Abstract:This paper describes the HISTALP database, consisting of monthly homogenised records of temperature, pressure, precipitation, sunshine and cloudiness for the 'Greater Alpine Region' (GAR,(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(43)(44)(45)(46)(47)(48)(49). The longest temperature and air pressure series extend back to 1760, precipitation to 1800, cloudiness to the 1840s and sunshine to the 1880s. A systematic QC procedure has been applied to the series and a high number of inhomogeneities (more than 2500) and outliers (more than 5000) have been detected and removed. The 557 HISTALP series are kept in different data modes: original and homogenised, gap-filled and outlier corrected station mode series, grid-1 series (anomaly fields at 1°× 1°, lat × long) and Coarse Resolution Subregional (CRS) mean series according to an EOF-based regionalisation. The leading climate variability features within the GAR are discussed through selected examples and a concluding linear trend analysis for 100, 50 and 25-year subperiods for the four horizontal and two altitudinal CRSs. Among the key findings of the trend analysis is the parallel centennial decrease/increase of both temperature and air pressure in the 19th/20th century. The 20th century increase (+1.2°C/+1.1 hPa for annual GAR-means) evolved stepwise with a first peak near 1950 and the second increase (1.3°C/0.6hPa per 25 years) starting in the 1970s. Centennial and decadal scale temperature trends were identical for all subregions. Air pressure, sunshine and cloudiness show significant differences between low versus high elevations. A long-term increase of the high-elevation series relative to the low-elevation series is given for sunshine and air pressure. Of special interest is the exceptional high correlation near 0.9 between the series on mean temperature and air pressure difference (high-minus low-elevation). This, further developed via some atmospheric statics and thermodynamics, allows the creation of 'barometric temperature series' without use of the measures of temperature. They support the measured temperature trends in the region. Precipitation shows the most significant regional and seasonal differences with, e.g., remarkable opposite 20th century evolution for NW (9% increase) versus SE (9% decrease). Other long-and short-term features are discussed and indicate the promising potential of the new database for further analyses and applications.

Long‐term in situ observations are widely used in a variety of climate analyses. Unfortunately, most decade‐ to century‐scale time series of atmospheric data have been adversely impacted by inhomogeneities caused by, for example, changes in instrumentation, station moves, changes in the local environment such as urbanization, or the introduction of different observing practices like a new formula for calculating mean daily temperature or different observation times. If these inhomogeneities are not accounted for properly, the results of climate analyses using these data can be erroneous. Over the last decade, many climatologists have put a great deal of effort into developing techniques to identify inhomogeneities and adjust climatic time series to compensate for the biases produced by the inhomogeneities. It is important for users of homogeneity‐adjusted data to understand how the data were adjusted and what impacts these adjustments are likely to make on their analyses. And it is important for developers of homogeneity‐adjusted data sets to compare readily the different techniques most commonly used today. Therefore, this paper reviews the methods and techniques developed for homogeneity adjustments and describes many different approaches and philosophies involved in adjusting in situ climate data. © 1998 Royal Meteorological Society

Abstract. The COST (European Cooperation in Science and Technology) Action ES0601: advances in homogenization methods of climate series: an integrated approach (HOME) has executed a blind intercomparison and validation study for monthly homogenization algorithms. Time series of monthly temperature and precipitation were evaluated because of their importance for climate studies and because they represent two important types of statistics (additive and multiplicative). The algorithms were validated against a realistic benchmark dataset. The benchmark contains real inhomogeneous data as well as simulated data with inserted inhomogeneities. Random independent break-type inhomogeneities with normally distributed breakpoint sizes were added to the simulated datasets. To approximate real world conditions, breaks were introduced that occur simultaneously in multiple station series within a simulated network of station data. The simulated time series also contained outliers, missing data periods and local station trends. Further, a stochastic nonlinear global (network-wide) trend was added.Published by Copernicus Publications on behalf of the European Geosciences Union. V. K. C. Venema et al.: Benchmarking monthly homogenization algorithmsParticipants provided 25 separate homogenized contributions as part of the blind study. After the deadline at which details of the imposed inhomogeneities were revealed, 22 additional solutions were submitted. These homogenized datasets were assessed by a number of performance metrics including (i) the centered root mean square error relative to the true homogeneous value at various averaging scales, (ii) the error in linear trend estimates and (iii) traditional contingency skill scores. The metrics were computed both using the individual station series as well as the network average regional series. The performance of the contributions depends significantly on the error metric considered. Contingency scores by themselves are not very informative. Although relative homogenization algorithms typically improve the homogeneity of temperature data, only the best ones improve precipitation data. Training the users on homogenization software was found to be very important. Moreover, stateof-the-art relative homogenization algorithms developed to work with an inhomogeneous reference are shown to perform best. The study showed that automatic algorithms can perform as well as manual ones.

