Active on-demand service method based on event-driven ... - CiteSeerX

Computers & Geosciences 56 (2013) 1–11

Contents lists available at SciVerse ScienceDirect

Computers & Geosciences journal homepage: www.elsevier.com/locate/cageo

Active on-demand service method based on event-driven architecture for geospatial data retrieval Minghu Fan, Hong Fan n, Nengcheng Chen, Zeqiang Chen, Wu Du State Key Laboratory for Information Engineering in Surveying, Mapping and Remote Sensing (LIESMARS), Wuhan University, 129 Luoyu Road, Wuhan 430079, China

a r t i c l e i n f o

abstract

Article history: Received 7 September 2012 Received in revised form 14 January 2013 Accepted 27 January 2013 Available online 8 February 2013

Timely on-demand access to geospatial data is necessary for environmental observation and disaster response. However, traditional service methods for acquiring geospatial data are inefficient and cumbersome, which is not beneficial for timely data acquisition. In these service methods, data are obtained and published by managers and are then left to users to discover and to retrieve them. To solve this problem, we propose an event-driven active on-demand data service method, for which a prototype based on sensor web technologies is demonstrated. First, we select a subset of observed properties as the attributes of an observation event of a data service system. Event-filtering technologies are then employed to find the data desired by users. Finally, the data that meet the subscription requirement are pushed to subscribers on time. The aims of the implementation of the method are to test the suitability of the observation and measurement (O&M) profile for Earth observation and OGC event pattern markup language (EML) specification. We determined the attributes of observation events according to the requirement of the data service and encoded observation event information using the OGC Observations and Measurements specification. We encoded the information under filtering conditions using the OGC Event Pattern Markup Language specification. We implemented a data service method that is based on event-driven architecture via a combination of some sensor web enablement services. Finally, we verified the feasibility of the method using MODIS data from the forest fires that occurred on February 7, 2009, in Victoria, Australia. The results show that the proposed method can achieve actively pushing the desired data to subscribers in the shortest possible time. O&M profiles for Earth observation and EML are suitable for the metadata encoding of observation events and the encoding of subscription information respectively. They match well for the data service in the system. & 2013 Elsevier Ltd. All rights reserved.

Keywords: Active on-demand Data service Event-driven architecture Sensor web MODIS Fire

1. Introduction With the development of space science and technology, scientists, environment planners and managers have quickly obtained a wide range of Earth observation data (Franklin and Wulder, 2002; Rogan and Chen, 2004). These data are currently widely used in environmental management and planning in fields such as meteorology, hydrology, and natural disasters. (Leese, 1987; Luscombe and Hassan, 1993; Tralli et al., 2005). For many environmental events such as natural disasters, access to observational data are required in a timely manner. However, current geospatial data service methods are inefficient and cumbersome (Di et al., 2002; Heinen et al., 2009; Li et al., 2012). In these service methods, data are obtained and published by managers and it is left to the

n

Corresponding author. Tel.: þ86 1862771 6767; fax: þ86 27 6877 8229. E-mail address: [email protected] (H. Fan).

0098-3004/$ - see front matter & 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.cageo.2013.01.013

user to discover and to retrieve them when needed. If we can automatically filter observational data according to the requirement of users and then push the desired data to these users when such data are available, then the efficiency and the quality of data services will be significantly improved. Recently developed sensor web technologies can better achieve this idea, but only a limited number of relevant studies are available on this subject. The use of sensor web technologies to achieve active on-demand services for geospatial data is the main focus of our study. Traditional geospatial data services all are user-driven. Portal and web service are the most commonly used service styles. Portals provide users with a tool to find and download their desired data through HTTP or FTP. Common portals include the MODIS website,1 Global Land Cover Facility,2 United States

1 MODIS website, http://modis.gsfc.nasa.gov/data/, (accessed 29 August, 2012).

2

M. Fan et al. / Computers & Geosciences 56 (2013) 1–11

Geological Survey Global Visualization Viewer3 , etc. Web services provide users with some interfaces to query and obtain data. Most of the web services focus on data products delivery (Davies et al., 2009; Di, 2004; LAADS, 2012; Vannan et al., 2011). Some studies on the web services also explored data management methods (Horsburgh et al., 2009; Zhao et al., 2012). The above data service methods can meet the requirements of users to query, subscribe, and download archived data. However, such methods can have some limitations that restrict timely access to data when environmental events occur, mainly in the following aspects:

(1) Services are inefficient and cumbersome, which results in poor efficiency in data acquisition. Data are left to users to discover them and users wait for data to emerge. (2) Data are manually extracted by users when services lack an active notification mechanism. Consequently, users may not receive data on time when environmental events occur. (3) Only archived data are provided when services cannot be customized to filter future data, which is not conducive for the timely handling of environmental events.

