- Email: [email protected]
Weather radar and hydrology
Advances in Water Resources 32 (2009) 969–974
Contents lists available at ScienceDirect
Advances in Water Resources journal homepage: www.elsevier.com/locate/advwatres
Preface
Weather radar and hydrology 1. Introduction The growing concerns about environmental and climate change issues, as well as the emergence of the concept of sustainable development enshrined in the European Water Framework Directive (WFD) legislation, has modified the requirements for future hydrological observation and modelling initiatives and techniques. Originally, the focus at larger catchment scales was on the water stream flow at a limited number of locations, essentially for operational purposes such as flood prediction, dam regulation and hydropower production as well as hydrological process understanding. Water is now considered under its multiple functions such as water supply, hydrological risk, hydropower, irrigation, tourism, ecological value, to name a few. The demand for better hydrological understanding and techniques has thus moved to a more integrated prediction of the water balance components (rainfall, runoff, water storage, transpiration, evaporation, groundwater levels, etc.) at every point within a region [21]. In addition, consideration of land-use and human-induced landscape modifications is now a major concern for water management problems (i.e. for flood forecasting) and impact studies of land use change on stream flow, pollutants and/or sediments transport. For most of these questions, knowledge of the general characteristics of water balance components is not sufficient, more detailed process understanding (i.e. storm responses) throughout the landscape are required. The development of distributed and/or semi-distributed hydrological models, where there is a greater representation of the land-surface heterogeneities and the characterization of distributed inputs, are thus increasingly necessary. The usefulness of distributed hydrological models has been questioned, mainly due to overparameterization problems, parameter estimation and validation limitations [9,43,8]. Such difficulties cannot be ignored but the continuous development of hydrological observation and landscape characterization techniques are likely to offer renewed possibilities in this field. We must recognise there is an inevitable compromise between the complexity of the model proposed, and the spatial data that is available to drive the model (inputs) and robustly evaluate its predictions. We suggest weather radar technology offers a unique means for characterizing the rainfall variability over the range of scales and with the space-time resolutions required for a large variety of hydrological problems. This is especially true for risk management in urban and mountain watersheds characterized by fast dynamics and for which rainfall monitoring and nowcasting are critical [21,37]. For the largest scales, continental rainfields derived from radar networks may also provide valuable information in terms of water budget in complement with climatologic networks and satellite imagery. 0309-1708/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.advwatres.2009.03.006
This article is a preface to a special issue journal that has been initiated in parallel to the International Symposium Weather Radar and Hydrology, held in Grenoble and Autrans, France, in March 2008. This event continues a series of symposia originally called Hydrological Applications of Weather Radar that started in 1989 at the University of Salford, UK. This first symposium was followed by the Hannover (1992), Sao Paulo (1995), San Diego (1998), Kyoto (2001) and Melbourne (2004) meetings. The main objective of the 2008 symposium was to promote the use of quantitative precipitation estimates (QPE) derived from weather radar technology in hydrological sciences and their applications. The symposium was composed of two events: (i) A 3-day conference held on 10–12 March 2008 in Grenoble. The programme was composed of 5 keynote speeches, 52 oral and 80 poster presentations covering the multiple aspects of radar physics, radar QPE and distributed hydrological modelling. This scientific material (abstracts, extended abstracts, oral presentations) can still be accessed on the symposium web site at the following address: http:// www.wrah-2008.com. (ii) A 2-day workshop held on 13–14 March 2008 in Autrans (Vercors). The workshop was dedicated to thematic discussions aimed at summarizing the symposium results and investigating future research directions. In the following, a presentation of the most promising techniques for radar rainfall QPE is proposed in Section 2 prior to a discussion about some of the challenges of distributed hydrological modelling (Section 3). A set of recommendations is made in Section 4 to foster the use of weather radar in hydrologic sciences. Section 5 briefly introduces the articles of the special issue. 