...

Info
icon false title Table of Contents
Table of Contents maxLevel 5

Preamble

History of modifications to this User Guide

List of versions covered by this document

Acronyms

Introduction

Spatial climate analyses that extend back over many decades are an important basis for monitoring climate variations and long-term change (e.g. van der Schrier et al., 2013). They also serve as input for modelling environmental systems (e.g. ecosystems and glaciers, Kittel et al., 2004), and for calibrating climate reconstructions with proxy data (tree rings, Frank & Esper, 2005).

...

Data

The construction of LAPrec builds on two data components, namely a dataset of high-quality long-term station series that contributes information on long-term variations, and a dataset of high-resolution grid data that enhances information on spatial variability. The two data components are described in the following.

Long-term station dataset

HISTALP is an international data collection consisting of monthly homogenised climate records for the greater Alpine region (Auer et al., 2007). To meet the basic requirements of climate variability and change studies, it aims to provide instrumental data that extend far back in time, are dense in space, quality-improved, homogenized, multiple and user-friendly. The station series were subject to an intensive homogenization procedure. The homogenization method HOMER (Mestre et al. 2013) was used to check the homogeneity of the stations data. For the detection of breaks, the pairwise detection algorithm was applied. This algorithm is similar PRODIGE (Caussinus and Mestre 2004) and uses a maximum likelihood approach. A break signal was only considered a break if at least 50 % of the available reference stations detected it. For correction an ANOVA is used. Precipitation is a multiplicative parameter in regards to homogenization, meaning that the adjustment is a monthly factor that is multiplied to the original value. At least five reference stations have been needed, but all stations with a correlation of at least 0.7 have been used. Only in a small number of cases the five stations of such a high correlation have not been available. The homogenization of the stations has been done in groups of climatic zones in the HISTALP-GAR (Greater Alpine Region) area, to ensure the use of appropriate reference stations. As the case occurred that quite a number of stations had missing data within the same time period or didn't have any reference stations at the beginning of the time series, special networks where created for those stations in order to be able to find the best possible solution for them. As far as possible, found break points where shifted according to appropriate metadata information, but the small number of available metadata for some of the stations left quite a number of breaks unexplained.
The number of available monthly precipitation series rises continuously from 5 in 1800 to 40 in 1853 (Fig. 1). With the foundation of national meteorological institutes in the 1850s and 1860s the network quickly becomes denser and encompasses 246 stations by 1901. The high number of around 250 series is maintained until the most recent years when data collection and updating of data still are in progress.

...

Anchor
figure2 figure2

Figure 2: Map showing the stations utilized with the two periods, starting in 1871 (blue) and additionally starting in 1901 (green).

High-resolution grid dataset

Anchor
figure3 figure3

Figure 3: Mean annual precipitation (mm per year) for the period 1971-2008 in the high-resolution grid dataset APGD.

The Alpine precipitation grid dataset (APGD, Fig. 3) is a high-resolution grid dataset of daily precipitation covering the entire Alpine region. It builds on efforts initiated 20-years back in compiling and jointly analysing data from all the high-resolution rain-gauge networks in the region (Frei and Schär 1998). MeteoSwiss has re-established the APGD as part of the EURO4M project (Isotta et al. 2014). APGD incorporates more than 5500 rain-gauge measurements on average per day for the Alpine sections of Austria, Croatia, France, Germany, Italy, Slovenia and Switzerland. With 10-15 km station spacing, the dataset is one of the densest in-situ observation networks over a high-alpine topography worldwide. The procedure of spatial analysis uses local regression (PRISM) to incorporate physiographic influences on precipitation and angular distance weighting (SYMAP) for daily anomalies (Isotta et al. 2014). Currently, APGD is a static (not updated) dataset spanning the period 1971-2008. Recently, in the UERRA project, a probabilistic version of the APGD has been developed, providing ensembles of area-average precipitation over more than 500 hydrological catchments in the Alps. APGD is freely available upon registration and for non-commercial use at (http://www.meteoswiss.ch).

Method and Settings

The reconstruction method adopted to derive LAPrec is denoted as "Reduced Space Optimal Interpolation" (RSOI). It combines information from the primary long‐term precipitation series with statistical information distilled from a high‐resolution gridded analysis. In essence, RSOI establishes a linear model between station data and high-resolution grids, calibrated over a period when both are available. Technically, RSOI involves a Principal Component Analysis (PCA) of the high-resolution grid dataset, followed by an Optimal Interpolation (OI) using the long-term station data. Here, we summarize these two steps briefly and then detail the settings used in the present application. Detailed technical descriptions of RSOI are given in Kaplan et al. (1997), Schmidli et al. (2001), and Schiemann et al. (2010). Further insight into the potential of this method can be found in Masson & Frei (2016) and Isotta et al. (2019).

...

