No abstract

Stochastic weather generators are used in a wide range of studies, such as hydrological applications, environmental management and agricultural risk assessments. Such studies often require long series of daily weather data for risk assessment and weather generators can produce time series of synthetic daily weather data of any length. Weather generators are also used to interpolate observed data to produce synthetic weather data at new sites, and they have recently been employed in the construction of climate change scenarios. Any generator should be tested to ensure that the data that it produces is satisfactory for the purposes for which it is to be used. The accuracy required will depend on the application of the data, and the performance of the generator may vary considerably for different climates. The aim of this paper is to test and compare 2 commonly-used weather generators, namely WGEN and LARS-WG, at 18 sites in the USA, Europe and Asia, chosen to represent a range of climates. Statistical tests were selected to compare a variety of different weather characteristics of the observed and synthetic weather data such as, for example, the lengths of wet and dry series, the distribution of precipitation and the lengths of frost spells. The LARS-WG generator used more complex distributions for weather variables and tended to match the observed data more closely than WGEN, although there are certain characteristics of the data that neither generator reproduced accurately. The implications for the development and use of stochastic weather generators are discussed.

Spectral analysis of the sub-tropical percentage data (Fig. 4) indicates coherence with the 23±19-kyr oscillations (precession) just below the 80% con®dence interval. The effects of the seasonality and precipitation produced by the increased amplitude of the seasonal cycle of solar radiation appear to have had a signi®cant in¯uence on the distribution and composition of the sub-tropical vegetation in the Hungarian late Pliocene.The pollen spectrum of both taxonomic groups (that is, boreal and sub-tropical) is, however, dominated by a strong low-frequency component of ,124 kyr. Even though there are only at most three cycles to be resolved in this data set, such variance is clearly visible in the record (Fig. 3). This is of particular signi®cance, as the 124-kyr peak is non-existent in the calculated insolation forcing (Fig. 4) 15 .There is some evidence from oceanic records of signi®cant environmental change at the 95±124-kyr interval. Dust data in deep sea cores exhibit a strong response at this period in the time interval considered here 20 . Changes in atmospheric dust content are thought to re¯ect directly changes in continental aridity and terrestrial vegetation cover 21 , but until now there has been little direct pollen evidence to support this suggestion. The 124-kyr variance observed in our pollen record strongly supports the link between the dust content observed in marine cores and terrestrial vegetation change.The results from Pula thus indicate that in addition to forcing at the orbital frequencies of precession and obliquity, internally driven nonlinear responses of the climate system at a period of ,124 kyr were as important (if not more important) in driving terrestrial vegetation dynamics and were presumably associated with this broad-scale environmental change. This terrestrial sequence provides the basis for beginning to understand the physical relationships between vegetation, ice volume and insolation forcing during a critical period in the Earth's climate system. M

A 1961-1990 mean monthly climatology for a 'greater European' region extending from 32"W to 66"E and from 25" to 81"N has been constructed at a resolution of 0.S"latitude by 0.5" longitude for a suite of nine surface climate variables: minimum, maximum, and mean air temperature; precipitation totals; sunshine hours; vapour pressure; wind speed; and (ground) frost day and rain day ( > 0.1 mm) frequencies. This climatology has been constructed from observed station data distributed across the region. Station frequencies range from 936 (wind speed) to 3078 (precipitation). Over 95 per cent of these data are based on observations between 1961 and 1990 and over 90 per cent were supplied by individual national meteorological agencies (NMAs) on specific request. For four variables, some standardization of the data had to be performed because different countries supplied data under different definitions. Thus cloud cover had to be converted to sunshine hours, relative humidity to vapour pressure, air frost days to ground frost days and rain days > 1 mm to rain daysThe interpolation of the station data to the grid used elevation as one of the predictor variables and thus enabled three climate surfaces to be produced for each variable, reflecting the minimum, mean, and maximum elevation within each 0.5" by 0.5" cell. Subsets of stations were used for the interpolation of each variable, the selection being based on optimizing the spatial distribution, source priority and length of record. The accuracy of the various interpolations was assessed using validation sets of independent station data (i.e. those not used in the interpolation). Estimated mean absolute errors (MAE) ranged from under 4 per cent for vapour pressure to about 10 per cent for precipitation and up to 20 per cent for wind speed. The accuracy of the interpolated surfaces for minimum and maximum temperature was between 0.5"C and 0.8"C.We believe these results constitute the first climatology that has been constructed for this extensive European region at such a fine spatial resolution (0.5" by 0.5") from relatively dense station networks, for three different elevation surfaces and for a wide range of surface climate variables, all expressed with respect to a standard 30-year period. The climatology is already being used by researchers for applications in the areas of ecosystem modelling, climate change impact assessment and climate model validation, and is available from the authors.

Global warming is predicted to result in a rise in sea level which will lead to increased flood risk. Two other factors that will affect high water are the existing trends in mean sea level and changing tides. We illustrate here that in the Bay of Fundy and Gulf of Maine these two are related. An analysis of long-term sea level records shows that, independent of global warming related to climate change, sea level and tidal range have been increasing in this system. Our numerical model investigation indicates that recent changes in sea level, attributed in part to post-glacial rebound, are giving rise to increasing tides. The combined effects of modern sea level rise, global warming induced sea level rise, and the expanded tidal range they induce, will produce a significant increase in the high water level, much greater than that found when considering modern climate-induced sea level changes in isolation. We are predicting a dramatic increase in the risk of flooding at higher high water during the twenty-first century.RÉSUMÉ [Traduit par la rédaction] Le réchauffement climatique devrait entraîner une hausse du niveau de la mer, d'où l'augmentation des risques d'inondation. Mais deux autres facteurs sont à l'origine de cette hausse, soit les tendances actuelles liées au niveau moyen de la mer et au changement dans l'amplitude des marées. Dans notre étude, nous démontrons que ces deux facteurs sont interreliés dans la baie de Fundy et dans le golfe du Maine. L'analyse des relevés du niveau de la mer à long terme démontre que le niveau de la mer et l'amplitude des marées sont en hausse dans ce système, abstraction faite du réchauffement causé par les changements climatiques. L'étude que nous avons effectuée au moyen d'un modèle numérique fait ressortir que les changements survenus récemment dans le niveau de la mer, attribuables en partie au relèvement post-glaciaire, expliquent l'amplitude croissante des marées. L'effet de l'élévation actuelle du niveau de la mer, conjugué à celui de l'élévation du niveau de la mer provoquée par le réchauffement climatique et de l'amplitude croissante des marées que ces deux phénomènes induisent augmenteront considérablement le niveau des hautes eaux, beaucoup plus que ce qu'indiquent les conclusions des études sur les changements du niveau de la mer induits par le climat actuel pris isolément. Nous prévoyons une augmentation importante des risques d'inondation dans des conditions de pleine mer supérieure au cours du XXI e siècle.