2. The most promising techniques for radar rainfall estimation Real-time rainfall detection and estimation as well as rainfall nowcasting have been the main motivations for the deployment of operational weather radar networks over the past six decades. Although the qualitative value of radar networks is unquestioned, the complex technology of weather radar systems combined with the variety of uncertainty sources and the strong variability of precipitation at all scales [30] has limited the quality of radar precipitation estimates [37] and therefore the use of radar QPE in operational hydrology. Measurement artefacts (e.g., clear air echoes, ground and sea clutter, anthropogenic targets such as wind mills, airplanes, etc.) are a first very important limitation for quantitative use. In spite of automated identification techniques now being routinely
970
Preface / Advances in Water Resources 32 (2009) 969–974
implemented for conventional radars (i.e. single-polarization noncoherent Doppler radar in the following), residual artefacts may significantly alter rain reflectivity values and the subsequent radar data processing steps. An agreement exists now about the main radar error sources related to both instrumental and sampling properties [75,45,37]: calibration, attenuation due to wet radome, ground clutter and beam blockages, attenuation due to gas and rain, vertical heterogeneity of the hydrometeors (referred to as the vertical profile of reflectivity –VPR- in the following), reflectivity-rain rate conversion (Z–R relationship), rainfall advection. Volume-scanning protocols and physically-based data processing are now generally implemented operationally to cope with such error sources. Although some noticeable progress has thus been obtained [39,68,69,26] some inherent limitations will continue to exist due for instance to the difficulty in performing an absolute calibration, to the sampling problems (radar resolution varies as a function of range, non-uniform beam filling, non-observation of the low-level altitudes above ground level beyond 60–80 km), to entangled errors (e.g., combination of calibration, attenuation and VPR errors at C-band) and to the high intra-field variability of the Z–R relationship [72]. A major advent in the recent years is the implementation of dual polarization and Doppler techniques in a pre-operational mode for several weather radar networks [76,62,42]. In addition to the reflectivity Z, the differential reflectivity ZDR, the cross correlation qhv and the differential phase shift /DP between the horizontal and vertical polarizations can be measured for each radar resolution volume. It is worth noting that the potential for dualpolarisation techniques to eliminate non-meteorological echoes [42] has actually been the main reason why several countries have decided to deploy a network of polarimetric radars. Classification algorithms are used to distinguish several classes of liquid and frozen hydrometeors [67] prior to the application of specific rainrate estimators using various combinations of the radar measurables [62]. Polarimetry also offers some interesting possibilities such as self-calibration of reflectivity, attenuation correction [70,41] and DSD retrieval [16,57]. Of lesser direct value for radar QPE, the Doppler capability allows for a better description of the rain system dynamics [15]. Refractivity measurements from the phase variation of ground targets [31] also offer potential for the characterization of low-level humidity fields. There is an agreement about the critical importance of calibration and maintenance operations for quantitative applications. The parameters of the radar systems need to be monitored continuously and alarms have to be triggered when predefined thresholds are exceeded. Data from the monitoring system are critical metadata and need to be archived along with the radar observations and products. Polarization means more measured variables and thus more information, but also increased technical complexity and workload in terms of system maintenance. Despite the promises of polarimetry, there is evidence that the density of existing operational radar networks is insufficient for a proper hydrological coverage at the national/continental scale if one considers the 100-km range as a maximum practical limit. Two complementary solutions are currently being investigated: The X-band frequency offers attractive design characteristics compared to the S-band and C-band frequencies (reduced transmitted power and antenna size; reduced sensitivity to ground clutter) that are counterbalanced by a dramatic sensitivity to rain attenuation. In the recent years, a breakthrough has been achieved in radar QPE at X-band through the use of polarimetry and range-profiling algorithms [70,2,40,51,53]. Several industrial (e.g. (http://www.casa.umass.edu/) and research projects have been initiated to further develop this concept which has certainly a great future for radar QPE in rugged topography and/or in urban areas.
