JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 96, NO. C6, PAGES 10,739-10,764, JUNE 15, 1991 Wind Wave Cast in the Mediterranean Sea LUIGI CAVALERI, LUCIANA BERTOTTI, AND PIERO LIONELLO Istituto per lo Studio della Dinamica delle Grandi Masse, Consiglio Nazionale delle Ricerche, Venice, Italy An advanced, third-generation wave model has been repeatedly applied to the Mediterranean Sea. We have studied the accuracy of the results and the factors that control it. The grid resolution, when less than 0.5ø , is shown to have in general negligibleeffect in the open sea. The related accuracy of description of the coastal border can have a relevant effect on the wave field on the side and on the wake of the coastaldetails, for a distanceof a few grid points. The difficulty of correctly evaluating the wind in the Mediterranean basin is dramatically shown by comparing the results obtained using wind fields with different resolution. The crucial role of the orography in shapingthe wind fields and the need for higher-resolutionmodels is established. It is found that owing to the strong dependenceof wave height on wind speed, for resolution coarser than 70 km the loss of accuracy of the wind fields, even if meteorologically acceptable, makes them useless for an efficient evaluation of the wave conditions.The exception, explained by the local absence of relevant mountain ridges, is the case of a southerly wind in the eastern Mediterranean. The accuracy of wave forecasting is obtained by comparing analysis and forecastfieldsfor a very severestorm. In connection with the preferential west to east flow of most of the Mediterranean storms, and with the scarcity of information on the Atlantic Ocean, the equality of the meteorological forecastrapidly deterioratesbeyond 1 or 2 days of forecast. This leads to an even faster deterioration of the wave forecast, whose practical limits in the western Mediterranean can at present be establishedat 1 day. 1. INTRODUCTION We discuss herein the application of an advanced numer- ical wave model to the Mediterranean Sea. We study in particular the accuracy of the results, the elements that control their precision, and the possibilityfor forecasts. For this aim we have used a recently developed third-generation model, namely WAM (wave model). WAM is the product of the joint efforts of a group of scientists. Its declared principle is to approach the problem from the physical point of view, neglecting, as far as possible, any parameterizationof the physical processes. Still within the limits of our current knowledge, this approach makes the model flexible and capable of very good performances in a wide variety of meteorological conditions. In applying the wave model to a complex basin, we implicitly carry out a double test. On the one hand we test the WAM model on complicated geometrical and meteoro- logical patterns. After the previous extensive tests carried out in the open ocean, this is an obvious step toward the coastal areas, where most economic interests are located. On the other hand, with a reliable wave model at hand, we can test the accuracy of the input and the associatedsensi- tivity of the output. As the wave estimateis the product of both wind and wave modelsworking in series, in each applicationit is difficult to decide which of these models might be blamed for the errors in the output. We have approached this point by usingwind fields from different sources for the same storm. Keeping all the other conditions the same, the differences in the outputs reflect uniquely those in the inputs. The Mediterranean Sea is very attractive for such a test. Historically and politically relevant, it has been sailed for more than 40 centuries when ship traffic was practically interrupted 6 months a year during winter storms. During Copyright 1991 by the American GeophysicalUnion. Paper number 91JC00322. 0148-0227/91/91J C-00322505.00 this extendedperiod of time a heavy burden has been paid by sailors. Bascom [1976, p. 5] estimates that more than 30,000 ships lie at the bottom of the Mediterranean. Even nowa- days, more than six large ships are lost on the average every year [Bascom, 1980, p. 160]. A closedbasin large enoughfor substantialstorms to develop, the winter Mediterranean sea is quite different from the ideal image often reported in the international literature. We begin this paper (section 2) with a description of the basin, its geographicalcharacteristics, and the main systems of winds and meteorological patterns. This is followed (sections 3 and 4) by a compact description of the WAM model, of its equations and numerical schemes, and of the implementations in the Mediterranean basin. We then turn our attention to the meteorologicalinput sources and related wind fields, described in section 5. The introductory part is completed by the description of the three storms usedfor the test (section 6). The main body of the paper follows in the next three sections. In section 7 we analyze the influence of the grid resolution on the final results. For this we use a single set of input wind fields and two different resolutions for the WAM model. In section 8 we carry out the other side of the test; that is, we use different input wind fields for the same storm. The analysisof the resulting differencesin the outputs shows the crucial role of the accuracy of the input wind fields. The test is completedby two more storms that help to identify the reasons for the former discrepancies. In section 9 we consider the problem of the predictability of a storm, and we discuss the reasons for its limitations. The results are sum- marized and discussed in section 10. The definition of the statistics used throughout the paper is given in the appendix. 2. THE MEDITERRANEAN SEA The Mediterranean Sea (Figure 1) extends longitudinally for almost 3800 km from a 6øW (Strait of Gibraltar) to the coasts of Syria and Israel, at 36øE. In the north-south direction it is bounded by Libya at 30øN to the south and at 10,739