (s)

BBOX (◦ )

Size (MB)

1 2 3 4 5 6 7 8 9 10

0.35 0.67 5.3 11.9 13.8 17.5 48.3 561 786 963

(10.16N,99.33E,13.60N,101.20E) (11.16N,99.33E,13.73N,101.20E) (13.16N,100.33E,14.50N,103.0E) (13.16N,100.33E,15.73N,103.0E) (13.16N,100.33E,17.20N,103.0E) (11.16N,100.33E,17.20N,105.30E) (11.16N,100.20E,17.20N,107.20E) (11.16N,101.20E,17.20N,101.20E) (10.58N,102.75E,17.20N,106.5E) (10.58N,103.75E,17.20N,107.6)

0.2 1.2 11.2 50.6 66.6 86.3 100.6 500.0 1456 3223

Fig. 9. The visualization results of MODIS Flood detection SOS.

For any given beam mode, the same beam angle and profile are used for both transmission and reception. The receiver detects the echo resulting from backscatter of the transmitted signal from the Earth’s surface. The detected signal is then digitized and encoded prior to transmission to the on-ground data reception facility. Data transmission may occur in real time as the data is collected, or the data may be stored on the on-board tape recorders (OBR) for later transmission. 3.4.1. Results The Geo-Sensor displays the final flood classified map by GeoServer in WCS and requests the data source from the WCS server. The workflow has the capability of outputting the flood area in vector format. The Geo-Sensor is a client that can visualize the local data source and OGC standard web service or web sources such as WMS, WCS, etc. The visualization results of MODIS (shown in Fig. 9) and RADARSAT (shown in Fig. 10) data are shown on Geo-Sensor. The red and pink areas are representative of flood areas that are encoded in O&M. The performance of sensor web environment and GetObservation application for flood detection. The visualization of data contains 10 data set; each dataset cover different central part of Thailand represent in Table 1 and Fig. 11, the total of data size of the central part in Thailand at a particular moment is about 3223 MB. Respond time of SOS contains 10 data set; each dataset cover different part as shown in Table 2 and Fig. 12, SOS contains 10 data

Fig. 10. The visualization results of RADARSAT Flood Detection SOS.

Fig. 11. The visualization of data contains 10 data set; each dataset cover different central part of Thailand. Table 2 The SOS contains 10 data set; each dataset cover different central part of Thailand. Data set

Time (s)

BBOX (◦ )

Size (MB)

1 2 3 4 5 6 7 8 9 10

0.50 0.76 6.4 13.5 15.6 18.5 50.3 641 856 1053

(10.16N,99.33E,13.60N,101.20E) (11.16N,99.33E,13.73N,101.20E) (13.16N,100.33E,14.50N,103.0E) (13.16N,100.33E,15.73N,103.0E) (13.16N,100.33E,17.20N,103.0E) (11.16N,100.33E,17.20N,105.30E) (11.16N,100.20E,17.20N,107.20E) (11.16N,101.20E,17.20N,101.20E) (10.58N,102.75E,17.20N,106.5E) (10.58N,103.75E,17.20N,107.6)

0.2 1.2 11.2 50.6 66.6 86.3 100.6 500.0 1456 3223

set; each dataset cover different part of Thailand, and the total data size of the central part in Thailand at a particular moment is about 3223 MB (Table 3). Response time of WPS based on flood detection as shown in Table 1 and Fig. 13, WPS contains 10 data set; each dataset cover different central part of Thailand, and the total data size of the central

Fig. 12. The SOS contains 10 data set; each dataset cover different central part of Thailand.

Author's personal copy 254

K. Auynirundronkool et al. / International Journal of Applied Earth Observation and Geoinformation 14 (2012) 245–255

Table 3 WPS contains 10 data set; each dataset cover different central part of Thailand. Data set

Time (s)

BBOX (◦ )

Size (MB)

1 2 3 4 5 6 7 8 9 10

1.20 2.68 9.6 20.5 32.3 40.5 100.3 378 1203 5563

(10.16N,99.33E,13.60N,101.20E) (11.16N,99.33E,13.73N,101.20E) (13.16N,100.33E,14.50N,103.0E) (13.16N,100.33E,15.73N,103.0E) (13.16N,100.33E,17.20N,103.0E) (11.16N,100.33E,17.20N,105.30E) (11.16N,100.20E,17.20N,107.20E) (11.16N,101.20E,17.20N,101.20E) (10.58N,102.75E,17.20N,106.5E) (10.58N,103.75E,17.20N,107.6)

0.2 1.2 11.2 50.6 66.6 86.3 100.6 500.0 1456 3223

Fig. 13. WPS contains 10 data set; each dataset cover different central part of Thailand.

Fig. 14. Time summarize for all processing.

part in Thailand at a particular moment is about 3223 MB (Fig. 14 and Table 4). The total of the data summarize for response time of data visual, response time of SOS GetObservation and response time of WPS based on flood detection.