Although much less established, microwave links rainfall measurement also offers interesting potential with the deployment of wireless communication networks all over the world [56,49]. A major advantage is related to the fact that such measurements are made close to the ground. First studies indicate that effects of rainfall variability are limited if the link length and frequency are chosen correctly [6,50]. For such settings, the main error sources are related to the wetting of antenna covers and to the resolution of power measurements.
3. Distributed hydrological modelling Recent and ongoing research initiatives such as DMIP (Distributed Model Intercomparison Project; [66]) and PUB (Prediction in ungauged basins; [34]) are focused on distributed hydrological modelling. For PUB the concept is to reduce the uncertainties in the modelling process by improving our understanding and representation in models. We briefly review hereafter some of the related challenges. One of the difficulties in hydrological modelling is that the spatial and temporal scales of hydrological processes span several orders of magnitude: from a few m2 and minutes for soil infiltration to thousands of km2 and several years for groundwater [13]. Furthermore, depending on the climate and the characteristics of the catchment (topography, land use, pedology, etc.), only one or several of these processes may be dominant. For instance, it is usually recognized that runoff (both surface and sub-surface runoff) is dominant in the case of flash-floods, but that evapotranspiration must be accounted for in long-term simulations. According to the scale of interest, the dominant processes might also change. For example while soil hydraulic characteristics are influential for runoff generation through infiltration excess and ponding at the plot scale, the presence of macropores and preferential flow might be the dominant factor when studying the runoff generated at the hillslope scale. Finally, there are several thresholds in hydrology (e.g., appearance of ponding, snow melting, saturated areas, flooding, etc.) which make the modelling more complex due to the associated high level of non-linearities in hydrological responses [65]. A second type of difficulty arises from the combined heterogeneities of the land surface and the sub-soil. The large spatial heterogeneity of the land surface is driven by topography, land use, geology and pedology and increasingly by human actions. To simplify the problem, hydrologists try to define so-called homogeneous zones, i.e. areas where the hydrological behaviour can be considered as homogeneous. This has led to the definition of Hydrological Representative Units (HRU) [33,10], Representative Elementary Areas (REA) [74], Representative Elementary Watersheds [61]. In terms of modelling, this range of concepts has led to various landscape discretizations. Models are mostly based on regular meshes that are usually derived from Digital Terrain Models (e.g. [1,11,60,18]. However, the continental surface heterogeneity is not well described by regular grids and other discretizations have been proposed such as triangular irregular networks [44], hillslopes [71] and hydrolandscapes [24]. For process representations, two approaches can be distinguished [64]. The first one is an upward approach where processes are modelled using equations established at very small scales (such as the Richards equation for water transfer within the saturated/ unsaturated zone, the St-Venant equation for water flow within a river channel, or the Boussinesq equation for water transfer within an aquifer) and extended to larger scales. Effective parameters, relevant to the scale of discretization must therefore be specified if not calibrated [13]. This approach is difficult to verify due to limitations with current hydrological observation techniques (e.g., for subsurface flow processes). Given the non-linearity in hydrological
Preface / Advances in Water Resources 32 (2009) 969–974
processes, it is also difficult to derive theoretical approaches describing how the model parameters might change across scales. The second approach is a downward approach where process conceptualization is derived from the available data, with a minimum of parameters, without representing what happens at smaller scales. The simplest approach to this is data based modelling techniques; other examples of differing complexity of representation can be found in [46,29]. But the establishment of such models requires that data are available at the modelling scale, which is not always the case, and the extrapolation to ungauged catchments and/or potentially more extreme observational events that are not yet characterised within the data is difficult. In terms of modelling practice, a change in modelling paradigm is certainly required as well as accounting for the uncertainties in model prediction. Modellers were trying to answer the following question: ‘‘Which model is suited to my problem?”