1. Page 1 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 Sensor-Web-technology based Sensor-model Integration for Supporting the GEOSS Societal Benefit Areas Liping Di Center for Spatial Information Science and Systems George Mason University ldi@gmu.edu

2. Page 2 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 Introduction Earth observation (EO) through scientific sensors is the most important method for measuring the current state of the Earth system (ES). –In recent years, one of the most significant technological developments in EO is the EO sensor web. –The EO sensor web is a web of interconnected, heterogeneous, geospatial sensors that are interoperable, intelligent, dynamic, flexible and scalable. The sensor web approach employs new data acquisition strategies and systems for integrate ES models (ESM) simulate the ES and/or its components, for understanding the functioning of the system and predicting its future state. –ESM requires initial states to run as well as ground truth to validate the model. –Those information can be provided through sensor web observation Although both areas have been heavily invested, traditionally there has been little integration and coupling between EO and ESM. This presentation discusses a service-oriented general framework for facilitating the dynamic connection and interoperation between the EO sensor web and Earth system models (model Web) and provides examples of its applications in GEOSS societal benefit areas.

3. Page 3 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 The Self-adaptive Concept The self-adaptation concept is the central piece of the control theory widely and successfully used in engineering and military systems. Such a system contains a predictor and a measurer –The predictor makes a prediction based on initial conditions –The measurer then measures the state of a real world phenomenon. –A feedback mechanism is built in that automatically feeds the measurement back to the predictor. –The predictor takes the measurement against the prediction to calculate the prediction error and adjust its internal state based on the error. Thus, the predictor learns from the error and makes a more accurate prediction in the next step. With the prediction-feedback mechanism, a system can soon learn the pattern of the real-world phenomenon and make accurate predictions.

4. Page 4 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 Self-adaptive Earth Prediction Systems (SEPS) In Earth science domain, we can consider the sensor Web (SW) as the measurer and the ESM as the predictor We apply the self-adaptation concept to Earth system prediction by proposing the concept of Self-adaptive Earth Predictive System (SEPS) A SEPS consists of an EO component (SW) and an ESM component (MW), coupled by a SEPS framework component (The Connector) –EO SW measures the ES state while ESM predicts the evolution of the state. –A feedback mechanism processes EO measurements and feeds them into ESM during model runs, as well as initial conditions. –A feed-forward mechanism compares ESM predictions with scientific goals for scheduling or selecting further optimized/targeted observations. –The SEPS framework automates the feedback and feed- forward mechanisms (called the FF loop).

5. Page 5 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 Standard-based SEPS Framework with SOA The importance for improving the model prediction and emergency responses through rapid interaction between sensor systems and Earth system models has been recognized in recent years –The multi-agency (NSF, NOAA, NIH) research program on Dynamic, Data Driven Application System (DDDAS) is an example. The current problem includes: –Not recognize the sensors and model as an integrated system. –Ignore the feed-forward from model to sensor for dynamic sensor planning and refined observation. –The coupling between the sensors and the Earth system models are done case-by-case in ad hoc fashion. Cannot be reused with different models and different sensor systems. A standards-based general SEPS framework is needed: –Can work with various ESMs and EO sensors/data systems. –Fully support the automation of the FF loop. –Provide generalization through standard interfaces and services under service-oriented architecture.

6. Page 6 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 Overall Architecture 1.FTP, HTTP, WCS-T, WFS-T 2.SPS 3.THREDS, ECHO, CSW 4.SPS 5.OpenDAP, WCS, WFS 6.SOS 7.CSW, WFS, WCS 8.ESM state 9.WCS, WFS 10.WNS 11.WCS, WFS

7. Page 7 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 The Standard Interfaces and Components of SEPS Framework The framework consists of five subcomponents interconnected by standard interfaces. –Data Discovery and Retrieval Services (DDRS) –Data Pre-processing, Integration, and Assimilation Services (PIAS) –Science Goal Monitoring Services (SGMS) –Data and Sensor Planning Services (DSPS) –Coordination and Event Notification Services (CENS). Three major groups of standard interfaces in the framework –Interfaces to the sensor world (Sensor Web and data archives). –Interfaces to the Earth system models (model webs and legacy models). –Internal data and service interfaces between components of the framework.

8. Page 8 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 SEPS Standard Interfaces for Connection to Sensor Web ISO 19130 – Sensor Data Model for Imagery and Gridded Data, provides the conceptual framework for enabling the sensor web. OGC Sensor Web Enablement (SWE) initiatives have developed six implementation specifications for the sensor web –SensorML, a general model and XML encoding of the geometric, dynamic, and observational characteristics of a sensor; –Observation and Measurements (O&M), a framework and GML encoding for measurements and observations; –Sensor Planning Service (SPS), by which a client can determine the feasibility of a desired set of collection requests for one or more mobile sensors/platforms; –Web Notification Service (WNS) by which a client may conduct an asynchronous dialog with one or more other services; –Sensor Alert Service (SAS) by which a client can register for and receive sensor alert messages, part of the OGC Sensor Alert Service Interoperability Experiment; –Sensor Observation Service (SOS) providing an API for managing deployed sensors and retrieving sensor data.

