RELEASE NOTES FOR CRU TS v3.24.01: 25 January 2017 The 3.24.01 release of the CRU TS dataset covers the period 1901-2015, and corrects various errors in release 3.24. It supercedes this and all earlier releases. The reference to use is: Harris, I., Jones, P.D., Osborn, T.J. and Lister, D.H. (2014), Updated high-resolution grids of monthly climatic observations – the CRU TS3.10 Dataset. Int. J. Climatol., 34: 623–642. doi: 10.1002/joc.3711 Differences Between CRU TS v3.24 and previous releases, Known Issues, and the forthcoming 4.00 release. 1. Differences Between Releases 1.1 Between 3.23 and 3.24 In addition to updating the dataset with 2015 data, many new stations have been added, mainly for TMN and TMX but also - in Iceland - for TMP, PRE and CLD. The CHERRAPUNJI precipitation error (in 1980) has also been corrected: this was a late change which means that PRE and WET have a different run ID (1610031104) to the other variables (1609301803), though this will only be apparent in the NetCDF headers and the header lines of the CRU CY derived files. 1.2 Between 3.24 and 3.24.01 This corrected release addresses two sources of error: misassigned WWR observations, (global; TMP, PRE, TMN, TMX), and mis-scaled WWR observations (Myanmar, PRE). The misassigned WWR observations affected various regions, in particular (for TMP) Northwestern France, Eastern Siberia, Reunion Island, and Sudan. The time period was approximately 1990 to 2013. Additionally, the CRU half-degree wind climatology (used in calculation of PET) was reproduced from the higher-resolution 10' climatology in order to correct coverage and range issues. The has resulted in PET for some regions being scaled by around 8-12%, though variability is unchanged. Similarly, WET has been improved through the use of better 2.5-degree gridded normals for synthetic production. Again, changes are principally issues of scale, not signal. 2. Known Issue: Interpolation near the International Date Line Affects: Eastern Siberia, all variables. In April 2014, a user notified BADC that they had discovered a discontinuity in data fields near to the International Date Line (ie, Eastern Siberia). This manifests as a vertical (ie, along a line of longitude) discontinuity approximately 2.5° (five gridcells) to the West of the Date Line (so, at about 177.5°E). Investigation of this phenomenon concluded that it was due to the gridding approach used: this is explained in the documentation, and consists of triangulation, followed by interpolation to the chosen grid points. The triangulation is not 'spherical', and so triangles do not form across the Date Line, this is causing the anomalous data. Despite exhaustive testing of available routines within the package that is being used, none have been found to deliver fields that are free from unwanted artifacts. This has been a significant driver in the forthcoming move to ADW gridding. 3. Known Issue: Anomalous Canadian data. Affects: Canada (esp. Eastern Canada), temperature variables. In May 2014, a user contacted CRU with statistical work showing a cold bias in CRU TS mean temperature data (TMP) over Eastern Canada, when compared to other datasets and models. This bias apears to begin in 2005 and runs until the present. Resources have not been available to investigate thoroughly. Affects: Canada, wet days variable. The country-level spatially-averaged wet day series shows unlikely behaviour, with the day count steadily rising until 1990, at which point there is a precipitous fall (of approximately two days) to levels previously seen in the 1930s. 4. Known Issues: General differences from previous release. Affects: All variables, all releases At first glance, an update process should not change any existing published data unless it includes retrospective data for the previous time period, or introduces new stations. However, these are several ways in which new updates to existing stations may affect published data from any time: a) New data for a station will change the standard deviation of that station. This is used to assess each observation of that station and to eliminate excessive values. So in 'borderline' cases, new data may block a previously-allowed datum, or alternatively may allow through a previously-blocked value. b) New metadata for any station can result in a relocation. This may simply be a correction, or improved accuracy: but any move will alter the triangulation process, and this can propogate over a wide area as the triangles are not constructed independently of each other. So a station move of perhaps 0.05 degrees could change an interpolated value many degrees away (and all of the others in between). This change will also apply to the whole station series, changing all triangles at every time step. This has been another significant driver in the imminent move to ADW gridding. 5. The Forthcoming 4.00 Release The 4.00 release has been further delayed. However, this has allowed investigation of methods for error estimation. As a first stage, accurate station counts (the number of observations contributing to each datum) will be published alongside (inside, for NetCDF versions) the data files. The 4.00 release will utilise Angular-Distance Weighting (ADW) gridding, promising more accurate results with far greater adjustability and logging. It will cover the same spatial, temporal and variate spaces as version 3.23 (land areas excluding Antarctica at 0.5°x0.5°, monthly from 1901 to 2014 with no missing values, 10 variables). 3.xx and 4.xx versions will run concurrently for a period, after which the new (ADW) approach will be used. This is to allow comparisons between the methods and results to be made by users of the dataset. As always, please contact BADC in the first instance if you have any questions, observations or suggestions. Ian Harris NCAS-Climate Climatic Research Unit School of Environmental Sciences University of East Anglia Norwich NR4 7TJ