Many long instrumental climate records are available and might provide useful information in climate research. These series are usually affected by artificial shifts, due to changes in the conditions of measurement and various kinds of spurious data. A comparison with surrounding weather-stations by means of a suitable two-factor model allows us to check the reliability of the series. An adapted penalized log-likelihood procedure is used to detect an unknown number of breaks and outliers. An example concerning temperature series from France confirms that a systematic comparison of the series together is valuable and allows us to correct the data even when no reliable series can be taken as a reference. Copyright 2004 Royal Statistical Society.

Abstract. The COST (European Cooperation in Science and Technology) Action ES0601: Advances in homogenization methods of climate series: an integrated approach (HOME) has executed a blind intercomparison and validation study for monthly homogenization algorithms. Time series of monthly temperature and precipitation were evaluated because of their importance for climate studies. The algorithms were validated against a realistic benchmark dataset. Participants provided 25 separate homogenized contributions as part of the blind study as well as 22 additional solutions submitted after the details of the imposed inhomogeneities were revealed. These homogenized datasets were assessed by a number of performance metrics including i) the centered root mean square error relative to the true homogeneous values at various averaging scales, ii) the error in linear trend estimates and iii) traditional contingency skill scores. The metrics were computed both using the individual station series as well as the network average regional series. The performance of the contributions depends significantly on the error metric considered. Although relative homogenization algorithms typically improve the homogeneity of temperature data, only the best ones improve precipitation data. Moreover, state-of-theart relative homogenization algorithms developed to work with an inhomogeneous reference are shown to perform best. The study showed that currently automatic algorithms can perform as well as manual ones.

The paper describes the development of a dataset of 192 monthly precipitation series covering the greater alpine region (GAR,(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18). A few of the time series extend back to 1800. A description is provided of the sometimes laborious processes that were involved in this work: from locating the original sources of the data to homogenizing the records and eliminating as many of the outliers as possible. Locating the records required exhaustive searches of archives currently held in yearbooks and other sources of the states, countries and smaller regional authorities that existed at various times during the last 200 years. Homogeneity of each record was assessed by comparison with neighbouring series, although this becomes difficult when the density of stations reduces in the earliest years. An additional 47 series were used, but the density of the sites in Austria and Switzerland was reduced to maintain an even coverage in space across the whole of the GAR. We are confident of the series back to 1840, but the quality of data before this date must be considered poorer. Of all of the issues involved in homogenizing these data, perhaps the most serious problem is associated with the differences in the height above ground of the precipitation gauges, in particular the general lowering of gauge heights in the late 19th century for all countries, with the exception of Italy. The standard gauge height in the early-to-mid 19th century was 15-30 m above the ground, with gauges being generally sited on rooftops. Adjustments to some series of the order of 30-50% are necessary for compatibility with the near-ground location of gauges during much of the 20th century. Adjustments are sometimes larger in the winter, when catching snowfall presents serious problems. Data from mountain-top observatories have not been included in this compilation (because of the problem of measuring snowfall), so the highest gauge sites are at elevations of 1600-1900 m in high alpine valley locations. Two subsequent papers will analyse the dataset. The first will compare the series with other large-scale precipitation datasets for this region, and the second will describe the major modes of temporal variability of precipitation totals in different seasons and determine coherent regions of spatial variability.

Ensembles used for probabilistic weather forecasting tend to be biased and underdispersive. This paper proposes a statistical method for postprocessing ensembles based on quantile regression forests (QRF), a generalization of random forests for quantile regression. This method does not fit a parametric probability density function (PDF) like in ensemble model output statistics (EMOS) but provides an estimation of desired quantiles. This is a nonparametric approach that eliminates any assumption on the variable subject to calibration. This method can estimate quantiles using not only members of the ensemble but any predictor available including statistics on other variables. The method is applied to the Météo-France 35-member ensemble forecast (PEARP) for surface temperature and wind speed for available lead times from 3 up to 54 h and compared to EMOS. All postprocessed ensembles are much better calibrated than the PEARP raw ensemble and experiments on real data also show that QRF performs better than EMOS, and can bring a real gain for human forecasters compared to EMOS. QRF provides sharp and reliable probabilistic forecasts. At last, classical scoring rules to verify predictive forecasts are completed by the introduction of entropy as a general measure of reliability.

Abstract. We review work on extreme events, their causes and consequences, by a group of European and American researchers involved in a three-year project on these topics. The review covers theoretical aspects of time series analysis and of extreme value theory, as well as of the deterministic modeling of extreme events, via continuous and discrete dynamic models. The applications include climatic, seismic and socio-economic events, along with their prediction.Two important results refer to (i) the complementarity of spectral analysis of a time series in terms of the continuous and the discrete part of its power spectrum; and (ii) the need for coupled modeling of natural and socio-economic systems. Both these results have implications for the study and prediction of natural hazards and their human impacts.