9 Jan 2017 THE PERMANENT SERVICE FOR MEAN SEA LEVEL: FURTHER INFORMATION Since 1933, the Permanent Service for Mean Sea Level (PSMSL) has been responsible for the collection, publication, analysis and interpretation of sea level data from the global network of tide gauges. It is based at the National Oceanography Centre, Liverpool, United Kingdom and is a regular member of the International Council for Science - World Data System (ICSU-WDS). It is supported by the U.K. Natural Environment Research Council. The PSMSL reports to the International Association for the Physical Sciences of the Oceans (IAPSO) Commission on Mean Sea Level and Tides (President Prof. Gary T. Mitchum, USA). It also a service of the the International Association of Geodesy (IAG) and works with the Global Sea Level Observing System (GLOSS) and the Global Geodetic Observing System (GGOS). More information about these organisations can be found via http://www.icsu-wds.org/ http://iapso.iugg.org/ http://www.iag-aig.org/ http://www.gloss-sealevel.org/ http://www.ggos.org/ DESCRIPTION OF PSMSL 'RLR' AND 'METRIC' DATASETS The database of the Permanent Service for Mean Sea Level (PSMSL) contains monthly and annual mean values of sea level from almost 2000 tide gauge stations around the world. (Note that, in common with most other climate variables, by 'monthly' we mean CALENDAR monthly means.) The PSMSL receives monthly and annual mean values of sea level from almost 200 national authorities, distributed around the world, responsible for sea level monitoring in each country or region. Data from each station are entered directly as received from the authority into the PSMSL raw data file for that station (usually called the METRIC file in PSMSL publications). The monthly and annual means so entered for any one year are necessarily required to be measured to a common datum, although, at this stage, datum continuity between years is not essential. While the PSMSL makes every attempt to spot inconsistent or erroneous data, the responsibility for the monthly and annual means entered into the METRIC files in this way is entirely that of the supplying authority. A description of data checks routinely made by the PSMSL is given below and in Woodworth, Spencer and Alcock (1990) and IOC (1992). In order to construct time series of sea level measurements at each station, the monthly and annual means have to be reduced to a common datum. This reduction is performed by the PSMSL making use of the tide gauge datum history provided by the supplying authority. To date, approximately two thirds of the stations in the PSMSL database have had their data adjusted in this way, forming the 'REVISED LOCAL REFERENCE' (or 'RLR') dataset. For scientific purposes, the RLR dataset is normally superior to the 'METRIC', although the latter, which contains the total PSMSL data holdings, can also be analysed bearing in mind the above datum continuity considerations. (See below for further comments on METRIC and RLR differences). IN GENERAL, THOUGH, ONE SHOULD ONLY USE THE RLR DATASET FOR TIME SERIES ANALYSIS. The RLR datum at each station is defined to be approximately 7000mm below mean sea level, with this arbitrary choice made many years ago in order to avoid negative numbers in the resulting RLR monthly and annual mean values. The detailed relationships at each site between RLR datum, benchmark heights, tide gauge zero etc. are not normally required by analysts of the dataset, but is available from the individual station pages. The contents of the PSMSL dataset used to be described in 'Data Holdings of the PSMSL' printed reports, of which the last was Spencer and Woodworth (1993). Regular updates are now made via the web. Further information about the PSMSL, and about the spatial and temporal distribution of the PSMSL data set, can be found in Holgate et al. (2013) and earlier journal papers referenced therein. The detailed current contents of the PSMSL databank are described via files accessed from the PSMSL web page: http://www.psmsl.org/ In particular, this PSMSL home page points to http://www.psmsl.org/data/obtaining/ which describes how to access the MSL data sets. The primary identifier in the PSMSL database is the station ID. This is a unique integer that is assigned to each tide gauge site and will not change going forward. All files (both data files and pages on the website) will use this unique identifier to refer to a station. Prior to 2010, stations were referred to by a three digit coastline code (previously called country code) and a three digit station code. These have be retained for continuity. As an example of coastline codes, Iceland is defined as '010', followed by Jan Mayen ('012') and the Faeroe Islands ('015'), and then progressing around the world coastline in essentially an eastward direction until Greenland ('980'). Antarctica ('999', a change from the previous value of 'A ') finishes off the list. As the list progresses through the various oceans and seas it also includes the nearby island gauges. Separate coastline codes are used for countries abutting more than one ocean