Assessing water observation network settings by hydro-geological sub-sampling of a large data set for Sweden

Developing cost-effective methods for hydrological observations is an identified research objective of the WMO Hydrological Research Strategy 2022-2030. Network settings of hydrological observational networks are central for data collection and monitoring efforts to fulfill observation needs. In this work, we ask the question: Where and how densely (location and time) measurement stations need to be placed to gain sound scientific insights into hydro-meteorological conditions of a region?  

We address this question through information theory concepts and calculate entropy, joint entropy, and mutual information for an existing large dataset of hydro-meteorological parameters. The dataset spans 36 years (1981-2017) of daily data for Sweden based on the national S-HYPE hydrological model. Hydrological data include runoff, inflow, and streamflow as computed values and meteorological data encompass temperature and precipitation as measured and corrected data. We chose Sweden as a study domain to look at a Nordic region with a large number of water basins and an overall well-sampled region allowing to assess interesting network settings through sub-sampling.  

Sub-sampling and analysis for potential network settings is done for the seven hydrological clusters across Sweden as they are defined in Girons Lopez et al. (2021). Random sub-sampling (by 10%, 25%, 50%, 75%, and 90%) of each of the seven clusters shows a narrow range of (Shannon) entropy indicating excellent assignment of catchments to the seven clusters.  

Focusing on three clusters, which span Sweden’s North-South extend and mainly feature forested areas and a cluster covering mostly coastal lakes, we assess how much information remains in the data set if sub-sampled by hydro-geological parameters, such as baseflow and flashiness. Such tests allow us to determine ideal and minimum network settings with respect to observational and computational efforts based on different criteria relevant to scientific investigations and decision-making needs. 

 

Girons Lopez, M., Crochemore, L., and Pechlivanidis, I. G.: Benchmarking an operational hydrological model for providing seasonal forecasts in Sweden, Hydrol. Earth Syst. Sci., 25, 1189–1209, https://doi.org/10.5194/hess-25-1189-2021, 2021