Getting your money's worth: Testing the value of data for hydrological model calibration

Abstract

Despite the big data era, observational data continue to be a limiting factor in the environmental sciences. To collect the most informative field data, studies on the value of data are essential. This article describes a model‐based approach to assess the value of data. While we discuss the approach for hydrological model calibration, the approach is applicable across the environmental sciences. The overall goal is to provide guidance on optimal data collection strategies, that is, what to measure, where, and when.

A recipe for why and how to set up and sustain an experimental catchment

“…As computer power is becoming less expensive, field work more risky, and field data collection increasingly limited by financial and logistical constraints (Burt & McDonnell, 2015;Seibert et al, 2024), does it make sense to keep collecting experimental data, in some cases at high spatial and temporal resolution, that we can hardly analyse, at least during a typical PhD or postdoc time-span? Are these new big data and tools, combined with measurements from existing experimental catchments set in the past and often publicly available through data repositories, not enough to advance our understanding of how catchments function?…”

“…Here I highlight some reasons for which the answer is positive. i. Hydrology is data-hungry and still limited in ground data (Seibert et al, 2024). Experimental catchments are open-air laboratories of paramount significance for collecting new data, using them to generate hypotheses and rigorously testing them, and validating or falsifying hydrological models or model components (Beven, 2018;Pfister & Kirchner, 2017).…”

“…Then, according to the available funds and research goals, other sensors can be installed, possibly also extending the size of the monitored area (step iv). Decisions on whether to keep monitoring some variables or not might be based on the approach suggested bySeibert et al (2024), which includes calibrating a hydrological model on rainfall-runoff and other ancillary data collected at different times and catchment locations, and comparing the simulated and observed streamflow to determine the model performance. Although models could be differently sensitive to the incorporation of different data, this approach potentially allows assessing what, when, and where to measure in the catchment to obtain the most informative data for model calibration(Seibert et al, 2024).…”

Can eXplainable AI Offer a New Perspective for Groundwater Recharge Estimation?—Global‐Scale Modeling Using Neural Network

“…However, apart from a few regions with accessible and consistent national survey data, there has been the chronic deficiency of groundwater recharge data. Not only data-driven models but also existing process-based models at the regional and global scale usually have been limited by a relatively small number of field measurements in the subsurface (Seibert et al, 2024) which limits the verifiability and thus constrains the model's credibility. However, recently, comprehensive data sets of ground-based groundwater recharge estimates (i.e., estimates based on field measurements, excluding satellite-based measurements) have been presented by Mohan et al (2018), Moeck et al (2020) andMacDonald et al (2021), which, if used collectively, could address the problem of insufficient data.…”