The sensor web is a collaborative observation infrastructure that can promptly provide various kinds of observational data according to the observation task. Over the past 10 years, the standards and the applications of the sensor web have made great progress in environmental observation (Di et al., 2010; Moe et al., 2008). OGC and ISO have been formulating standards and protocols for the sensor web. The OGC Sensor Web Enablement (SWE) initiatives have developed a suite of standards that include observations and measurements (O&M, 1.0) (Cox, 2007a, b), sensor model language (SensorML, 1.0) (Botts, 2007), sensor observation service (SOS, 1.0) (Na and Priest, 2007), sensor planning service (SPS, 1.0) (Simonis, 2007), sensor alert service (SAS) (Simonis, 2006), and web notification service (WNS) (Simonis and Echterhoff, 2006). The standards of SWE 2.0 are partly approved, with a portion of it still under discussion and in the process of obtaining full approval. SWE 2.0 evolved from SWE ¨ 1.0 (Broring et al., 2011). O&M 1.0, SensorML 1.0, SOS 1.0, and SPS 1.0 upgraded to O&M 2.0, SensorML 2.0, SOS 2.0, and SPS 2.0, respectively. SAS evolved to sensor event service (SES), whereas WNS remains undeveloped. A new specification, event pattern markup language (EML) (Everding and Echterhoff, 2008), has been developed for the processing of complex events and event streams. In addition to SWE, OGC has developed a set of interoperable protocols such as web map service (Beaujardiere, 2006), web coverage service (WCS) (Whiteside and Evans, 2008), and web processing service (WPS) (Simonis, 2007). Sensor web technologies based on the above standards and protocols have been applied to data acquisition in environmental observation. Conover et al. used sensor web protocols in environmental data acquisition and management (Conover et al., 2010). Chen et al. combined WCS and SOS to implement near real-time ondemand retrieval of remote sensing observations (Chen et al., 2011). Chen et al. also used WPS to retrieve AutoChem data in the Antarctica Spatial Data Infrastructure automatically (Chen et al., 2010a, 2010b).

2 Global Land Cover Facility, http://glcf.umiacs.umd.edu/data/, (accessed 29 August, 2012).

3 USGS Global Visualization Viewer, http://glovis.usgs.gov/, (accessed 29 August, 2012).

Williams et al. similarly used WPS to receive real-time weather data based on O&M standards from SOS (Williams et al., 2011). The above studies all are based on the traditional serviceoriented architecture, which mainly applies to the request– response model and does not meet the requirement of environmental observation in which response to events is necessary (Kong et al., 2009). The event-driven architecture (EDA) compensates for this deficiency (Papazoglou and Heuvel, 2007). EDA can be used for timely response on various events and for coordination with business process integration in ubiquitous enterprises (Kong et al., 2009; Michelson, 2006). Environmental observation is a perceptual process that requires prompt response to environmental events and coordination for events processing. OGC has exerted a significant effort for some years to apply event-driven technology to the sensor web. First, a publish/subscribe mechanism (Eugster et al., 2003) was supported by SAS. As the successor of SAS, SES has supported complex event processing and event stream processing (Luckham, 2001). WNS can be used to push messages via various protocols. SPS 2.0 defined an event-driven mechanism for observation tasks. The OWS-6 SWE Event Architecture Engineering Report (Everding and Echterhoff, 2009) (we will abbreviate it to OWS-6 SWE EAER) describes an abstract event architecture which uses event-driven technology to extend the existing OGC architecture. Some studies used event-driven ideas. Yu et al. used an event-driven method in the geospatial workflow (Yu et al., 2008). In self-adaptive Earth predictive systems, an event-driven mechanism was used to achieve the feed-forward cycle (Yu et al., 2010). For the aforementioned problems, we propose an event-driven active on-demand data service method based on the event-driven technologies of the sensor web. The publish/subscribe mechanism is used for users to request for archived and future observational data. An event stream processing technology is employed to filter observational data and to find the data desired by users. Observation events and ordering events drive the data service process and the active data pushing. The implementation of the data service is based on EDA. The remainder of this paper is organized as follows: The conceptual framework of the event-driven active on-demand service method is illustrated in Section 2.1. The defining, encoding and management of observation events are described in Section 2.2. The encoding and use of event filters is discussed in Section 2.3. The event-driven process of the data service is presented in Section 2.4. A specific implementation of the data service is elaborated in Section 3. In Section 4, characteristics of the proposed method are discussed. Finally, Section 5 states the conclusions and recommendations for future research.

2. Methods 2.1. Conceptual framework The conceptual framework is shown in Fig. 1. Compared with the traditional data service framework, an event-driven framework primarily adds two modules: event channel and event service. In the traditional service process, geospatial data are received and saved as data products in the data service center after processing. Users can subsequently query and access the data products through the portal or web services. During the entire process, data and users are, in many cases, in a passive waiting state. Additional modules are used to implement an event-driven service process and to achieve the automatic conversion of service states, thereby enhancing the efficiency of the service process. In Fig. 1, the data service center corresponds to the traditional data center, which is responsible for receiving, processing, and


3

Fig. 1. Event-driven active on-demand data service conceptual framework.