. The question should move to ‘‘Which combination of processes is relevant to this problem?” [48]. Adopting this paradigm has two consequences; firstly the spatial and temporal discretization of the model should be adjusted so as to meet the main objectives of the modelling study and the data availability, and secondly the complexity and spatial heterogeneity of model process representation should be adapted to the model spatial and temporal discretization. Addressing these challenges is one way to reconcile the downward and upward approaches of model conceptualisation that provides pragmatic solutions to modelling problems [66]. Some authors have described this process as a learning framework for hydrological modelling [28]. Progress in computer science, such as object-oriented modelling, and the appearance of modular modelling platforms [3], which allow the user to couple various models in an efficient way, makes the exploration of different model representations more possible now and for the future. This new perspective of pragmatic model development to meet the needs of the application effectively in terms of spatial discretization and process conceptualization will benefit from the availability of new observational techniques that describe landscape heterogeneity and provide hydrological observations at various scales. Weather radar rainfall fields will contribute to progress in this direction. Understanding the variability of storm event rainfalls over catchment areas will improve the quantification of the uncertainties involved in the modelling process by better defining the input uncertainties to models, potentially moving away from the necessarily simplified representation of these estimates in current techniques [36]. Continuous increase in computing power makes uncertainty assessments possible now even for the more complex physically-based models [12]. This should be an important consideration in such studies, especially when using diverse data to assess model structures spatially [35]. 4. Recommendations Besides the necessary development of operational radar networks for rainfall detection and nowcasting purposes, two main recommendations can be formulated to foster the use of weather radar data in hydrological sciences (see also [37]). 4.1. Rainfall re-analyses for hydrological sciences Huge amounts of radar data have been collected for the last 30 years in almost all types of climate and for many regions all over the world. This wealth of data has not yet been fully exploited. Inspired by what has been done for atmospheric data sets (e.g., NCAR or ECMWF meteorological re-analyses), efforts should be encouraged for using archived radar and rain gauge data to produce long-term rainfall space-time data sets with relevant quantification of the associated errors. However, rainfall re-analysis, mainly
971
consisting in gathering large data sets, performing extensive quality checks and combining them to produce rainfall estimates, is a very demanding task and requires a community wide international effort. A first issue that needs to be addressed concerns the data that are available to perform re-analyses. Concerning rain gauge data, rain gauge networks date back to the 1950s. However, the diversity in data quality, type, consistency and accessibility is a serious issue. The effort and cost required to develop and maintain rain gauge networks is significant and there seems to be a trend of reducing the density of rain gauge networks as the radar network is improved. Concerning radar data, there has been a strong evolution of the radar networks during the previous two decades both in terms of density and operating protocols. Moreover radar systems collect huge amounts of numerical data, which raises the issue of efficient, safe and fast access archiving solutions. This issue will become even more acute as polarimetry becomes a standard for operational weather radar networks. Although radar networks date back to the 1980s, exploitable radar data will probably cover a much shorter period due to unreliable numerical archiving. In order to produce data sets that are consistent at the continental scale, the question of the homogeneity of different radar networks is important. In Europe, the OPERA project (http://www.knmi.nl/opera), which aims at harmonizing radar data from 28 countries and at promoting data exchange, is an important initiative in this respect. The second issue concerns the choice of methodologies to apply when performing a rainfall re-analyses from radar and other sources of rainfall data.
