National and Kapodistrian University of Athens
Νέες μέθοδοι ανάλυσης και πρόγνωσης θαλασσίων κυμάτων
Abstract
dc:descriptionThe main objectives of this thesis were to study the factors affecting the predictability of sea wave models and to develop the appropriate methodology for the improvement of their predictive capacity to simulate sea surface state. The impact of local characteristics and seasonal variability in relation to the temporal horizon of the predictability were also investigated. In order to achieve these goals, the data assimilation algorithms were improved and new ones have been developed by including Kalman filters. Then, a detailed control of the sensitivity and performance of the new integrated forecasting sea waves system was made. The experimental procedure of studying the predictive ability of sea wave simulation models was composed of a series of accomplishments for a long time, in areas with different climatic characteristics. Furthermore, different methodologies of using the available measurements were used by the sea wave data assimilation schemes (spectral and altimeter), as part of the wave forecasting system. The results from the analysis performed, are summarized as follows: • Strong dependence - of the predictive ability of the wave model using altimeter data assimilation patterns - by the climatic characteristics of the area applied. In particular, wind-wave areas were little affected by assimilation, as the ocean wave measurements reflect mostly long-period waves (swell). Furthermore, the altimeter data assimilation scheme did not corrected the wind fields introduced by the atmospheric model, but only the significant wave height. In this way, it is not possible to correct possible systematic errors in a long horizon as well as discrepancies related to the propagation direction and the energy distribution spectrum of sea waves. • The number of available assimilated measurements and their correlation with the corresponding model results play an important role in the degree of improvement of the latter. More measurements and less correlation give more improvement. • Local-morphological characteristics of the sea area affect the performance of these simulation models. Restricted areas of complex shoreline, islands and well-separated parts result in limited impact and not spreading of the data assimilation corrections. • In each case, the impact of data assimilation decreases significantly with increasing temporal distance from the beginning of the forecast (analysis time). So there is little dependence of wave modeling results by the initial conditions. More important is the role of the imported - by the atmospheric model - winds. Thus any errors in the simulated winds, in extent and direction, lead to the reappearance of the pre-assimilation differences in the wave model results, especially if they are systematic. The analysis of the results from the sensitivity experiments that examine the impact of the spectral data assimilation on the model output, revealed the following: • Large utility of the wave-energy spectrum data, as, in addition to wave height, they contain information about the frequency and directional spectrum and, consequently, for many parameters that describe sea waves (e.g. wave period, direction and swell). This information, when assimilated by WAM, amends totally the wave field and improves considerably the results. • The main problem of this assimilation method is the relatively low availability of such measurements. Also, improvements are needed in the relevant data assimilation schemes, as there are problems with the covariance matrices used (since there is still not sufficient time series covering the whole earth). Moreover, improvement is required in the way used to separate and digest the various components of sea waves. • Finally, it is important to underline the limited duration of effect of spectral assimilation on wave forecasts, as in the assimilation of altimeter wave height. To overcome the fact that data assimilation improves the wave forecast model results only for a limited time, a combination of techniques was proposed: combination of mathematical filters and data assimilation. The use of Kolmogorov/Zurbenko and then Kalman filters onto the results of WAM in areas with stable flow of observations ensures: • Alignment of the observational time series with the ones of the model, by filtering high-frequency variations that are not represented by the latter. • Significant reduction of possible systematic deviations of the model results compared with the relevant observations and • Creation of improved time series, free from significant part of bias. Subsequently, these improved forecasts are used as "new data" during a second execution of the sea-wave simulation system, with the use of data assimilation into the forecasting period, in order to "extend" the improved results at the neighborhood, which probably shows similar behavior with the relevant observations. The results confirm the predictions for extending the influence of data assimilation throughout the forecasting period, as well as for reducing the size and variability of the differences between the final estimates and the relevant observations. The limitations on the application of this methodology can be summarized as follows: • It is required a continuous flow of measurements in the area of application. • A second model simulation is necessary so to exploit the advantages of data assimilation. Based on these findings, we attempted to address part of these problems by integrating the Kalman filter in the data assimilation scheme, in order to exploit the advantages mentioned above, even in real time. In this way, the ‘generated’ observations coming by Kalman filtering during the forecasting period, are actually improved predictions of WAM and, with the use of data assimilation, offer significant advantages: • Restrict important part of systematic errors. • The data assimilation scheme is then ‘spreading’ this improvement over a wider area than the one of the observations. • The use of Kalman filters, as part of the wave forecasting system, ensures consistency and continuity of its results with the physical laws describing the simulated wave system. • The duration of the effect extends even until the end of the forecasting period. The overall investigation of the new sea wave simulation and prediction system with assimilation of all types of available data (from satellites and ground stations) in conjunction with Kalman filters included a series of model simulations for different seasons, where the following were noted: • The greatest improvement was achieved for the transitional seasons (autumn and spring). • When using measurements from a single satellite, most important is the role of the number of measurements and of the small repetition cycle, rather than the higher spatial resolution. Similar are the results in extreme events. • In the occurrence of extreme events, the characteristics of the satellite data used in the assimilation do not affect the predictability. Further investigation of the themes reached in the present study could focus on the following issues: • Extension of the methodology of using Kalman filters at the assimilation of spectral data. The specific recommendation relates to better use of the information contained in such data and related to the distribution of the sea waves energy in various sea-wave components and to a number of parameters derived from them. • Explore the possibility for successful use of such filters in areas with non-systematic existence of measurements, by creating the necessary parameters for the filters (e.g. covariance matrices) by a climatic database created mainly by satellite data, with different values for each study area.
Degree
thesis:*- Grantor dc:publisher
- National and Kapodistrian University of Athens
- Year dc:date
- 2010
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Emmanouil, Georgios
- Εμμανουήλ, Γεώργιος
Subjects
dc:subject × 16- Θαλάσσια κύματα
- Σύστημα ανάλυσης και πρόγνωσης
- Αριθμητικά μοντέλα
- Αφομοίωση δεδομένων
- ΦΙΛΤΡΑ ΚΑΛΜΑΝ
- Δορυφορικά δεδομένα
- Sea waves
- Analysis and forecasting system
- Numerical models
- Data assimilation
- Kalman filters
- Satellite data
- Φυσικές Επιστήμες
- Γεωεπιστήμες και Επιστήμες Περιβάλλοντος
- Natural Sciences
- Earth and Related Environmental Sciences
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/26081