Fig. 2. Definition of observation event and its description based on O&M.

storing data. This module can provide users data query and access services through the portal or the web service. We also use this module to encode observation event messages and to send the encoded message to the event channel. The event channel serves to collect and to distribute observation events from various sensors. The event service is responsible for processing data subscription tasks and for finding data that meet subscription conditions and for notifying users when their desired data are available. Data Pub/Sub is where data are published and subscribed, which is usually implemented via the portal and/or web service. Data Pub/Sub is also used to encode subscription information and to submit encoded information to the event service. Users access data according to notifications. To implement the above framework, we must address the following key issues: modeling observation events and encoding their information, encoding subscription information, and designing and implementing the event-driven process of data service. 2.2. Design event 2.2.1. Determining observation event attributes An observation event of a data service must contain the elements related to data query and data download. The OWS-6 SWE EAER presents a detailed (Everding and Echterhoff, 2009) discussion about events and gives a nice OGC event definition in the end: ‘‘An event is anything that happens or is contemplated as happening at an instant or over an interval of time’’. So, time is the most basic element of an event. In some contexts, particularly in Earth and environmental sciences, the term ‘‘observation’’ is used to refer to the result. Acquired remotely sensed images can be viewed as results or products of observations, which are what the user is most concerned about. Remote sensing data are generally stored according to sensors’ bands. Usually, a sensor is on a specific platform and contains one or more bands. In remote sensing applications, the band is often the most basic unit. A specific application is often related in a specific area. Besides

the above elements, other properties, such as data resolution, may be needed by data services. Based on the above ideas and the discussion, we give a following description of the observation event in data service, as shown in Fig. 2. Fig. 2(A) shows that an observation event contains nine elements. In Earth observation, platform and sensor usually refer to a satellite and to a space-borne sensor, respectively. BandID is the number of a band. EventTime is the date and the time of an observation. Location is the geographical area of observed object. Attributes refers to the properties of an observation. Result denotes observation data and its properties. 2.2.2. Encoding observation event Encoding events is necessary for open interoperability. OWS-6 SWE EAER thinks an O&M observation is a typical event. We use the Earth Observation Metadata profile of Observations & Measurements (Gasperi et al., 2012) as the encoding specification of an O&M observation event. In the profile, matching fields can be found for all parameters of an observation event, as shown in Fig. 2(B). Platform and Sensor correspond to the observationEquipment field. BandID may correspond to the observedProperty field. EventTime corresponds to the phenomenonTime field. Location corresponds to the featureOfInterest field. Attributes corresponds to the metaDataProperty field. Result corresponds to the result field and the resultTime field. More details on these fields are presented in the profile. 2.2.3. Observation event management An observation event is encoded in the data service center, stored and distributed through the event channel, and is then processed by the event service. When an observation is finished, the data service center extracts the observed information and encodes such information as event messages. These messages are sent to the event channel, which classifies and stores them as

4


streams according to the band and to the sensor. Meanwhile, all event streams are ordered by date and time. In other words, a sensor band corresponds to a chronological event stream. Upon receipt of the event subscription, the event channel extracts one or more event streams and sends them to the subscriber. Some constraint conditions such as sensor, band, date, and/or time can be used to extract a section of a stream. The event service handles event messages according to the user’s subscription information. Because observation events come from an observation system, there may be false events in the event stream if the system is abnormal. For example, when a sensor malfunctions, the corresponding observation result may be invalid. Special handling for such events is needed. However, related issues are beyond the scope of this study. We do not intend to have any further discussion. 2.3. Filtering observation events 2.3.1. Encoding filtering conditions Encoding filtering conditions is the key to filter observation events and then find data of interest for users. In a data query, a user will input some data constraint information, including temporal information, spatial information, and other data properties information. Such information should be encoded as a filtering condition of observation events. In sensor web technologies, EML can be used to encode the above information. EML is developed to describe event patterns for event (stream) processing and analysis. SES provides multistage filters that are built on EML for incoming events/event streams. These filters use a combination of logical, spatial, temporal, arithmetic, and comparison operators as well as functions to define the filter criteria and can handle queries in data streams. With these filters, all kinds of observation events can be identified and filtered according to the subscription of the users. 2.3.2. Using filtering conditions In Earth observations, observations are composed of continuous observation streams. A sensor usually constantly observes the Earth along some of its orbits. Observational data streams have certain continuity in time and space along the observational track. However, users are often interested in a specific area. To acquire observational data on the area, one or more observations may be needed. By constantly filtering incoming data streams according to the subscription conditions, the data required by users can be found in time. The method of filtering data flows can also be used for archived data, as long as the metadata information of such archived data is encoded as observation events. In the

