Application of generic data assimilation tools (DATools) for flood forecasting purposes
Introduction
The importance of an accurate and early warning in the case of river flooding cannot be overstated. Within several European funded projects like IRMA-Sponge (http://www.irma-sponge.org/) and EFFS (http://effs.wldelft.nl/), this wide felt need resulted in the development of a prototype open architecture flood forecasting system. In 2003, the UK Environment Agency (EA) commissioned an assignment to Delft Hydraulics for the development of redesigned version of this prototype now known as Delft-FEWS. Delft-FEWS is an ETL (Exchange, Transform, and Load) system. With this system many types of hydrological and meteorological data can be imported into a database. Subsequently, this data can be automatically validated and transformed if necessary. Through a published interface (PI) every participating organization can couple its models to the central database. Once coupled, many types of tasks can be performed automatically such as running forecasts, importing data, updating the web or intranet sites, etc. More information about this system can be found in Werner et al. (2004), Werner and Heynert (2006) or is available at http://www.wldelft.nl/soft/fews/int/index.html. Currently, the EA is using Delft-FEWS for operational forecasting of water levels and discharges in England and Wales in the national flow forecasting system (NFFS). This year, Delft-FEWS will be used for semi-operational forecasting in the Netherlands for prediction of water levels and discharges for the Rhine and Meuse. Other users of Delft-FEWS are located in Germany (BfG), Scotland (SEPA), Switzerland (FOWG), Italy (ARPA), Singapore, Taiwan, and Austria.
Data assimilation is a key element of real time flood forecasting (Madsen et al., 2000), and most forecasting systems apply some form of data assimilation. Delft-FEWS includes an error correction module (Broersen and Weerts, 2005) for output correction. A more sophisticated form of data assimilation is sequential data assimilation. Sequential data assimilation or state updating is a feedback system where the process models are conditioned using the information on the current state of the modelled system. These process models can be considered as a set of equations containing parameters and state variables (Refsgaard, 1997), where state variables are transient in time, and the parameters are generally held constant at some value determined in the calibration of the model prior to application in the real time environment. The primary goal of data assimilation is to guarantee an up to date representation of the state variables in model terms, making use of most recent available measurement information. This state is then used as an initial state for subsequent forecasts.
Data assimilation techniques are widely used in areas like meteorology, oceanography, and hydrology. Despite the fact that current Monte Carlo type filters are model independent, most implementations of these sequential data assimilation methods however are custom implementations specially designed for, and integrated with the code of a particular model. This is probably a consequence of the lack of generic data assimilation software packages and tools. The use of custom implementations has a number of disadvantages (COSTA, information on COSTA available at http://www.costapse.org):
- •
cost: the development and implementation of these methods is very time consuming and therefore expensive;
- •
incompatible: it is hard to reuse these data assimilation methods and tools for other models than for which they have originally been developed for.
Therefore, the objective of this study was the development of a generic sequential data assimilation module (DATools) for use within Delft-FEWS, and which can also be used standalone. The developed DATools software will be tested using a similar case study as used in Weerts and El Serafy (2006), which was performed in the recent past using a custom implementation instead of DATools. Secondly, a test inside the operational system FewsNL of the Rhine is carried out to show the ability of DATools to work within Delft-FEWS. During the course of this study, a lot of information has been exchanged with the COSTA project leading to the specification of identical interfaces for building blocks of both COSTA and DATools. The focus of the COSTA project is the development of a programming environment and is therefore more directed towards the programmers instead of the operational forecasters, which is more the focus of the DATools project.
Section snippets
General
The basic idea of data assimilation is to adjust the prior estimate of the model states given one or multiple measurement(s). To obtain the prior estimates of the model states one needs to define a stochastic model which requires specification of input and/or model uncertainty. On the other hand to adjust the prior estimate of the model states given the measurements one needs to know the measurement uncertainty. Therefore, the data assimilation process is split up into three distinct parts: the
Description of the case studies
In the following two paragraphs the description of the two case studies is given. The first test case (Ober Sieg HBV-96 model – twin experiment) is a similar case study as used in Weerts and El Serafy (2006), which was performed in the recent past using a custom implementation instead of DATools. This test is to show that the implementation of the available filters within DATools are working properly. The second test case (Meuse basin HBV-96 model – DATools within Delft-FEWS) is to show that
Results and discussion
In the following two paragraphs the results of the two case studies are shown and discussed.
Conclusions
DATools is a generic software package for data assimilation. Using DATools, it is possible to apply data assimilation methods for existing and new models. The focus of DATools lies in enabling data assimilation methods for operational forecasters using Delft-FEWS, but can also be used for academic studies (standalone version). This means that the PI interface of Delft-FEWS is supported, although some additional requirements with respect to state exchange have been added to the PI interface.
Acknowledgements
The hydrological model of the German part of the Rhine basin and the Ober Sieg River used in this paper has been developed by the Federal Institute of Hydrology in Germany (Federal Institute of Hydrology (Bfg), 2001) for the Institute for Inland Water Management and Waste Water Treatment in the Netherlands (RIZA). The hydrological model of the Meuse used in this paper has been developed for the Institute for Inland Water Management and Waste Water Treatment in the Netherlands (RIZA). Discharge
References (10)
- et al.
The data uncertainty engine (DUE): a software tool for assessing and simulating uncertain environmental variables
Computers and Geoscience
(2007) - et al.
Development and test of the distributed HBV-96 hydrological model
Journal of Hydrology
(1997) - Broersen, P.M.T., Weerts, A.H., 2005. Automatic error correction of rainfall-runoff models in flood forecasting...
- El Serafy, G.H., Gerritsen, H., Hummel, S., Weerts, A.H., Mynett, A.E., Tanaka, M., 2007. Application of data...
- Federal Institute of Hydrology (BfG), 2001. Hydrological modeling in the Rhine basin, Part II – Report on hourly...
Cited by (46)
Large scale hydrologic and tracer aided modelling: A review
2023, Journal of HydrologyPipedream: An interactive digital twin model for natural and urban drainage systems
2021, Environmental Modelling and SoftwareCitation Excerpt :These studies seek to improve streamflow forecasts by using data assimilation techniques to correct initial soil moisture conditions (Seo et al., 2003; Thirel et al., 2010; Weerts et al., 2010), snow water storage (Dechant and Moradkhani, 2011), and rainfall forcings (Seo et al., 2003). A variety of data assimilation approaches have been investigated, including variational data assimilation (Seo et al., 2003), best linear unbiased estimation (Thirel et al., 2010), particle filtering (Dechant and Moradkhani, 2011; van Leeuwen, 2009), and ensemble Kalman filtering (Weerts et al., 2010; Dechant and Moradkhani, 2011; Pauwels and Lannoy, 2009). While most of the literature focuses on offline data assimilation, a few studies have examined data assimilation in a real-time context.
An automated calibration framework and open source tools for 3D lake hydrodynamic models
2020, Environmental Modelling and Software