Long-term timeseries in the GEMStat Water Quality Database - temporal trends and potential for model forcing with a focus on the UN SDG 6.3.2 indicator.

The GEMStat database, under the auspices of the UN Environment Programme and hosted by the GEMS/Water Data Centre at the International Centre for Water Resources and Global Change, currently holds over 29 million measurements for over 600 water quality parameters, from more than 18,000 stations in 91 countries and covering the time period from 1906 to 2023. The data is available via an online platform. Furthermore, the measurements under the GEMStat „open“ data policy are currently being compiled to be made available in an online repository.

The corresponding publication focusses on long-term timeseries, with continuous data records of at least 10 years, for the five parameter groups that can be included in submissions for the UN SDG 6.3.2 Level 1 indicator (acidification, nitrogen, oxygen, phosphorus, and salinity). The incentive behind this course of action is to support the use of modelling approaches for future data submission on the SDG 6.3.2 indicator.

Overall, 1,082 timeseries stations were identified providing data for at least one parameter group, with 2,904,185 data points in total. At most stations (405), data for four of the selected parameter groups are available, followed by stations where data on five parameter groups are available (285 stations). River stations make up the largest part of the stations (978 or 90%). In addition, lake stations (84 or 7.8%), groundwater stations (17 or 1.6%) and reservoir stations (3 or 0.3%) contribute to the total number of stations.

Timeseries length ranges from 10 to 114 years with a mean duration between 23.7 (Oxygen) and 30.5 years (Salinity).

The number of measurements per year ranges from 3 to 233, with a mean frequency between 9.4 (oxygen) and 14.1 (pH), indicating an overall tendency of monthly measurements for many timeseries.

Timeseries were also analyzed for significant trends in the data using prewhitened nonlinear trend analysis.

In total, 4,050 of 7,019 timeseries (57.70%) showed significant trends (p<0.05). The fraction of significant timeseries stations within a parameter group was highest for nitrogen (1,299 of 2,094 stations or 62.03%), followed by phosphorus (587 of 978 stations or 60.02%), acidification (546 of 914 stations or 59.74%), salinity (1,158 of 1,995 stations or 58.05%), and oxygen (460 of 1,038 stations or 44.32%).

The length of these timeseries together with the high sampling frequency predestines their application for model forcing in support of the SDG 6.3.2 indicator. Furthermore, this study looked at one fairly specific application of GEMStat timeseries data and more timeseries are available for other water quality parameters and could advance model development with a range of other foci. The additional trend analysis indicated potential effects of global change in more than 50% of the timeseries, highlighting regional hotspots for further in-depth analysis.