9. Page 9 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 Standard interfaces among the SEPS components OGC Data Interoperability Protocols –Web Map Services (WMS), –Web Feature Services (WFS) –Web Coverage Services (WCS) –Catalog Services - Web Profile (CSW) OPeNDAP data access protocols W3C and OGC web geo-processing protocols –WSDL, BPEL, UDDI, SOAP. –The OGC web geoprocessing specifications.

10. Page 10 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 Data Discovery and Retrieval Services (DDRS) Component DDRS is one of the two subcomponents in the framework that connect to the EO sensor and data world. The major functions of DDRS are to discover and obtain data. It uses open standards for interfacing with data sources, allowing a web sensor or an EO data system equipped with standard interfaces to be easily plugged into the framework. –For data discovery, OGC CSW will be used for all data sources. –For data access to traditional sensor data systems, the OGC WCS, OGC WFS, and OPeNDAP protocols will be used. –For connection to web-enabled sensors, OGC SWE protocols (e.g., SOS, SensorML) will be used. –The interfaces between DDRS and PIAS will be CSW for catalog search and WCS and WFS for data communication. DDRS hides the complexity of the EO world behind its standard interfaces. To EO systems, DDRS is their client. But to PIAS, DDRS is the single point of entry for all required data from the EO world.

11. Page 11 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 Data Preprocessing, Integration, and Assimilation Services (PIAS) Component ESMs need high-level geophysical products in a model- specific form as input. Such products are normally not available directly from sensor measurements, but are derived from multi-sensor measurements and assimilated with data from other sources. The major function of PIAS is to generate ESM-specific products from data obtained through DDRS. In order for the SEPS framework to support easy plug- and-play of ESMs, a general approach for product generation is needed. We use the customizable virtual geospatial product approach. ESMs will use products generated by PIAS in two ways: 1) as the boundary conditions, and 2) as run-time ground truth.

12. Page 12 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 Science Goal Monitoring Services (SGMS) Component SGMS takes ESM outputs and in many cases the geophysical products generated by PIAS as inputs to analyze against science goals for determining whether additional or refined geophysical products are needed by the model to meet these goals. Use of geophysical products in SGMS rather than sensor measurements hides the complexity of the EO world from science-goal analysis and enables dynamic plug-in and removal of sensors without affecting SGMS. Like PIAS, SGMS needs a general method to evaluate whether diverse science goals are met. We will use the same approach in SGMS as that proposed for PIAS. Objective metrics such as information content and system uncertainty will be used to ensure that SGMS can provide accurate description of the additional or refined geospatial products needed. The interfaces between SGMS and PIAS to be WCS and WFS. SGMS will take ESM outputs in either netCDF or the ESMF state through WCS, OPeNDAP, or FTP.

13. Page 13 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 Data and Sensor Planning Services (DSPS) Component The major function of DSPS is to translate the requirements for geophysical products to EO data requirements and to schedule the acquisition of such data. The translation service uses geospatial processing models to trace back and identify all raw datasets required for generating the products. –From the specifications on geophysical products (e.g., time, spatial location, accuracy), the service will determine the specifications for raw data. –Then DSPS will find the best source for such data and schedule data acquisition through working with CSW portal in DDRS. –If the source is a data system, DSPS will provide information to CENS for scheduling the ordering and retrieval. –If the source is a schedulable Web sensor, DSPS will use OGC SPS to schedule observations and register the observation event in CENS. Because a geophysical product can be traced back to multiple raw datasets, a request for the refined product may result in scheduling or discovering simultaneous multiple-sensor observations. –A coincident search capability is provided in the CSW portal to help DSPS discover simultaneous observations.

14. Page 14 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 Coordination and Event Notification Services (CENS) Component CENS works as a rhythm controller for the FF loop. Since everything in the framework is implemented as a Web service, the loop can be implemented as a large workflow (FF-loop workflow). Two mechanisms, dynamic data retrieval, and asynchronous data notification, enable the dynamic data flow in the FF loop. –These two mechanisms share the same goal of using dynamic data but from the ESM and EO standpoints. –For dynamic data retrieval, a request is initiated from an ESM. CENS then acts upon the request, searches, prepares, and downloads data, and then generates and feeds the specific products to the model. –For asynchronous data notification, a sensor or data event will initiate the FF loop. A notification service will manage the event subscription and registry and monitor the state change of the registered events. –Event-based communication will be established between the server or the notification service and the client or the ESM. The notification service keeps a register of events supported by the sensor web. OGC WNS and SAS will be the key service protocols for the asynchronous mechanism.

