A global dataset of surface water and groundwater salinity measurements from 1980-2019

Abstract

Salinization of freshwater resources is a growing water quality challenge, which may negatively impact both sectoral water-use and food security, as well as biodiversity and ecosystem services. Although monitoring of salinity is relatively common compared to many other water quality parameters, no compilation and harmonisation of available datasets for both surface and groundwater components have been made yet at the global scale. Here, we present a new global salinity database, compiled from electrical conductivity (EC) monitoring data of both surface water (rivers, lakes/reservoirs) and groundwater locations over the period 1980-2019. The data were assembled from a range of sources, including local to global salinity databases, governmental organizations, river basin management commissions and water development boards. Our resulting database comprises more than 16.3 million measurements from 45,103 surface water locations and 208,550 groundwater locations around the world. This database could provide new opportunities for meta-analyses of salinity levels of water resources, as well as for addressing data and model-driven questions related to historic and future salinization patterns and impacts.

Fig. 1

Global overview of station density and measurement distributions. The global map of panel (a) shows the total number of stations per country with electrical conductivity (EC) observations included in our database, over the full data period (1980–2019). The zoomed panels highlight high-density station regions of each continent, whereas the numbers given for each water type is the total number of stations for associated continent. Panel (b) shows number of stations per country for the different decades included in the database (1980–1989, 1990–1999, 2000–2019). Panel (c) shows the distribution of sampled water types (as percentages of total samples) over the three decades, per continent. No data is represented as striped columns. Panel (d) shows violin plots of the distribution of number of measurements, per water type, over the same time periods.

Fig. 2

Data selection and harmonisation flowchart. The figure illustrates the processing and harmonizing steps of each dataset (divided into surface and groundwater parts) after initial data collection.

Fig. 3

Validation of converted TDS to EC for groundwaters. Time-series plot and scatter correlations of measured vs. predicted electrical conductivity (EC), using regional conversion factors. Panel (a) shows an example time-series from the groundwater station with the highest number of measurements (estimated from the “max” function in R) in Australia (data source: Water connect, n = 538) and panel (b) shows its corresponding scatter correlation (R² = 0.99). Panel (c) shows the correlation between measured and converted EC for the full dataset of all groundwater stations from Water connect (n= 37,819, R² = 0.98). Panel (d) and (e) shows correlations between measured and predicted EC data, for groundwaters in Texas (data source: TWDB, n = 59,985, R² = 0.91) respectively California (data source: GAMA, n = 4,706, R² = 0.98). All scatterplots were done in R, using the “ggscatter” function from the ggpubr package and estimating correlation coefficients using the “pearson” function.