or sea, e.g. France (Atlantic) and France (Mediterranean). While previously called a "country" code, the value had no political significance, but was used in a geographical subdivision sense only. However, while a change in country name would not impact this scheme, it is necessary to introduce a new coastline code when a country splits. Thus, this scheme is no longer used as the primary identifier of a tide gauge site. We note that our use of "country" above with regard to a particular tide gauge site does not imply the expression of any opinion whatsoever concerning the legal status of any country or territory, or of its authorities, or concerning the delimitation of the frontiers of any country or territory. Station pages provide most of the station information. (NOTE: replace the #STATION ID# below with the appropriate number) http://www.psmsl.org/data/obtaining/stations/#STATION ID#.php Each station page shows the station ID, location (latitude and longitude), GLOSS ID (if applicable), coastline code, station code, country, and frequency code. In addition, there is a time span and completeness of the data. The "date of last update" is the last time the record was altered in the database. If a quality flag exists for the station, a warning will be displayed advising further inspection of the station documentation. Frequency codes are defined as follows: Code Meaning Integer n n measurements/day C Integration from continuous recording HL Mean of high and low waters (i.e. Mean Tide Level) Below this station information is a data section. Plots of the monthly and annual data are shown if the station is RLR. Links to the data for this station are in this section. A documentation section follows the data section. This will include a link to the RLR information, if appropriate, as well as any comments relevant to the station. Finally, the authority information is listed, along with any authority comments that are relevant. For some of the longer sea level records, different authorities may have been responsible for the operation of the tide gauge and the analysis of its data at different times. The authority-codes shown here refer to the authority most recently responsible. A change of authority is indicated within the documentation embedded with the PSMSL data files, while a full history of the operation of a particular tide gauge can be obtained from the PSMSL. We also have generated a data catalogue (catalogue.dat) which details the entire holdings of the PSMSL. The stations are ordered first alphabetically by country, territory or area name (starting with ÅLAND ISLANDS), then grouped by coastline code, and finally listed within each group by station code. Each station entry starts with station code, station name, latitude, longitude, authority code/frequency code (as described above), and GLOSS ID (if applicable). Each entry is then finished with separate lines for the Station ID, metric data time span, metric completeness, RLR time span, and RLR completeness. A second file (nucat.dat) condenses the information for each station down to one line. Each station line of this file contains the PSMSL ID, station code, station name, latitude, longitude, authority code, frequency code, GLOSS ID, metric data time span, metric completeness, RLR time span, and RLR completeness. These files can be found at http://www.psmsl.org/data/obtaining/catalogue.dat http://www.psmsl.org/data/obtaining/nucat.dat FURTHER COMMENTS ON SEVERAL RLR RECORDS FROM A STATION You will see that several stations contain more than one PSMSL RLR record. RLR records are kept separate if, for example, they are from completely separate sets of measurements from different locations in the area of the station. Another reason is if the benchmark datums of the separate records epochs cannot be geodetically connected. A further reason is if one section of data is MSL and a further set is MTL. However, there are a small number of pairs of RLR records which can be combined into composites, with some reservations about care in using them because of, perhaps, long periods of time between them. These include: Aberdeen I (ID: 361) which is MSL and Aberdeen II (ID: 21) which is MTL but which are geodetically connected to a common benchmark Venezia (San Stefano) (ID: 39) and Venezia (Punta della Salute) (ID: 168) in Venice, which are from separate sites but which are geodetically connected to a common benchmark. North Point (ID: 333) and Quarry Bay (ID: 1674) in Hong Kong, which are from quite separate sites in areas of reclaimed land but which are geodetically connected to a common, if somewhat distant, benchmark. Auckland-Waitemata Harbour (ID: 217) and Auckland II (ID: 150) from Auckland, New Zealand, the former being provided by the RNZN Hydrographic Office, although gappy, and the latter being provided by the work of Prof. John Hannah and more complete but not up-to-date. Port Stanley (ID: 1082) and Port Stanley II (ID: 1796) in the Falkland Islands (Malvinas) which are geodetically connected to a common benchmark but are separated by a large period of time In each of these cases, the