process of data servicing, we can define some filters according to the subscription conditions of users and then use these filters to handle observation event messages. The filtered results are messages that users are interested in and would like to access. 2.4. Event-driven process In data service, we use the modeling idea of an event-driven application presented in the SPS 2.0 specification. In the specification, events represent state changes of tasks and events drive the execution of tasks. The specification defined events and states for observational task and request task, and use task state machines to describe the event-driven process of tasks. Similarly, we defined events and states of data service tasks and designed the event-driven process of data service task. A data service task state machine is shown in Fig. 3. The data service task state machine has three states (InAcquisition, Waiting, and Filtering) and eight events (DataAcquired, DataPublished, SubscriptionSubmitted, SubscriptionFailed, SubscriptionUpdated, SubscriptionCancelled, SubscriptionExpired, and SubscriptionFinished). These events cause the state changes of the data service task among the three states. The event we are most concerned about is the DataAcquired event. A DataAcquired event represents the end of an observational task and signifies the start of information processing for the observational data. In Fig. 3, Notify indicates that a notification is sent to the subscriber after a corresponding event occurred. The event-driven process is as follows: first, a user submits an order. If the order is invalid or if the subscription is not feasible, the subscription fails, and the process ends. Otherwise, the data service task enters into the InAcquisition state. In this state, the Waiting state is maintained if observation events do not occur or the task enters into the Filtering state. In the Filtering state, messages of observation events are filtered, and a notification is sent to the user when some messages meet the requirements of the order. Before the end of the data service task, the process always returns to the Waiting state when no message is being handled. In the InAcquisition state, the user can update the order. If the user cancels the order or if the order cannot be fulfilled, it is expired, or is finished, and the process ends.

3. Experiments and results 3.1. Architecture and components In OWS-6 SWE EAER, Everding and Echterhoff describe an abstract event architecture, which contains two implementation

Fig. 3. Diagram of the data service task state machine.


5

Fig. 4. Architecture of the event-driven active on-demand data service.

methods of event-driven applications. The first method is to combine SES and other web services, which has good compatibility with existing OGC services but is coarse-grained to event handling. The second method is that every web service implements interfaces related to EDA, which is fined-grained to event handling but not compatible with existing OGC services. We use the first method because it is enough to verify the feasibility of our method. Our architecture of the event-driven active ondemand data service is shown in Fig. 4. The components and their functions of the above architecture are as follows. The observation event cloud is where an observation event originates and is stored. This component consists of the data service center, event provider (can be a web application, a web service, or another SES), and event library. The data service center is responsible for the storage of observational data and for encoding the messages of observation events. The event provider is employed to store encoded messages in the event library and to handle the event subscription. The observation event cloud provides the user with an event distribution service according to the event subscription. Upon receipt of an event subscription, the event provider extracts event messages according to the subscription information and sends such messages to the subscriber. SES handles the data subscription task. The Subscribe and Notify operations of SES are used to receive the subscription information and event messages, respectively. WNS is employed to send the notification to the subscriber asynchronously if necessary. The DoNotification operation of WNS receives message to be sent to the subscriber. The data service portal accepts and encodes data subscription information and then delivers encoded information to SES. 3.2. Environment and usage case of the experiment In environmental observations, people tend to extract information on environmental events occurring in a certain time and place from observed data. Thus, data service is always related to a specific spatial and temporal event. As with our usage case, we select an observation of a fire event. We assume that users want to obtain the Terra MODIS data to extract essential information about Victoria, Australia on February 7, 2009, when a forest fire occurred. Related MODIS data products are described in the succeeding paragraph. Many Terra MODIS data products exist (Justice et al., 2002b), including fire products (Justice et al., 2002a). The products that can be used to extract fire information are MOD02, MOD03, MOD14, MOD14A1, MOD14A2, etc. Fig. 5 shows the coverage of Terra MODIS data in Victoria, Australia on February 7, 2009. The yellow region is Victoria. The white polygons represent data coverage of MOD02, MOD03, and MOD14 products, and correspond to landscape data, each of which contains data observed within 5 min. The figure at the upper right corner of each polygon

Fig. 5. MODIS data coverage at Victoria, Australia on February 7, 2009.

