Research plan and timetable
Task 1. Setup of operational procedures for the HF radar network.
In collaboration with IPA IIIc HAZADR project, where two HF radars will be installed in the middle Adriatic area (including communication issues and acquisition of data), operational procedures will be developed following recent investigations on that topic (e.g., Cosoli et al., 2012). The proposed work within NEURAL project will include an assessment of the radial vector measurements, and computations of realistic hourly current fields that will be displayed in real time on the project webpage.
Operational procedures for HF radar network ready to use (month 4).
Task 2. High-level quality control of new and historic HF radar data.
High-level data quality control procedures that will be applied in delayed mode to the HF radar measurements, and will include different mathematical procedures which with an aim of detection, flagging and removing or interpolating the problematic values. Also, the same procedures will be used in re-assessment and re-computation of the historic HF radar data collected in the northern Adriatic through the NASCUM project (www.izor.hr/nascum
), altogether about 1 year of data series (August 2007-August 2008).
Historical hourly HF radar data from the northern Adriatic re-assessed through delayed-mode quality control procedures (month 6).
HF radar data from the operational network in the middle Adriatic assessed through automatic delayed-mode quality control procedures (months 12-24).
Task 3. Mesoscale reanalysis and forecast of the atmospheric fields by Aladin/HR and WRF-ARW models.
Operational Aladin/HR mesoscale numerical model will be run for two periods with HF radar measurements: (i) during August 2007-August 2008, when HF radar measurements were carried out in the northern Adriatic, and (ii) from installation and acquisition of high-quality HF radar data using procedures from Tasks 1 and 2 (at the beginning of the project) till month 24 of the project. The latter period will be split in two: the first 12 months will be used for training of neural networks, and the next two years will be used for estimation of model skill and for assessment of the hindcasting/forecasting system. Hourly surface fields, i.e., wind at 10 m, will be extracted from the model for the areas covered by HF radar measurements. The second input information will be provided by a high-resolution version of the non-hydrostatic research WRF-ARW model nested into the Aladin model and operating in real time in a research mode. The WRF-ARW coupled into Aladin would be carried out by the project Co-PI and part of her group doing high-resolution mesoscale modelling. The two dataset would be intercompared and validated with other available direct wind observations.
Aladin/HR surface model fields available for the northern Adriatic area (month 6) and middle Adriatic area (month 14) covered by HF radar measurements.
WRF-ARW surface model fields available for the northern Adriatic area (month 8) and middle Adriatic area (month 16) covered by HF radar measurements.
Task 4. Tuning the neural network parameters for optimal performance.
All collected and quality-checked HF radar data and model fields will be prepared for the Self-Organizing Maps (SOM) neural network algorithm. Model coverage of atmospheric models relevant for the HF radar measurements will be tested and agreed. Testing of SOM parameters, map size, lattice structure, neighbouring structure, will be done first on northern Adriatic HF radar data and joint HF radar + meteo data, to compare with existing publications (Mihanović et al., 2011). Then, the testing will be repeated for the middle Adriatic area. The neural network solutions with lowest quantization and topographic errors will be chosen for further analysis and operational forecasts.
Neural network parameters tested and the best parameters chosen, northern Adriatic site (month 9), middle Adriatic site (month 17).
Task 5. Obtaining characteristic patterns of surface currents and meteo model data through learning process.
Characteristic surface current and meteo patterns will be computed and visualised, first for the historical northern Adriatic site and then for the middle Adriatic site. Each data series will be split in two equal parts; the first half of the series will be used for training of the neural network and the second half of the series will be used for assessment and skill performance of the hindcast. The patterns will be associated to the major wind patterns and processes in both regions. Conditional statistical averaging for certain meteo parameters (e.g., winds) will be done and compared to the neural network solutions.
Characteristic patterns of surface ocean currents computed for both northern (month 10) and middle Adriatic HF radar sites (month 18).
Task 6. Assessing of the abilities and skill performance of the system.
Reliability of SOM patterns learned during the training process will be assessed by computing quantization error and topographic error for the second half of the data series, being a nonlinear measure of reproduction of the patterns by the SOM method. Also, complex correlation coefficients, root-mean-square errors and other skill parameters between the hindcast SOM patterns and the associated HF radar measurements and meteo fields will be estimated and assessed. An evaluation of the errors and skill parameters will be done, and, if not satisfactorily, the tuning of neural network parameters and HF radar and meteo data will be repeated.
Skill parameters of the modelling systems computed and assessed (months 12, 20).
Task 7. Setting up procedures and protocols for operational ocean forecast.
Once the skill performance of the system is at the satisfactorily level and ocean currents are satisfactorily forecasted, the procedures and protocols for operational maintenance and forecasts will be developed. These includes: (i) flow chart of the whole operational system, (ii) parameters and thresholds necessary for taking automatic decisions within the flow chart (e.g., threshold for reliability of forecasted surface current pattern), (iii) algorithms and procedures necessary for obtaining the forecast, (iv) scripts and programs which will automatically transfer the necessary operational Aladin/HR products to the NEURAL forecasting system.
Flow chart, procedures and protocols of the operational system developed (month 21).
Task 8. Creating dynamic web content with forecast of ocean currents.
Project web pages will be developed at the beginning of the project; however, visualisation of the surface current forecast, including design, storage and archiving of the forecasts will be developed. Dynamic scripts and programmes will be constructed for attractive visualisation of the acquired ocean forecast. The database containing HF radar measurements, meteo fields and associated ocean current and meteo neural network patterns will be developed and maintained.
Dynamic web content with forecasts of ocean currents created (month 23).
Task 9. Dissemination of project results.
Dissemination of the project information and achieved results is considered to be of the highest importance. The dissemination will be directed towards (i) national governmental authorities, operational services and civil protection agencies, and (ii) international programmes and projects. The first will be realised through organisation of a stakeholder workshop at the end of the project, and by informal meetings and communication. The latter will be done through targeted presentations at their annual and general meetings. In addition, all of these potential users and supporters will be invited to the Final Workshop, in which the project results will be presented in detail as well as reports on the applicability of the neural network ocean currents forecasting system in the frame of operational oceanography and weather forecasting. Also, further research and networking with the international researchers, research groups, projects and programmes will be proposed. Finally, the public outreach should not be forgotten - a series of open-access lessons and presentations will be given, in the frame of other science outreach activities (e.g., the Festival of Science organised in all of Croatia every year), and through different public media (tv, radio, newspapers, internet)
The information on project results and application possibilities will be systematically disseminated to potential users (months 12-24)
Stakeholder workshop organized (month 24)