Depending on the archived data (raw volume data versus already processed QPE products), the possibility to apply physically-based processing algorithms may vary significantly. Pragmatic and standardised approaches should be implemented to control the stability of the radar signal, to determine the radar detection domain, to identify and remove residual artefacts, to quantify and correct for attenuation effects, and to characterize the vertical profile of reflectivity. Radar–radar merging techniques should receive more attention especially for relatively dense networks operating in rugged topography: an earlymerging of volume data may be preferable to merging the QPE products of various radars. As already mentioned, conventional radar technology does not guarantee an accurate electronic calibration and the Z–R relationship is highly variable. Therefore recourse to other sources of rainfall data is needed to optimize the radar parameters and/or to control the radar QPE quality. Rather than a simple ‘‘radar calibration”, such an operation should be seen as a multi-sensor merging requiring a detailed knowledge of the space-time structure of rainfall and of the instrumental/sampling errors associated with the various sensors used [37,5,73]. The lack of absolute rainfall reference is often invoked; rain gauge networks with adequate densities remain the only practical solution for obtaining such reference values for past data. The issue of representativeness and discrepancies in sampling scales is significant; time integration is a possible way to limit this influence. Other types of rainfall data such as disdrometers, microwave links, and satellite data (in particular TRMM and GPM that offer suitable resolutions) could be progressively incorporated in such merging procedures. A variety of applications can benefit from the possibility of accessing a standardized long-term rainfall re-analyses (in a similar way to the extensive use of atmospheric re-analyses for many
972
Preface / Advances in Water Resources 32 (2009) 969–974
research or operational applications). For example, rainfall re-analyses will be useful for research and engineering applications like the analysis of extremes, the derivation of space-time ‘‘design rainfall”, the detection of climatic trends, the evaluation of meteorological numerical models, the compilation of multiscale water budgets, the forcing of distributed hydrological models. Critical information that has to be provided in conjunction with the reanalyses is the quantification of the uncertainties associated with precipitation estimates to enable ensemble and probabilistic approaches.
4.2. Hydrometeorologic observatories A synergy between various disciplines including hydrology and meteorology, but also neighbouring disciplines such as geochemistry, ecology, geography, applied mathematics is required to achieve progress in process understanding and modelling in hydrology. Better techniques to regionalise the most appropriate model structures and to define in the landscape what drives the dominant modes of hydrological behaviour are critically needed. Hydrometeorologic observatories (CUASHI initiative in the USA – http://www.cuashi.org; HYDRATE project in Europe – http:// www.hydrate.tesaf.unipd.it/; [25]) are now proposed worldwide to provide the required datasets. International scientific experiments like AMMA (http://amma-international.org/) and HyMeX (http://www.cnrm.meteo.fr/hymex) offer additional opportunities for coupled studies on the water cycle including its oceanic, atmospheric and continental components. Operational and research radars, together with the other instruments evoked in Section 2, have the potential to dramatically improve the description of rainfall patterns. Dense raingauge networks are required to further validate radar QPE and characterize their error structure. In addition, hydrologic experimental design should be based on nested instrumented catchments, to address process understanding and model evaluation at various scales as well as understanding the uncertainties in our observational data: Detailed hillslope instrumentation is required to advance the understanding of factors controlling soil moisture redistribution and for upscaling from local to the hillslope scales. Besides conventional instrumentation, hydrogeophysics offers now a variety of new instruments to characterize the soil profiles and the water dynamics. Small- and medium-scale catchments (1–100 km2) instrumentation should include the distributed monitoring of soil moisture, evapotranspiration and the other terms of the water and energy budget. Distributed hydrometry could be achieved using new remote sensing techniques such as large-scale particle imagery velocimetry [22]. Very high resolution satellite images and lidar