brief documentation should be consulted with regard to the construction of composite time series. MIXED MTL AND MSL DATA Above were given a few examples of MTL and MSL records that had been kept separate. Unfortunately, however, there are some examples where the two data types have been combined into a single record. In some of the more extreme cases, the annual average difference between these two mean levels can be over 10 cm, which could introduce an artifact into estimates of the long-term trends. In an attempt to make these combined records more transparent, and to cause the minimum disruption to the current set of records, we have introduced a flag 010 (described further in DATA FORMATS below) indicating MTL values in a MSL record. In addition, we have applied an estimate of the annual average difference (MTL-MSL) to the RLR time series, so that these values should be more directly comparable with the rest of the MSL records. Note that the metric files have the MTL values flagged but do not have the correction applied. Caveats: For the differences between MTL and MSL, we are generally using results provided by Philip Woodworth in conjunction with his manuscript on the subject. The difference does have a seasonal component, and we are only using an annual average for the correction. Thus, the seasonal portion of the MTL and MSL values still cannot be compared, even after we have applied the correction. Typically, we are using recent data to estimate a correction applied to data in the 1800s or early 1900s. Thus, we are explicitly assuming that the difference has not changed in time. In some cases, there is not any high-frequency data needed to conduct an analysis. In others, we have reason to believe that the correction is highly uncertain (at greater than the 1 cm level). Both cases will be addressed through flagging described below. The zip files contain a file (mtl_msl_corrections.csv) that lists the Station ID, the start and end of the MTL data within the time series, and the value of the MTL-MSL estimate applied to the data [applied as MTL value – (MTL – MSL estimate) = MSL estimate]. This should enable one to easily remove the correction [MSL estimate + (MTL – MSL estimate) = original MTL value] and substitute your own if desired. The flag also would enable one to ignore the MSL estimates values completely in the analysis if desired. If we have no estimate of the MTL-MSL difference, the value will be set to -99999 in this file, but will effectively be 0 in the data file, i.e. the MTL value will remain unchanged in the distributed data file, but will have the flags 011. In the case of the flags 011 and a value not equal to -99999, this indicates that we have reason to believe that this value may be uncertain to greater than 1 cm. A note will be placed in the documentation file as to why the ‘flagged for attention’ was set. We would recommend that values with the flags 011 not be used in analysis of long-term trends. FURTHER COMMENTS ON THE USE OF 'METRIC' DATA Without the provision of full benchmark datum history information, records will remain as 'Metric only' in the databank and not as 'RLR'. It is a good general rule, therefore, that 'Metric' records should NEVER be used for time series analysis or for the computation of secular trends; without datum continuity their only use is in studies of the seasonal cycle of mean sea level. If there is any doubt about the datums for a particular record, the PSMSL would be pleased to supply clarification. There are, however, some 'Metric only' records which almost certainly can be used for time series work, even though the PSMSL does not have full benchmark datum histories. They include, in particular, a number of German 'Metric' records which are measured with respect to Normal Null (NN). Any such information is included in the station comments in the relevant documentation sections of the data sets supplied by the PSMSL. Even though these records are expressed relative to the national levelling systems, they are, in effect, relative to a local level as required for RLR purposes i.e. the records do not (as far as we know) contain datum shifts contributed by re-levelling adjustments. In general, however, measurements relative to national levelling systems may well reflect such adjustments, which explains why the PSMSL has traditionally steered clear of classifying such data as RLR. We have relaxed this classification slightly for a few of the German stations. In the late 1990s and early 2000s, the PSMSL received data from these stations that could be related to a tide gauge benchmark as well as NN. Using these relationships, we have reclassified these entire time series as RLR. In the past, the PSMSL also included the the Netherlands data in the above category of Metric records acceptable for time series work. These records are expressed relative to the national level system Normaal Amsterdamsch Peil (NAP). However, a recent re-levelling of NAP in 2005 introduced a small datum shift for the tide gauge time series. In order to maintain utility of these long records, we have