In our application for LAPrec, the OI model is established, again, over the period 1971-2008, and using square-root transformed, rather than original precipitation data. As for the configuration of the number of scores retained (truncation L, see above), our choice was L =27 in the case of the long reconstruction LAPrec1871 (starting in 1871, 85 stations) and L=50 for the short reconstruction LAPrec1901 (starting in 1901, 164 stations). These settings were inferred by cross-validation experiments. The accuracy of the reconstruction does barely improve with settings beyond these marks. It is worth mentioning that the results are not overly sensitive to details of the truncation settings. The inherent penalty for variance in high-order modes limits the risk of overfitting with undue values of L.
Technically, the reconstruction is implemented in R (R Core Team 2012). It is computationally rather inexpensive for the 5-km grid over the Alps, the calculation being completed within less than 10 seconds on a 2.8 GHz core, including the PCA.

Production and Access

The LAPrec dataset is provided in NetCDF format, following the CF-1.6 convention (see http://cfconventions.org/cf-conventions/v1.6.0/cf-conventions.html). Besides the monthly precipitation fields, fields of geographical variables (longitude, longitude and elevation above sea level) and details on the coordinate reference system are included. The fields consist of a regular 5 km x km grid that is spatially referenced in ETRS89 / ETRS-LAEA (European Terrestrial Reference System 1989-Lambert Azimuthal Equal Area, see https://spatialreference.org/ref/epsg/3035/).
Two variants of the dataset exist reflecting the two different starting dates (January 1871 and January 1901) and respective different station numbers (currently 85 and 164). Updates of the dataset might change the exact number of ingoing stations but keep three criteria for missing data defined in chapter 1.1. as well as the starting years 1871 and 1901. Homogenised station series data (no deliverable of this contract) are available from the HISTALP database of ZAMG and according files. In principle, the station series are updated annually. Due to national issues, the timing of the update varies within the year.
Versioning of the LAPrec dataset is in the format vX.Y where an increment of X denotes methodical adaptations and an increment in Y denotes updates in input station data. The file names consist of the dataset acronym, the starting year and the version number, e. g. LAPrec1871.v1.0.nc and LAPrec1901.v1.0.nc.

Results

Reconstruction examples

The LAPrec datasets permit to analyse fields of monthly precipitation sums in the remote past, when only few stations were available. Figure 4 depicts examples of monthly sums in June 1876, October 1907 and June 1910, when heavy precipitation events occurred in the Alpine region. The examples illustrate how RSOI introduces small-scale structures (via the APGD grid dataset) that are not resolved explicitly by the low density station measurements (dots in Fig. 4).

...

Figure 5: APGD (first column) and reconstructions (LAPrec1901, second column and LAPrec1871, third column) of the monthly precipitation in December 1974 and June 2008, in mm/month. Note that reconstructions and references are made strictly independent, by excluding, in turn, data from each test month from the calibration (e.g., the direct interpolation of a test month was not available for the PCA and OI).

Trends

Anchor
figure6 figure6

Figure 6: Linear trend in yearly and seasonal precipitation in % change per 10 years relative to the yearly and the seasonal mean precipitation, respectively, in the period 1981–2010. Results are shown for the RSOI‐based spatial reconstructions (fields) and for the sets of long‐term stations used in the reconstruction (colored dots). Trends are shown for three periods: (first column) 1864–2017, (second column) 1901–2017, and (third column) 1961–2017. Statistically non-significant trends are displayed with stippling. Trends are estimated with the Theil‐Sen slope. Statistical significance is assessed with the Mann‐Kendall trend test, followed by a Benjamini‐Hochberg meta‐test with a critical false discovery rate of 0.05 (for detail, see section 5.2).

Most yearly and seasonal trends of precipitation over the whole available period of LAPrec1901 and LAPrec1871 are statistically not significant. The one with at least a small significant region in the Alps are shown in Figure 6, namely the yearly (left) and the winter (right) trends from 1871-2017. In the yearly analysis, precipitation is increasing with amounts of about 1% per decade with respect to the mean in the period 1981-2010 in the north-east part of France, in north-east Switzerland, west Austria and southern Germany. For winter, the whole area northern of the main Alpine ridge has a trend around 2% per decade.
Trends are estimated using the Theil‐Sen slope (Sen, 1968; Theil, 1950) and the statistical significance is assessed based on the Mann‐Kendall test (Kendall, 1975; Mann, 1945) applied individually to all grid points and a Benjamini‐Hochberg meta test of the pertinent p values (Benjamini & Hochberg, 1995). The latter controls the false discovery rate at 5 %.

User Guidance

The construction of LAPrec deviates from the principles commonly used in the generation of spatial climate datasets. The underlying station dataset encompasses (almost) continuous time series only, with the compromise of a much coarser spatial density. The stationarity of the input data avoids disturbances of temporal consistency that have to be expected when station series start or end, or when the number of input data gradually varies over time. Such inconsistencies have been documented in datasets that are produced, more conventionally, by integrating all available information at any particular time (Hofstra et al. 2010; Begueria et al. 2016; Frei 2014; Isotta et al. 2019).