also offer new opportunities for a better description of surface water pathways. For regional scale catchments (100–10,000 km2), the collection and organisation of all the existing operational data sets is required. Improving the accuracy of flood discharge measurements using 1D/2D hydraulic modelling and image analysis is necessary. Satellite data providing estimate of soil moisture should also be acquired and processed in order to provide a regional view of soil moisture variability. As a complement to detailed instrumentation, post-flood event field campaigns have to be further developed to gather and analysis the large number of evidence left by extreme flood events even on ungauged catchments. Such surveys appear to be an efficient way of progressing the hydrology of extremes and testing distrib-
uted hydrological models beyond their common calibration limits [38,20,14]. The availability of accurate radar QPE is again a key element for successful post-flood surveys. In the next years, one challenge for distributed hydrological models is thus to be able to integrate all these new sources of data together with their error characteristics in a consistent way and to progress in the knowledge of water pathways and fluxes. A second challenge is to tackle the problem of process conceptualisation at various scales (change of scale problem). This must be based on the recognition that when moving to larger scales, the form of the equations should perhaps be modified [47]. McDonnell et al. [54] go one step further in this direction by promoting the adaptation of methodologies borrowed from ecology, focusing on the hydrological functions which can be associated to patterns of heterogeneity. In situ measurements and models will benefit from a mutual enrichment. One of the promising ways to reach this goal is by extending the technique of sensitivity analysis and data assimilation, used in meteorology and oceanography to hydrology. Recent results [17] are very promising and show that these techniques can be useful in improving measurement networks by identifying the places and time where additional observations can be most informative. 5. Overview of the special issue This special issue of Advances in Water Resources contains twelve articles, which represent only a partial sample of the scientific material presented during the International Symposium Weather Radar and Hydrology (http://www.wrah-2008.com/). About radar QPE, Diss et al. [27] offer a convincing assessment of a dual-polarization Doppler X-band radar in mountainous terrain with reference to raingauge measurements and the estimates of an operational S-band radar. With regard to merging techniques, Velasco-Forero et al. [73] propose a non-parametric blending methodology intended to be used in real-time operation for the radar-raingauge estimation while Cummings et al. [23] utilize microwave link measurements for the mean field bias adjustment of an operational radar. Three papers address more fundamental aspects related to the rainfall spatial variability. With X-band polarimetric Doppler measurements, Moreau et al. [58] analyze the rainfall spatial correlation function in order to assess the representativeness error of rain gauges with respect to areal rainfall and to more precisely characterize the radar instrumental error. In the same line, Mandakapa et al. [52] present a theoretical framework for estimating the spatial correlation of the radar rainfall error. Berne et al. [7] develop a method based on indicator variograms to automatically determine the anisotropic rainfall spatial structure from radar data. Two articles are about quantitative precipitation forecast (QPF). This important subject is not specifically addressed in the present preface and the interested readers are referred to [55] for a recent review. Herein, Fabry and Seed [32] analyze the accuracy of rainfall nowcasts from the US 3D radar mosaic and consider weatherdependent predictors to try improving the forecast and/or reducing its uncertainty. Barillec and Cornford [4] introduce a new variational Bayesian assimilation method for precipitation nowcasting using a spatial rainfall model based on radar data. Four articles are about hydrologic modelling with radar rainfall input. Focus is given to extremes for which weather radar offers unprecedented observations. Morin et al. [59] present a robust hydrological model utilizing radar data intended to be used for flash-flood warning in a semi-arid region. Bonnifait et al. [14] discuss the hydrologic and hydraulic modelling exercise that was conducted to put in coherence the various sources of data (including
Preface / Advances in Water Resources 32 (2009) 969–974