reclassified most of the Netherlands records as RLR and introduced different RLR factors for the periods before and after 2005. While these records do not meet the strict definition of RLR and may still include prior re-levelling adjustments, we believe this represents the best path forward. ELLIPSOIDAL LINKS FOR RLR DATA PSMSL are able to publish estimates of the height of our RLR data above a reference ellipsoid (GRS80) in some cases, where a permanent GNSS station has been installed near the tide gauge, and the two instruments have been connected through levelling. In these cases the link is listed on the station's RLR diagram page. Estimates are provided by Système d'Observation du Niveau des Eaux Littorales (SONEL). Full details are included at the page http://www.psmsl.org/data/obtaining/ellipsoid.php The zip files contain a file (ellipsoidal_links.csv) that lists the stations for which such a link has been made. DATA FORMATS The page http://www.psmsl.org/data/obtaining/notes.php gives detailed information on the formats of the data sets. Prior to 2010, the entirety of the PSMSL database had be distributed in condensed formats (e.g., psmsl.dat and rlrann.dat). However, with the update of the backend database, maintaining these formats was no longer possible. Data is now distributed as individual time series for each station. The entire dataset (including documentation) is distributed in zipped files, described further below. Separate time series files exist for the RLR monthly mean data, the RLR annual mean data, and the Metric monthly mean data (same strong warnings as above regarding the Metric data apply). Separate directories for each data type exist: http://www.psmsl.org/data/obtaining/rlr.monthly.data/ http://www.psmsl.org/data/obtaining/rlr.annual.data/ http://www.psmsl.org/data/obtaining/met.monthly.data/ Each directory contains a file list (filelist.txt) consisting of the station ID, latitude, longitude, station name, coastline code, station code, and quality flag. These files can be used translate between the new station ID and the old coastline/station codes. The quality flag has a value of "N" or "Y", where "Y" indicates that the documentation files should be consulted. More information on the quality flag is given three paragraphs below. Details of the data formats can be found on the notes.php page listed above. It should be noted that the time series file have internal flags for each data point. For example, the monthly files have an entry that indicates how many days were missing in the average. The annual file has an entry ("N" or "Y") that indicates, in the case of "Y", that either one month was missing or that at least 30 days were missing. (The annual average is dropped if more than one monthly value is missing. A monthly value is calculated as long as 15 or less days are missing in a month.) In both cases, a quality flag is present. A value of "001" indicates that the flag is present (see below). Each of the above directories will also contain the appropriate zip file: rlr.annual.zip, rlr.monthly.zip, or met.monthly.zip. These zip files will contain filelist.txt, described above, and two directories (docu and data). Two documentation files may exist for each station: station documentation (named docu/#STATION ID#.txt) and authority documentation (named docu/#STATION ID#_auth.txt). A data file for each entry in the filelist.txt will exist and be named data/#STATION ID#.rlrdata or data/#STATION ID#.metdata, as appropriate. Matlab code for reading in all of the data from one of these set is distributed with the zip file. Each set of PSMSL data files contain within them documentation warning flags which are used to point to either a suspect station record, a suspect station-year, or a suspect station-month. These flags are set on the basis of either the data checks described below or after further detailed scientific analysis. For monthly and annual data files, these flags will take the form of octal number, similar to UNIX permissions. The 'flag for attention' described above is "001". In addition, we have introduced the flag "010", which indicates that the value was a MTL value (perhaps corrected to MSL in the RLR file) in a MSL record. Both flags being true would be indicated by "011". Please see the above notes.php page for more information. PSMSL DATA CHECKS The following describes the checks made by the PSMSL on data received from national authorities. Details can also be found in PSMSL publications (e.g. Woodworth, Spencer and Alcock, Int. Hyd. Rev., 67(1), 131-146, 1990). In general, in years past the PSMSL has not received copies of original tide gauge hourly height measurements or continuous charts but has accepted monthly and annual mean sea level (MSL) values from national authorities on the understanding that these quantities have been computed accurately. Inevitably, this has always not been the case. The PSMSL has devised a range of tests on the supplied MSL information which guards against gross errors in the dataset such as transcription errors or large unrecorded datum changes. Some of these tests are 