Table 4 Time Summarize for all processing (V = RESPONSE TIME OF Data Visual; O = RESPONSE TIME OF SOS GetObservation; P = RESPONSE TIME OF WPS based on flood detection). Data set

1 2 3 4 5 6 7 8 9 10

Time (s)

Size (MB)

V (s)

O (s)

P (s)

1.20 2.68 9.6 20.5 32.3 40.5 100.3 378 1203 5563

0.35 0.67 5.3 11.9 13.8 17.5 48.3 561 786 963

1.20 2.68 9.6 20.5 32.3 40.5 100.3 378 1203 5563

0.2 1.2 11.2 50.6 66.6 86.3 100.6 500.0 1456 3223

4. Discussion NOAA AVHRR data has been successfully applied in monitoring flood processes for a long time because of its property of high temporal resolution. Also, the deployment of LANDSAT MSS and TM to estimate the land use type in inundated areas can also give a good result. However, the visible band and infrared band cannot function effectively during the monsoon seasons due to the cloud cover. SAR data are the priority data for flood monitoring. So RADARSAT images were the fundamental source of information and the MODIS data were used to overview the wider area. RADARSAT was used to classify flood areas because its accuracy is higher than MODIS and RADARSAT can be used in any weather conditions. Therefore, MODIS and RADARSAT imagery can be used to manage and classify flooding in an area. Sensor web technology can be useful and provides near real-time monitoring; it also does not require a large number of personnel in this process. We think this system will be superior to web service, as this architecture has flexibility. This paper proposes an automatic procedure for flood mapping service technology with the support of OGC web service, consisting of SOS, WPS, and WCS. The proposed methods have the benefits of being automatic and near real time and are able to monitor and map floods under sensor web. 4.1. Instant time update of the proposed method It is found that the current process uses a commercial program media, which requires time consuming processes or procedures for its work and various programs have limitations. After the process of analysis, to expand the availability to the public by means of a map and the export of information in the form of images and then to the website takes a long time. The proposed method is nearer real time than the existing flood area detection method. The traditional methods of flood monitoring and mapping are by manual operations. Temporal and spatial information are important when monitoring flood areas in Thailand, with real-time updated information speeding up the understanding of the dynamics of the environment in Thailand. RADARSAT provides high-resolution satellite images of the area of interest, which could allow clearer analysis, while MODIS data confirms the extent of the flood area. Thus, near real-time sensor processing and observation services for flooded area detection can be implemented using RADARSAT and MODIS data under sensor web. The proposed method could perform near real-time generalization of RADARSAT and MODIS data for flood monitoring and mapping. The different data sets can be in integrated because only RADARSAT dataset is not enough which MODIS its can be support 2 time per day. However, MODIS data have qualification of some weather which it is main problem. So that we use to sensor for flood detection in Thailand. RADARSAT imagery data can be distinguish flood area likable more MODIS which it necessary of flood detection in this paper. 4.2. Automation of the proposed method It was found that the current data acquisition is from satellites via ground receiver stations and then transferred by media. So every day, personnel will have to import the data and analyze data. The proposed method allows automatic updates. The existing temporal and spatial data update method requires considerable manual involvement in the individual steps. To implement automatic temporal and spatial data updates is a major challenge. The technology chains are MODIS, RADARSAT WPS, SOS, and WCS-T. The proposed method can be automatic as a workflow, invoked as a web service and executed dynamically to automatically achieve MODIS and

Author's personal copy K. Auynirundronkool et al. / International Journal of Applied Earth Observation and Geoinformation 14 (2012) 245–255

RADARSAT observational data acquisition, flood area detection processing, and live sensor map generation, unlike the data update by manual methods. Thus, an automatic information updating service can be implemented by the geo-processing workflow. 4.3. Further data set support Although this paper we use only RADARSAT and MODIS. However, we try to presented solution made the system much more flexible and useful as a tool in the research activity. So that a various observation service dataset can be plug-in and integration on this system. 5. Conclusions and outlook Flood emergency response is significant in a government’s work; as the core function model, flood detection in real time or near real time under a sensor web environment is still a problematic issue. Based on the SWE technologies mentioned in the paper, it is reasonable to suggest that time series of RADARSAT and MODIS imagery can be used to quantitatively flood detection. An automatic instant in time flood detection approach consisting of flood sensor observation service (SOS), flood detection web processing service (WPS) under the sensor web environment is presented to generate dynamically real-time flood maps. In this paper we use threshold analysis methods for distinguish approach under generalized flood assumption. For the data befit to integrate and processing its depend on capability of system which will take effected with time-consuming vary on magnitude of data set. The result expected to be achieved from the project is the classification of flooded areas by the RADARSAT and MODIS imagery, with the process automated using a technology called sensor web, allowing users to process and access data and create map layouts. It is demonstrated that the proposed approach is feasible for automatic instant flood detection in Thailand’s Central plain. The next step is to study how to evaluate and improve the quality of data processing and usability by other sensors. Acknowledgements This work was supported by grants from the National Basic Research Program of China under Grant 2011CB707101, Chinese 863 program under Grant 2011AA010500, and the National Nature Science Foundation of China program under Grant 41171315, 41023001 and 41021061. We sincerely thank Dr. David A. Tait, for proofreading the manuscript.