15. Page 15 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 Approach for Interoperation with Earth System Models The SEPS aims at bridging existing Earth science models and the Sensor Web. Major interoperation between ESM and SEPS takes place in the interfaces between PIAS and ESM and between SGMS and ESM Modern Sensor Web technology is based on service- oriented architecture that utilizes Web service standards and technology, while most of Earth science models still use conventional server/client architecture or even a single system under the Earth Science Modeling Framework (ESMF). A mechanism to bridge the two worlds together has been studied

16. Page 16 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 ESMF The modeling community has developed the standard for interoperability between models—The Earth Science Modeling Framework (ESMF). ESMF is a high-performance, flexible software infrastructure to increase the ease of use, performance portability, interoperability, and reuse in climate, numerical weather prediction, data assimilation, and other Earth science applications. –It defines an architecture for composing multi-component applications and includes data structures and utilities for developing model components. –It defines standard function methods for organizing a model component. –It also defines an ESMF data structure (ESMF state) for communication between components. –There are two states in ESMF: the import state and the export state.

17. Page 17 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 Interoperation between ESMF models and SEPS Services Interoperations between ESMF models and SEPS services is done by exchanging and modifying the import and export states of the outmost component of a model It is also possible some intermediate result or export states may be accessed and exposed through standard SEPS Web service interfaces The exchange of states between the SEPS framework and ESMF has been implemented as an OGC WPS process –The WPS endpoint is deployed at http://data.laits.gmu.edu:8081/wps/WebProcessingService for the testing purpose. http://data.laits.gmu.edu:8081/wps/WebProcessingService

18. Page 18 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 State exchange between SEPS and ESMF

19. Page 19 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 Example of Applications in GEOSS Societal Benefit Areas – Climate Experiments have been carried out on interoperating with selected Earth science models. –The Community Atmosphere Model (CAM) was one of the selected models. Further information about the model can be found at http://www.ccsm.ucar.edu/models/atm-cam/. http://www.ccsm.ucar.edu/models/atm-cam/ –A prototype was developed to establish a data link between CAM and SEPS through standard geospatial interfaces by providing/retrieving Import/Export states to/from the ESMF model. The test page is available at http://data.laits.gmu.edu:8081/wps/ESMFDemo_test.html. http://data.laits.gmu.edu:8081/wps/ESMFDemo_test.html –This demonstration shows that it is possible to change the initial state of the ESMF models and retrieve the resulting states of the running ESMF models. This preliminary results helped us refine the architecture to build interoperation between SEPS and ESMF- enabled Earth system models.

20. Page 20 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 Example of Applications in GEOSS Societal Benefit Areas – Weather –Generalize detection and tracking concepts into a workflow architecture to enable parallel development of exchangeable and interacting components. –Update WCS and WFS capabilities, developed detection and tracking data models and implemented Web Processing Services (WPS) for prototype algorithms. –Formulate operational steps and verification methods to demonstrate automated detection of enhanced-V severe weather signature for future public watches and warnings. –Demonstrate technical feasibility for supporting sensor web targeting applications with cloud-top detection and tracking components utilizing GOES and MODIS IR radiance observations MODIS Couplet Detector and Correlation Matrix for the Enhanced-V Severe Storm Signature The Detection and Tracking of Satellite Image Features Associated with Extreme Physical Events for Sensor Web Targeting Observing

21. Page 21 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 Example of Applications in GEOSS Societal Benefit Areas – Disasters Flood and drought are two most significant types of disasters which happens every year, causing significant damage to agriculture. Two systems have been prototyped to provide near-real time disaster information and post-disaster damage estimation by integrating sensors with models –Global Agricultural Drought Monitoring and Forecasting System (http://gis.csiss.gmu.edu/GADMFS/)http://gis.csiss.gmu.edu/GADMFS/ Drought measure algorithms/ prediction models On-demand monitoring and forecasting –A Remote-sensing-based Flood Crop Loss Assessment Service System (RF-CLASS) (http://dss.csiss.gmu.edu/RFCLASS)http://dss.csiss.gmu.edu/RFCLASS Flood loss model driven by sensors Provide information for flood insurance and mitigation Multi-sensor near-real-time and historic data at different spatial and temporal resolutions have been used to drive the models and algorithms

22. Page 22 CSISS Center for Spatial Information Science and Systems, George Mason University 3/26/2013 Example of Applications in GEOSS Societal Benefit Areas – Agriculture Near real-time crop condition and progress information for large geographic area are very important for agriculture decision making. The technology described in this presentation has been integrated into the National Crop Condition and Progress Monitoring System (NCCPMS) –Feeding the near real-time remote sensing data to models to generate the crop condition and progress information.