'common sense', others are 'statistical': (1) A check is made that the average of the quoted monthly mean values is consistent with the quoted annual mean. (2) 'Common sense' consistency checks are made on the data including checking that the datum information is consistent with previous knowledge. This includes reference to back correspondence and simple time series plotting. (3) A search for outliers is made on data for each calendar month of the year separately, and for the annual means, for all possible time-spans containing at least 20 years of data. A linear fit is made to the time series and any individual monthly mean value more than 4.5 standard deviations from the fitted line is flagged. (4) A search is made for incorrect datum information by performing a set of linear regressions of RLR annual mean values against the supplied datum correction factors ('RLR factors' in PSMSL terminology) in all possible 20-year time-spans. A correctly adjusted RLR time series should be uncorrelated with the RLR factors. (5) A search for jumps in the RLR time series is made for each calendar month of the year separately and for the annual means. The difference between a mean value and the corresponding value for the next year of data is histogrammed and any outlier more than 4.5 standard deviations from the mean-difference is flagged. (6) A test is made for 'upside down' data. In several countries the main research interest is the study of vertical land movements, rather than sea level changes, with the result that MSL data are often quoted as the distance below a benchmark height rather than above it. The most sensitive test to guard against such an error is an inspection of the seasonal cycle which, for 'upside down' data, would appear opposite to that observed in neighbouring records and opposite to oceanographic and meteorological expectations. (7) A set of 'buddy checking' is made in which the RLR data from one station is subtracted from that of a neighbouring station (or 'buddy') which is less than 400km away. Over this short distance most of the MSL variability due to oceanographic and meteorological forcings in the two records should be similar and will cancel out giving a difference time series primarily composed of relative vertical land movements, instrumental and datum errors and any small spatially varying ocean and weather influences. The previous tests are then applied to the difference time series and any discrepancies are flagged. These tests have been applied to the entire dataset and inconsistencies have been referred back to the national authorities, although the reasons for some apparent oddities are no doubt lost in history. RECOMMENDATIONS TO AUTHORITIES CONTRIBUTING DATA TO THE PSMSL a) General The Permanent Service appreciates the contributions from all organisations supplying mean sea level data and does not seek to impose unnecessary conditions upon contributors. Nevertheless a minimum of quality control must be exercised if the data bank is to be an authoritative reference. To this end the PSMSL requests the following information together with each set of monthly and annual mean sea level values supplied: i) the units used (metres, rarely feet), ii) a statement of the datum to which the values refer, iii) a statement of the measured depth of that datum below the primary tide gauge bench mark (TGBM), iv) an indication of incomplete or deduced data (see paragraph b), v) the number of observations per day used to calculate the monthly means, vi) any information of changes in datums, bench marks or relevant procedures since the previous batch of data, vii) any information on the availability of more frequent readings (e.g. hourly heights). Although data will be accepted in any format, mean heights should preferably be in the metric system to the nearest millimetre, and the datum to which the means refer should preferably be the tide gauge zero. Data will be gratefully received in any form (e.g. as paper tabulations in letters or via email). b) Treatment of incomplete records One of the most important things for users of the mean sea level data bank to know is the accuracy of the published figures. Details of the treatment of gaps in the tidal record are of particular interest. Therefore, the PSMSL makes the following recommendations: i) small gaps in observed tidal records should be interpolated, if possible before computing monthly and annual means, ii) the interpolation should be performed at an early stage in the processing. One principle to adopt is that of a comparison with the complete records from a nearby station. However we would stress that predicted values are not suitable for interpolation because of meteorological effects, iii) in cases where interpolation is impossible the monthly mean should be compiled from the incomplete data. Where more than 15 days are missing from a month a mean value should not be computed, iv) when sending mean values to the PSMSL, authorities are requested to indicate if interpolation has been effected or the exact number of missing days of data. These details should be sent as