represents the starting UTC time of the observation. The red polygon represents data coverage of MOD14A1 and MOD14A2 products, and corresponds to a tile, which may contain thermal anomaly data as observed in 1 or more days. The users region of interest is symbolized by the green polygon. If a user wants to access MOD02 data about the occurrence in Victoria on February 7, 2009, he has to submit the following information: platform, sensor, product type, area of interest, time period, etc. He can enter such information through the data service portal, as shown in Fig. 6. In the data service portal, selected input items include the platform (single option), sensor (single option), band (multiple option), resolution (multiple option), product type (multiple option), area of interest (input values via Google Map or manual input), and time period (input values via date/time controls or manual input). The user can use a combination of these conditions to subscribe to their desired data. 3.3. Encoding of subscription information Upon receipt of a data subscription, the data service portal uses EML to encode the subscription information. A sample of encoded subscription information is shown in Fig. 7. Fig. 7 contains a UML class diagram and some key coding examples of one subscription information. This example contains the five conditions that are mentioned above. For platform, sensor, and product type, we need only to specify their value (Fig. 7C). But for area of interest and time period, some operators are needed. In Fig. 7, the Intersects operator indicates that the observation events whose spatial range intersects with AOI_01 meet the requirement (Fig. 7B). The AnyInteracts operator indicates that the observation events whose temporal range intersects with TP_01 meet the requirement

6


Fig. 6. Data service portal.

(Fig. 7A). For this example, the above spatial condition and temporal condition must be met simultaneously (Fig. 7D). SES provides support for these operators. The other codes (Fig. 7E) are used to handle event stream.

information in Fig. 8 can therefore be available after a successful order in the LADDS site.

3.4. Encoding observation events

The process flow of the data service is shown in Fig. 9. The data service task consists of two processes: event publishing and subscription information processing. The process for event publishing is as follows: The data service center registers event publishers (sensors) through the RegisterPublisher operation of the event provider and then sends event messages (Fig. 8) from the publishers to the event provider through the Notify operation. Messages are stored in the event library by the event provider. Within the validity period of registration, the event provider accepts event messages of registered publishers and stores such messages. Subscription information processing has the following steps: first, the subscriber submits a data subscription through the Subscribe operation of SES. SES establishes the filter according to the subscription information (Fig. 7) and then submits the event subscription through the Subscribe operation of the event provider. The event provider then selects event messages according to the event subscription and sends such messages to SES

The description of the event information is based on the metadata of the observation. Typically, only a portion of the metadata is needed, because a common data query does not need all metadata. We selected some metadata and applied them to the observation event. The information of observation events is encoded with O&M. A sample is shown in Fig. 8. Fig. 8 contains a UML class diagram and some key coding examples of one observation information. This example meets the conditions shown in Fig. 7. Between Figs. 7 and 8, segment A, segment B and segment C correspond respectively. They are temporal information, spatial information and property information respectively. Segment D is result information in Fig. 8. Data access methods can be traditional web methods such as HTTP or FTP or data services such as SOS or WCS. In the example, we suppose that the MODIS data are stored in the LADDS site of NASA and that the user can access such data via FTP. The access

3.5. Process flow of the data service task


7

Fig. 7. Sample of encoded subscription information.

through the Notify operation of SES. In SES, the phenomenonTime, featureOfInterest, platForm, instrument, and productType of these messages are extracted and then compared with corresponding items of the previously built filter. If some messages match the conditions of the filter, the data access information of these messages is sent to WNS through the DoNotification operation of WNS. WNS sends the information to the subscriber. Finally, the subscriber accesses data from the LADDS site by using the access information. During the validity period of the data subscription, the process that consists of sending event messages, filtering event messages, and notifying the subscriber (if the message matches with the conditions of the filter) loops. Meanwhile, SES also notifies the subscriber via WNS if the data subscription task cannot be fulfilled or if such task is expired. In addition, the subscriber can update or cancel the data subscription through the Renew and the Unsubscribe operations, respectively.

3.6. Observation event simulation and experimental results We extract observation metadata from the satellite image data service to model the observation event because no real satellite sensors can be directly connected. An event corresponds to an observation that acquires a band data in landscape data. Typically, landscape data correspond to multiple events. An example is the landscape data with the start time of 00:30, as shown in Fig. 5, which correspond to 38 events (the MODIS sensor theoretically only has 36 bands, but the 13th and the 14th bands are divided into low and high bands during observation). Other landscape data with start times of 12:30 and 12:35 correspond to 16 events (in the night mode of the MODIS sensor, only the 20th to the 36th bands transmit data, except for the 26th band). In other words, on February 7, 2009, the observations of the Terra MODIS sensor in Victoria, Australia corresponded to 70 events. For higher-level data products, datasets no longer correspond to the bands of the

8


Fig. 8. Sample of encoded observation information.

MODIS sensor, but such datasets are seen as single-band data. For example, a tile data of the MOD14A1 product corresponds to eight events. Each event corresponds to 1-day data, which are stored as a dataset and as band data. Event Provider receives and handles above events and then stores them in a table of a database. If a new event satisfies a subscription of SES, then it will be send to the SES immediately before it is stored in event library. But for archived events, the Event Provider has to query event library. SES filters these events and extracts data information from the events that satisfied users’ subscriptions. As indicated by the information encoded in Fig. 7, 70 observation events are assessed

on whether they meet required conditions after filtering the observation event streams of Terra MODIS. Multiple band data of MODIS are generally stored in a single data file. Thus, the 70 events actually correspond to five data files that the user can obtain.