...

• Combination with other data: The strength of LAPrec is its length and reliable representation of temporal variations. The accuracy for time-mean precipitation on a local scale is, however, limited. If, for a particular application, there is reliable complementary information about mean precipitation, for instance from local measurements, it can be valuable to merge LAPrec with this complementary information. We suggest that the merging is constructed such that the final result preserves relative anomalies from LAPrec, i.e. with a multiplicative correction. Except from such an adjustment of means, we discourage combinations with other datasets. Most importantly, we disadvise replacing recent years in LAPrec with other spatial analyses, for instance, because of their higher resolution or better local accuracy. For example, the two variants of LAPrec should not be mixed together. There is a risk of disturbing the temporal consistency, which was the main motivation for the present development.

Conclusion

This user guide and technical report describes the two reconstruction grid datasets LAPrec1871 and LAPrec1901 for the European Alps. They are covering more than a century of monthly precipitation sums and are an important basis for monitoring climate variations and long-term change, satisfying high standards in climate consistency.

...

LAprec was developed in the framework of the Copernicus Climate Change Service (Contract C3S_311a Lot4) in a collaboration between ZAMG (Zentralanstalt für Meteorologie und Geodynamik, Austria) and MeteoSwiss (Federal Office of Meteorology and Climatology, Switzerland). In the same Contract, LAPrec was used as reference for a comparison with other grid datasets (E-OBS ERA5) and the RSOI method was applied for reconstruction in Fennoscandia ("NCGD_rec").

LAPrec data access

LAPrec, for long-term Alpine precipitation reconstruction, is a gridded precipitation dataset that extends back till 1871 and was specifically constructed to satisfy high standards in climate consistency. The dataset has monthly resolution and covers the mountain range of the European Alps with territory from eight countries. The development is a collaboration between ZAMG (Zentralanstalt für Meteorologie und Geodynamik, Austria) and MeteoSwiss (Federal Office of Meteorology and Climatology, Switzerland).

...

• a set of high-quality station series, taken from the HISTALP station data archive (e.g.Auer et al. 2007, see citation section). HISTALP is an initiative of the national weather services in the Alpine region, led by ZAMG, to assemble and analyze high-quality climate series in the European Alps and extends over the full period of interest, continuously without gaps. It informs about the temporal variation over the long term.
• a high-resolution spatial analysis, the Alpine Precipitation Grid Dataset APGD (Isotta et al. 2014, see citation section) developed at MeteoSwiss. APGD covers a few recent decades only but builds on data from thousands of rain-gauges. This component enriches the final result with spatial detail that is not resolved by the long-term stations alone.
  LAPrec is a cross-product that brings together the merits of the two initiatives in terms of long-term extent / consistency (HISTALP) and spatial resolution (APGD).

Actual version, user guide and update policy

Notice that the last six years of the dataset are provisional: Some stations may not be available in this period and the gaps are therefore filled.

For information on the current version, please see "List of versions covered by this document" at the beginning of the document.

Citation

All users of data uploaded on the Climate Data Store (CDS) must provide clear and visible attribution to the Copernicus programme and are asked to cite and reference the dataset provider. Please refer to How to acknowledge, cite and reference data published on the Climate Data Store for complete details.

...

Reference:
Isotta FA, Begert M, Frei C. 2019: Long-term consistent monthly temperature and precipitation grid data sets for Switzerland over the past 150 years. Journal of Geophysical Research: Atmospheres, 123. https://doi.org/10.1029/2018JD029910

Data providers

DWD (www.dwd.de)

OMSZ (www.omsz.hu)

...

ARPA Trentino (https://www.meteotrentino.it/)

References

Auer I., Böhm R., Jurkovic A., Orlik A., Potzmann R., Schöner W., ... Mercalli L. (2005). A new instrumental precipitation dataset for the Greater Alpine Region for the period 1800–2002. Int. J. Climatol., 25, 139–166. http://doi.org/10.1002/joc.1135

...

 Copernicus Climate Change Service

Info
icon false
This document has been produced in the context of the Copernicus Climate Change Service (C3S). The activities leading to these results have been contracted by the European Centre for Medium-Range Weather Forecasts, operator of C3S on behalf of the European Union (Delegation agreement signed on 11/11/2014). All information in this document is provided "as is" and no guarantee or warranty is given that the information is fit for any particular purpose. The users thereof use the information at their sole risk and liability. For the avoidance of all doubt , the European Commission and the European Centre for Medium - Range Weather Forecasts have no liability in respect of this document, which is merely representing the author's view.

Related articles

Content by Label
showLabels false max 5 spaces CKB showSpace false sort modified reverse true type page cql label in ("precipitation","in-situ") and type = "page" and space = "CKB" labels era-interim