suffixes after each monthly mean and shown in brackets: e.g. 2487(9) would mean 9 daily mean values were missing and not interpolated when computing the mean of 2487mm; 913(XX) would mean missing data were interpolated to provide the average of 913mm, v) if there are 11 or 12 monthly mean values available then an annual mean should be calculated. If the annual mean is computed by averaging the monthly means, the monthly means must first be weighted. The weight for each month should be the number of days for which readings were available. c) Computation of monthly and annual mean values The attention of data contributors is drawn to three IOC publications entitled 'Manual on Sea Level Measurement and Interpretation' which can be downloaded from http://www.psmsl.org/train_and_info/training/manuals/ The third one is especially worth considering with regard to monthly mean computation. The PSMSL will be pleased to assist with advice on methods of data processing and the determination of mean values. d) Preservation of original data Contributors are urged to preserve the original sea level data in permanent form. The information contained in such basic time series is of great value in many scientific studies, is irreplaceable, and should not be lost to posterity. Where original data are available in computer compatible form, the PSMSL would be grateful to receive copies (and in the case of GLOSS stations, it is a requirement that any such data are provided to GLOSS Archiving Centres, of which the PSMSL is one, see below). Such global databanking of higher frequency information will evolve considerably over the next few years as a result of GLOSS and related activities. LINKAGE TO THE GLOSS PROGRAMME The Global Sea Level Observing System (GLOSS) is a programme coordinated by the Joint Technical Commission for Oceanography and Marine Meteorology (JCOMM) of the Intergovernmental Oceanographic Commission (IOC) and the World Meteorological Organisation (WMO). It aims for the establishment of global and regional sea level networks for oceanographic, climate change and coastal research purposes. The main component of GLOSS is the 'Global Core Network' (GCN) for long term climate change and oceanographic sea level monitoring. For full details, see the page http://www.gloss-sealevel.org/ The development of GLOSS can be traced through the various Implementation Plans for the programme and through related documents (IOC, 1990; Woodworth and Player, 2003). The most recent Implementation Plan (IOC, 2012) can be downloaded from the above page. The PSMSL took a major lead in the planning of GLOSS which in the long term will result in a significant improvement in the quantity and quality of data delivered to the PSMSL. Further information on the development of GLOSS from a PSMSL perspective can be found via the above web pages and Woodworth (1998), Woodworth and Player (2003), and Holgate et al. (2013). The PSMSL has played a particularly important part in GLOSS's provision of training courses and training materials with courses held at the PSMSL, and with PSMSL-related scientists having taken part in courses overseas. Recent years have seen major efforts to collect higher frequency (typically hourly) sea level data as well as MSL information. In collaboration with the British Oceanographic Data Centre and the University of Hawaii Sea Level Center, the PSMSL also functions as a 'GLOSS Archiving Centre' for higher frequency sea level data. See the above-mentioned GLOSS web page and the page http://www.psmsl.org/data/hf/ for further details. References ---------- FAGS, 1989. Federation of Astronomical and Geophysical Data Analysis Services (FAGS). Chronique de l'UGGI, No.194, 1-67. Holgate, S.J., Matthews, A., Woodworth, P.L., Rickards, L.J., Tamisiea, M.E., Bradshaw, E., Foden, P.R., Gordon, K.M., Jevrejeva, S., and Pugh, J., 2013. New Data Systems and Products at the Permanent Service for Mean Sea Level. Journal of Coastal Research: Volume 29, Issue 3: pp. 493 – 504. doi:10.2112/JCOASTRES-D-12-00175.1. IOC, 1990. Global Sea Level Observing System (GLOSS)implementation plan. Intergovernmental Oceanographic Commission, Technical Series, No.35, 90pp. IOC, 1992. Joint IAPSO-IOC workshop on sea level measurements and quality control. Intergovernmental Oceanographic Commission, Workshop Report, No.81, 167pp. IOC, 2012. Global Sea Level Observing System (GLOSS) Implementation Plan – 2012. UNESCO/IOC, 41pp. Spencer, N.E. and Woodworth, P.L. 1993. Data holdings of the Permanent Service for Mean Sea Level (November 1993). Bidston, Birkenhead: Permanent Service for Mean Sea Level. 81pp. Woodworth, P.L., Spencer, N.E. and Alcock, G.A. 1990. On the availability of European mean sea level data. International Hydrographic Review, 67(1), 131-146. Woodworth, P.L. (ed.) 1998. Global Sea Level Observing System (GLOSS) Implementation Plan 1997. Intergovernmental Oceanographic Commission, Technical Series, No. 50, 91pp. & Annexes. Woodworth, P.L. and Player, R. 2003. The Permanent Service for Mean Sea Level: an update to the 21st century. Journal of Coastal Research, 19, 287-295.