4. Case study and discussions We implemented an active on-demand data service method based on event-driven technologies of the sensor web to obtain geospatial data. A data acquisition task verified the feasibility of the proposed method. The experiment used archived data.


9

Fig. 9. Process flow of the data service task.

However, the method is equally applicable to future data because the process of the data service task is the same, except that the subscriber has to wait until the future observation event occurs for notification. At present, the usage case only support common constraint conditions that can verify the feasibility of the event-driven method. Other conditions may also be added, e.g. resolution. The filtered results of the above usage case contain information on all bands of MODIS. However, if we add a resolution constraint such as 250 m, the filtered results contain only information of the first and the second bands of MODIS. Adding a condition to the method is simple. What we need to do is to insert the condition

into the subscription information, to add corresponding metadata in the observation information, and to implement the corresponding processing function in SES. The addition of multi-platform and/or multi-sensor conditions is similar. Aside from providing support to traditional data subscription methods, the usage case also aids in band-based data subscriptions. For example, the fire detection algorithm for MODIS requires data from the first, second, seventh, 21st, 22nd, 31st, and 32nd bands (Giglio et al., 2003). The proposed method can directly use the above bands as constraint conditions to subscribe to data. If the MODIS data are stored in data files, the result of the subscription is the same as the usage case. However, if the data

10


are stored in bands (e.g., data are served via WCS or SOS), the user can more accurately access their desired data. Compared with traditional data service methods, the proposed method has the following advantages:

(1) Active data push can improve the efficiency of data discovery and acquisition. Traditional data services are inefficient and cumbersome. Data managers publish data through portals or web services and it is up to the user to find and access the data. The proposed method can notify the subscribers when the data of interest are available and subscribers can then access the data for use. (2) An active notification mechanism can improve the efficiency and the quality of data services. In traditional data services, users themselves must manage the process of data discovery and acquisition, as well as find and locate problems that may occur during data acquisition. The proposed method can actively send sufficient information to the users based on the events that occurred in the data service tasks, which is beneficial for the users to understand and to handle data acquisition tasks in time. (3) Users can conveniently subscribe to both archived and future data. Traditional data services usually can only provide users with archived data and do not enable users to subscribe to future data. When new observations are added into the system, corresponding observation event messages are processed automatically, and the subscribers are then notified. Moreover, the proposed method does not need to change the existing data organization because the observation events only use some observation metadata information and the processing of events is irrelevant to data organization. Additionally, the solution of the proposed method has a good scalability, which benefits from the loose coupling of EDA. In Fig. 4, the event provider and SES are independent of each other. We can enhance data service capabilities by adding another event provider and/or SES. We can also add another data service center to provide more observation events. Meanwhile, we can reduce data service capabilities by simply disabling or deleting relevant services. A limitation of the proposed method is that the user must have a certain expertise on Earth observation. If we can establish some connection between phenomenon, event and observation, such limitation can be greatly reduced. Related research is in progress.

5. Conclusion and outlook Traditional data service methods are less efficient and do not facilitate easy, timely discovery, access and use of geospatial data. This paper proposed an event-driven active on-demand data service method based on OGC sensor web technologies. The method mitigates some limitations of traditional methods and improves the efficiency and quality of data services. First, the method implements active data push to ensure the delivery of data to users as soon as possible. Second, the method has active notifications that enable users to handle events that occurred during the process of data acquisition promptly. Third, the method covers future data subscription, which benefits users by enabling them to obtain and process their desired data in time. In summary, the proposed method can send the desired data to subscribers in the shortest possible time. In addition, EDA increases the flexibility of the data service system. The sensor web is a perceptive environment. The event-driven method can significantly enhance the reactivity of the sensor web

and enables it to react promptly to environmental changes, which is conducive to the discovery and the tracking of environmental events. In these applications, data events can be published through observation topics (a flexible category, e.g. sensor topic, product topic, hotspot topic, etc., each topic contains the next topic in turn) and a subscriber can subscribe to data according to these topics, and other relevant information such as spatial and temporal information. As for the encoding of subscription information and the process of observation events processing, they are similar to this method. Within a valid subscription period, data that meet the subscription information will be pushed to the subscriber in time. Future studies can investigate the use of the event-driven method in active on-demand observations of fire occurrences.

