Refine
Language
- English (26)
Is part of the Bibliography
- yes (26)
Keywords
- dynamics (4)
- catchment (3)
- floods (3)
- hydrological models (3)
- inference (3)
- machine learning (3)
- model (3)
- Bayesianism (2)
- aquifer (2)
- climate change impacts (2)
Machine learning (ML) algorithms are being increasingly used in Earth and Environmental modeling studies owing to the ever-increasing availability of diverse data sets and computational resources as well as advancement in ML algorithms. Despite advances in their predictive accuracy, the usefulness of ML algorithms for inference remains elusive. In this study, we employ two popular ML algorithms, artificial neural networks and random forest, to analyze a large data set of flood events across Germany with the goals to analyze their predictive accuracy and their usability to provide insights to hydrologic system functioning. The results of the ML algorithms are contrasted against a parametric approach based on multiple linear regression. For analysis, we employ a model-agnostic framework named Permuted Feature Importance to derive the influence of models' predictors. This allows us to compare the results of different algorithms for the first time in the context of hydrology. Our main findings are that (1) the ML models achieve higher prediction accuracy than linear regression, (2) the results reflect basic hydrological principles, but (3) further inference is hindered by the heterogeneity of results across algorithms. Thus, we conclude that the problem of equifinality as known from classical hydrological modeling also exists for ML and severely hampers its potential for inference. To account for the observed problems, we propose that when employing ML for inference, this should be made by using multiple algorithms and multiple methods, of which the latter should be embedded in a cross-validation routine.
Neutrons on rails
(2021)
Large-scale measurements of the spatial distribution of water content in soils and snow are challenging for state-of-the-art hydrogeophysical methods. Cosmic-ray neutron sensing (CRNS) is a noninvasive technology that has the potential to bridge the scale gap between conventional in situ sensors and remote sensing products in both, horizontal and vertical domains. In this study, we explore the feasibility and potential of estimating water content in soils and snow with neutron detectors in moving trains. Theoretical considerations quantify the stochastic measurement uncertainty as a function of water content, altitude, resolution, and detector efficiency. Numerical experiments demonstrate that the sensitivity of measured water content is almost unperturbed by train materials. Finally, three distinct real-world experiments provide a proof of concept on short and long-range tracks. With our results a transregional observational soil moisture product becomes a realistic vision within the next years.
[1] Spatial patterns of land surface and subsurface characteristics often exert significant control over hydrological processes at many scales. Recognition of the dominant controls at the watershed scale, which is a prerequisite to successful prediction of system responses, will require significant progress in many different research areas. The development and improvement of techniques for mapping structures and spatiotemporal patterns using geophysical and remote sensing techniques would greatly benefit watershed science but still requires a significant synthesis effort. Effective descriptions of hydrological systems will also significantly benefit from new scaling and averaging techniques, from new mathematical description for spatial pattern/structures and their dynamics, and also from an understanding and quantification of structure and pattern-building processes in different compartments ( soils, rocks, and land surface) and at different scales. The advances that are needed to tackle these complex challenges could be greatly facilitated through the development of an interdisciplinary research framework that explores instrumentation, theory, and simulation components and that is implemented in a coordinated manner
Distributed environmental models such as land surface models (LSMs) require model parameters in each spatial modeling unit (e.g., grid cell), thereby leading to a high-dimensional parameter space. One approach to decrease the dimensionality of the parameter space in these models is to use regularization techniques. One such highly efficient technique is the multiscale parameter regionalization (MPR) framework that translates high-resolution predictor variables (e.g., soil textural properties) into model parameters (e.g., porosity) via transfer functions (TFs) and upscaling operators that are suitable for every modeled process. This framework yields seamless model parameters at multiple scales and locations in an effective manner. However, integration of MPR into existing modeling workflows has been hindered thus far by hard-coded configurations and non-modular software designs. For these reasons, we redesigned MPR as a model-agnostic, stand-alone tool. It is a useful software for creating graphs of NetCDF variables, wherein each node is a variable and the links consist of TFs and/or upscaling operators. In this study, we present and verify our tool against a previous version, which was implemented in the mesoscale hydrologic model (mHM; https://www.ufz.de/mhm, last access: 16 January 2022). By using this tool for the generation of continental-scale soil hydraulic parameters applicable to different models (Noah-MP and HTESSEL), we showcase its general functionality and flexibility. Further, using model parameters estimated by the MPR tool leads to significant changes in long-term estimates of evapotranspiration, as compared to their default parameterizations. For example, a change of up to 25 % in long-term evapotranspiration flux is observed in Noah-MP and HTESSEL in the Mississippi River basin. We postulate that use of the stand-alone MPR tool will considerably increase the transparency and reproducibility of the parameter estimation process in distributed (environmental) models. It will also allow a rigorous uncertainty estimation related to the errors of the predictors (e.g., soil texture fields), transfer function and its parameters, and remapping (or upscaling) algorithms.
Data driven high resolution modeling and spatial analyses of the COVID-19 pandemic in Germany
(2021)
The SARS-CoV-2 virus has spread around the world with over 100 million infections to date, and currently many countries are fighting the second wave of infections. With neither sufficient vaccination capacity nor effective medication, non-pharmaceutical interventions (NPIs) remain the measure of choice.
However, NPIs place a great burden on society, the mental health of individuals, and economics. Therefore the cost/benefit ratio must be carefully balanced and a target-oriented small-scale implementation of these NPIs could help achieve this balance.
To this end, we introduce a modified SEIRD-class compartment model and parametrize it locally for all 412 districts of Germany. The NPIs are modeled at district level by time varying contact rates. This high spatial resolution makes it possible to apply geostatistical methods to analyse the spatial patterns of the pandemic in Germany and to compare the results of different spatial resolutions.
We find that the modified SEIRD model can successfully be fitted to the COVID-19 cases in German districts, states, and also nationwide. We propose the correlation length as a further measure, besides the weekly incidence rates, to describe the current situation of the epidemic.
Transverse dispersion, or tracer spreading orthogonal to the mean flow direction, which is relevant e.g, for quantifying bio-degradation of contaminant plumes or mixing of reactive solutes, has been studied in the literature less than the longitudinal one. Inferring transverse dispersion coefficients from field experiments is a difficult and error-prone task, requiring a spatial resolution of solute plumes which is not easily achievable in applications. In absence of field data, it is a questionable common practice to set transverse dispersivities as a fraction of the longitudinal one, with the ratio 1/10 being the most prevalent. We collected estimates of field-scale transverse dispersivities from existing publications and explored possible scale relationships as guidance criteria for applications. Our investigation showed that a large number of estimates available in the literature are of low reliability and should be discarded from further analysis. The remaining reliable estimates are formation-specific, span three orders of magnitude and do not show any clear scale-dependence on the plume traveled distance. The ratios with the longitudinal dispersivity are also site specific and vary widely. The reliability of transverse dispersivities depends significantly on the type of field experiment and method of data analysis. In applications where transverse dispersion plays a significant role, inference of transverse dispersivities should be part of site characterization with the transverse dispersivity estimated as an independent parameter rather than related heuristically to longitudinal dispersivity.