255

References Botts, M., Robin, A., 2007a. Bringing the sensor web together. Geosciences, 46–53. Botts, M., Robin, A. (Eds.), 2007b. OpenGIS® Sensor Model Language Implementation Specification (Version 1.0.0). OGC Document Number: 07-000, Wayland, MA, USA, 180 pp. Bricio, L., Negro, V., Diez, J.J., 2008. Geometric detached breakwater indicators on the Spanish northeast coastline. Journal of Coastal Research 24 (5), 1289–1303. Chen, N., Di, L., Yu, G., Gong, J., Wei, Y., 2009a. Use of ebRIM-based CSW with sensor observation services for registry and discovery of remote sensing observations. Computers and Geosciences 35 (2), 360–372. Chen, N., Di, L., Yu, G., Min, M., 2009b. A flexible geospatial sensor observation service for diverse sensor data based on Web service. ISPRS Journal of Photogrammetry and Remote Sensing 64 (2), 234–242. Cox, S., (Ed.), 2007a. Observations and Measurements – Part 1 – Observation schema. Open Geospatial Consortium (OGC) document number: 07-02r1. Wayland, MA, 724 USA. Cox, S., (Ed.), 2007b. ObservationsandMeasurements – Part 2 – sampling features. Open Geospatial Consortium (OGC) document number: 07-002r3, Wayland, MA, 724 USA. Delin, K., Small, E., 2009. The Sensor Web: Advanced Technology for Situational Awareness. Wiley Handbook of Science and Technology for Homeland Security, John Wiley & Sons. 2010. Flood-prone Areas. Multiple Cropping Center (MCC), Faculty of Agriculture. Gao, B.C., 1996. DWI-A normalized difference water index for remote sensing of vegetation liquid water from space. Remote Sensing of Environment 58, 257–266. Havens, S., 2007. OpenGIS® transducer markup language implementation specification (version 1.0.0). Open Geospatial Consortium (OGC), Wayland. Report No. 06-010r2. Jackson, T.J., Chen, D., Cosh, M., Li, F., Anderson, M., Walthall, C., 2004. Vegetation water content mapping using Landsat data derived normalized difference water index for corn and soybeans. Remote Sensing of Environment 92 (4), 475–482. Kuehn, S., Benz, U., Hurley, J., 2002. Efficient flood monitoring based on RADARSAT-1 images data and information fusion with object-oriented technology. In: Proceedings of Geoscience and Remote Sensing Symposium, vol. 5, pp. 2862–2864. Laugier, O., Fellah, K., Tholey, N., Meyer, C., De Fraipont, P., 1997. High temporal detection and monitoring of flood zone dynamic using ERS data around catastrophic natural events: the 1993 and 1994 Camargue food events. In: Proceedings of Third ERS Symposium on Space at the Service of our Environment, Florence, Italy, March 14–21, 1997, p. 216. McFeeters, S.K., 1996. The use of the Normalized Difference Water Index (NDWI) in the delineation of open water features. International Journal of Remote Sensing 17 (7), 1425–1432. Na, A., Priest, M., 2007. OpenGIS® sensor observation service implementation specification (version 1.0). Open Geospatial Consortium (OGC), Wayland. Report No. 06-009r6. Rogers, A., Kearney, M., 2004. Reducing signature variability in unmixing coastal marsh Thematic Mapper scenes using spectral indices. International Journal of Remote Sensing 25 (12), 2317–2335. Simonis, I., 2007. OpenGIS® Sensor Planning Service Implementation Specification (Version 1.0). Open Geospatial Consortium (OGC), Wayland. Report No. 07014r3. Simonis, I., Wytzisk, A., 2003. Web Notification Service (Version 0.1.0). Open Geospatial Consortium (OGC), Wayland. Report No. 03-008r2. Tong, P.H.S., Auda, Y., Populus, J., Aizpuru, M., Habshi, A.A., Blasco, F., 2004. Assessment from space of mangroves evolution in the Mekong Delta, in relation to extensive shrimp farming. International Journal of Remote Sensing 25 (21), 4795–4812. Wang, S.-x., Liu, Y.-l., Zhou, Yi, Wei, C.-j., 2002. Study on the method of establishment of normal water extent database for flood monitoring using remote sensing. In: Proceedings of Geoscience and Remote Sensing Symposium, vol. 4, pp. 2048–22050.

Suggest Documents