Acknowledgments This work was supported by grants from the National Natural Science foundation Projects of China (No. 40971243, PI: Dr. Hong Fan; No. 41023001, PI: Dr. Jianya Gong), the Chinese 863 program (No. SS2012AA121001, PI: Dr. Jianya Gong) and the National Basic Research Program of China (No. 2011CB707101, PI: Dr. Nengcheng Chen). References Beaujardiere, J.d.l (Ed.), 2006. OpenGISs Web Map Server Implementation Specification (Version 1.3.0). OGC Document Number: 06-042. Open Geospatial Consortium. Wayland, MA, USA, p. 85. Botts, M. (Ed.), 2007. OpenGISs Sensor Model Language (SensorML) Implementation Specification (Version 1.0.0). OGC Document Number: 07-000. Open Geospatial Consortium. Wayland, MA, USA, p. 180. ¨ Broring, A., Echterhoff, J., Jirka, S., Simonis, I., Everding, T., Stasch, C., Liang, S., Lemmens, R., 2011. New generation sensor web enablement. Sensors 11, 2652–2699. Chen, N., Chen, Z., Di, L., Gong, J., 2011. An efficient method for near-real-time ondemand retrieval of remote sensing observations. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 4, 615–625. Chen, N., Di, L., Yu, G., Gong, J., 2010a. Automatic on-demand data feed service for AutoChem based on reusable geo-processing workflow. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 3, 418–426. Chen, N., Li, D., Di, L., Gong, J., 2010b. An automatic SWILC classification and extraction for the AntSDI under a sensor web environment. Canadian Journal of Remote Sensing 36, S1–S12. Conover, H., Berthiau, G., Botts, M., Goodman, H.M., Li, X., Lu, Y., Maskey, M., Regner, K., Zavodsky, B., 2010. Using sensor web protocols for environmental data acquisition and management. Ecological Informatics 5, 32–41. Cox, S. (Ed.), 2007. Observations and Measurements: Part 1: Observation Schema (Version 1.0). OGC Document Number: 07-022r1. Open Geospatial Consortium. Wayland, MA, USA, p. 85. Cox, S. (Ed.), 2007. Observations and Measurements: Part 2: Sampling Features (Version 1.0). OGC Document Number: 07-002r3. Open Geospatial Consortium. Wayland, MA, USA, p. 46. Davies, D.K., Ilavajhala, S., Wong, M.M., Justice, C.O., 2009. Fire information for resource management system: archiving and distributing MODIS active fire data. IEEE Transactions on Geoscience and Remote Sensing 47, 72–79. Di, L., 2004. GeoBrain: a web services based geospatial knowledge building system. In: Proceedings of the NASA Earth Science Technology Conference 2004, Palo Alto, CA, USA. Di, L., Moe, K., Zyl, T.L.v., 2010. Earth observation sensor web: an overview. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 3, 415–417. Di, L., Yang, W., Deng, M., Deng, D., 2002. Interoperable access of remote sensing data through NWGISS. In: Proceedings of the IEEE International Geoscience and Remote Sensing Symposium, 2002, IGARSS ’02 255–257. Eugster, P.T., Felber, P.A., Guerraoui, R., Kermarrec, A.-m., 2003. The many faces of publish/subscribe. ACM Computing Surveys 35, 114–131. Everding, T., Echterhoff, J. (Eds.), 2008. Event Pattern Markup Language (EML) (Version 0.3.0). OGC Document Number: 08-132. Open Geospatial Consortium. Wayland, MA, USA, p. 58. Everding, T., Echterhoff, J. (Eds.), 2009. OGCs OWS-6 SWE Event Architecture Engineering Report (Version: 0.3.0). OGC Document Number: 09-032. Open Geospatial Consortium. Wayland, MA, USA, p. 168. Franklin, S.E., Wulder, M.A., 2002. Remote sensing methods in medium spatial resolution satellite data land cover classification of large areas. Progress in Physical Geography 26, 173–205.


Gasperi, J., Houbie, F., Woolf, A., Smolders, S. (Eds.), 2012. Earth Observation Metadata Profile of Observations and Measurements (Version: 1.0). OGC Document Number: 10-157r3. Open Geospatial Consortium. Wayland, MA, USA, p. 78. Giglio, L., Descloitres, J., Justice, C.O., Kaufman, Y.J., 2003. An enhanced contextual fire detection algorithm for MODIS. Remote Sensing of Environment 87, 273–282. Heinen, T., Kiemle, S., Buckl, B., Mikusch, E., R., D.G.L., 2009. The geospatial service infrastructure for DLR’s national remote sensing data library. Selected Topics in Applied Earth Observations and Remote Sensing 2, 260–269. Horsburgh, J.S., Tarboton, D.G., Piasecki, M., Maidment, D.R., Zaslavsky, I., Valentine, D., Whitenack, T., 2009. An integrated system for publishing environmental observations data. Environmental Modelling and Software 24, 879–888. Justice, C.O., Giglio, L., Korontzi, S., Owens, J., Morisette, J.T., Roy, D., Descloitres, J., Alleaume, S., Petitcolin, F., Kaufman, Y., 2002a. The MODIS fire products. Remote Sensing of Environment 83, 244–262. Justice, C.O., Townshend, J.R.G., Vermote, E.F., Masuoka, E., Wolfe, R.E., Saleous, N., Roy, D.P., Morisette, J.T., 2002b. An overview of MODIS land data processing and product status. Remote Sensing of Environment 83, 3–15. Kong, J., Jung, J.-Y., Park, J., 2009. Event-driven service coordination for business process integration in ubiquitous enterprises. Computers and Industrial Engineering 57, 14–26. LAADS Web, 2012. MODAPS web services. /http://ladsweb.nascom.nasa.gov/data/ web_services.htmlS (accessed 29.08.12). Leese, J.A., 1987. Remote sensing applications in the meteorology and operational hydrology programmes of WMO. Advances in Space Research 7, 49–57. Li, X., Yang, J., Liu, S., Chen, F., Liu, J., 2012. Study on web service-based remote sensing data service platform for digital Earth. In: 2012 International Conference on Technology and Management, ICTAM 2012, Jeju, Jeju-Island, Korea 403–410. Luckham, D.C., 2001. The Power of Events: An Introduction to Complex Event Processing in Distributed Enterprise Systems. Addison-Wesley Longman Publishing Co., Inc., Boston, MA, USA 376. Luscombe, B.W., Hassan, H.M., 1993. Applying remote sensing technologies to natural disaster risk management: implications for developmental investments. Acta Astronautica 29, 871–876. Michelson, B.M., 2006. Event-driven architecture overview. Patricia Seybold Group. /http://www.elementallinks.com/2006/02/06/event-driven-architecture-over view/S (accessed 29.08.12).

11

Moe, K., Smith, S., Prescott, G., Sherwood, R., 2008. Sensor web technologies for NASA earth science. In: Proceedings of the 2008 IEEE Aerospace Conference. Vol. 1–9, 2211–2217. Na, A., Priest, M. (Eds.), 2007. Sensor Observation Service (Version 1.0). OGC Document Number: 06-009r6. Open Geospatial Consortium. Wayland, MA, USA, p. 104. Papazoglou, M.P., Heuvel, W.-J.v.d., 2007. Service oriented architectures: approaches, technologies and research issues. VLDB Journal 16, 389–415. Rogan, J., Chen, D., 2004. Remote sensing technology for mapping and monitoring land-cover and land-use change. Progress in Planning 61, 301–325. Simonis, I. (Ed.), 2006. OpenGISs Sensor Alert Service Implementation Specification (Version 0.9). OGC Document Number: 06-028r3. Open Geospatial Consortium. Wayland, MA, USA, p. 110. Simonis, I. (Ed.), 2007. OpenGISs Sensor Planning Service Implementation Specification (Version 1.0). OGC Document Number: 07-014r3. Open Geospatial Consortium. Wayland, MA, USA, p. 186. Simonis, I., Echterhoff, J. (Eds.), 2006. Draft OpenGISs Web Notification Service Implementation Specification (Version 0.0.9). OGC Document Number: 06095. Open Geospatial Consortium. Wayland, MA, USA, p. 64. Tralli, D.M., Blom, R.G., Zlotnicki, V., Donnellan, A., Evans, D.L., 2005. Satellite remote sensing of earthquake, volcano, flood, landslide and coastal inundation hazards. ISPRS Journal of Photogrammetry and Remote Sensing 59, 185–198. Vannan, S.K.S., Cook, R.B., Pan, J.Y., Wilson, B.E., 2011. A SOAP web service for accessing MODIS land product subsets. Earth Science Informatics 4, 97–106. Whiteside, A., Evans, J.D. (Eds.), 2008. Web Coverage Service (WCS) Implementation Standard (Version 1.1.2). OGC Document Number: 07-067r5. Open Geospatial Consortium. Wayland, MA, USA, p. 133. Williams, M., Cornford, D., Bastin, L., Jones, R., Parker, S., 2011. Automatic processing, quality assurance and serving of real-time weather data. Computers and Geosciences 37, 353–362. Yu, G.E., Di, L., Moses, J.F., Li, P., Zhao, P., 2008. Geospatial workflow in a sensor web environment: transactions, events, and asynchrony. In: Proceedings of the International Geoscience and Remote Sensing Symposium (IGARSS), Boston, MA, United States. V132–V135. Yu, G.E., Di, L., Zhang, B., Wang, H., 2010. Coordination through geospatial web service workflow in the sensor web environment. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 3, 433–441. Zhao, Y., Hu, C., Shen, H., Ma, D., Li, X., Huang, Y., 2012. A hierarchical organization approach of multi-dimensional remote sensing data for lightweight web map services. Earth Science Informatics 5, 61–75.