operational data and post-event field estimates) collected for an extreme event. Sangati et al. [63] utilize high resolution rainfall fields and a distributed hydrologic model to test the sensitivity of flash-flood simulations to spatial aggregation of rainfall and soil properties over a range of scales. Finally, Cole and Moore [19] perform an ungauged modelling investigation using a distributed hydrologic model and radar rainfall inputs for instrumented nested catchments. Acknowledgements First author would like to thank the members of the scientific and organizing committees for their invaluable help for the organization of the WRaH 2008 international symposium. This event has been sponsored by the Centre National de la Recherche Scientifique, the Institut National des Sciences de l’Univers, Météo France and the Cluster Environnement of the Rhône-Alpes region. A contribution in time was funded by the UK Natural Environment Research Council (NERC), Flood Risk from Extreme Events (FREE) programme with the grant number NE/E002242/1. The guest editors wish to recognize the scientific contributions of the anonymous reviewers to the special issue. References [1] Abbott MB, Bathurst JC, Cunge JA, O’Connell PE, Rasmussen J. An introduction to the European hydrological system – système hydrologique European, (SHE), 1: history and philosophy of a physically-based distributed modelling system. J Hydrol 1986;87:45–59. [2] Anagnostou EM, Anagnostou MN, Krajewski WF, Kruger A, Miriovsky BJ. Highresolution rainfall estimation from X-band polarimetric radar measurements. J Hydrometeorol 2004;5(1):110–28. [3] Argent RM. An overview of model integration for environmental applicationscomponents, frameworks and semantics. Environ Model Soft 2004;19:219–34. [4] Barillec R, Cornford D. Data assimilation for precipitation nowcasting using Bayesian inference, Adv Water Resour 2009;32:1050–65. [Corrected proof, Available online 8 October 2008, ISSN: 0309-1708]. [5] Berenguer M, I Zawadzki. A study of the error covariance matrix of radar rainfall estimates in stratiform rain. Weather Forecast 2008;23:1085–101. [6] Berne A, Uijlenhoet R. Path-averaged rainfall estimation using microwave links: uncertainty due to spatial rainfall variability. Geophys Res Lett 2007;34:L07403. doi:10.1029/2007GL029409. [7] Berne A, Delrieu G, Boudevillain B. Variability of the spatial structure of intense mediterranean precipitation. Adv Water Resour 2009;32:1031–42. [Corrected proof, Available online 3 December 2008, ISSN: 0309-1708]. [8] Beven K. How far can we go in distributed hydrological modelling. Hydrol Earth Syst Sci 2000;5:1–12. [9] Beven K, Binley A. The future of distributed models: model calibration and uncertainty prediction. Hydro Proc 1992;6:279–98. [10] Beven KJ, Freer J. A dynamic TOPMODEL. Hydro Proc 2001;15(10):1993–2011. [11] Beven K, Kirkby MJ. A physically-based variable contributing area model of basin hydrology. Hydrol Sci Bull 1979;24:43–69. [12] Blasone RS, Madsen H, Rosbjerg D. Uncertainty assessment of integrated distributed hydrological models using GLUE with Markov chain Monte Carlo sampling. J Hydrol 2008;353(1-2):18–32. [13] Blöschl G, Sivapalan M. Scales issues in hydrological modelling – a review. Hydrol Proc 1995;9:251–90. [14] Bonnifait L, Delrieu G, Le Lay M, Boudevillain B, Masson A, Belleudy P, et al. Hydrologic and hydraulic distributed modelling with radar rainfall input – reconstruction of the 8–9 September 2002 catastrophic flood event in the Gard region, France. Adv Water Resour 2009;32:1077–89. [15] Bousquet O, Tabary P, Parent du Châtelet J. On the value of operationally synthesized multiple-Doppler wind fields. Geophys Res Lett 2007;34:L22813. doi:10.1029/2007GL030464. [16] Bringi V, Chandrasekar V, Hubbert J, Gorgucci E, Randeu WL, Schoenhuber M. Raindrop size distribution in different climatic regimes from disdrometer and dual-polarized radar analysis. J Atmos Sci 2003;60(2):354–65. [17] Castaings W, Dartus D, Le Dimet FX, Saulnier GM. Sensitivity analysis and parameter estimation for the distributed modelling of infiltration excess overland flow. Hydrol Earth Sys Sci 2007;4:363–405. [18] Ciarapica L, Todini E. TOPKAPI: a model for the representation of the rainfallrunoff process at different scales. Hydrol Proc 2002;16:207–29. [19] Cole SJ, Moore RJ. Distributed hydrological modelling using weather radar in gauged and ungauged basins. Adv Water Resour 2009;32:1107–20. [20] Costa JE, Jarrett RD. An evaluation of selected extraordinary floods in the United States reported by the U.S. geological survey and implications for future advancement of flood science. USGS Scientific Investigations Report 2008–5164, Reston, Virginia; 2008.
973
[21] Creutin JD, Borga M. Radar hydrology modifies the monitoring of flash flood hazard. Hydrol Proc 2003;17:1453–6. [22] Creutin JD, Muste M, Bradley AA, Kim SC, Kruger A. River gauging using PIV technique: proof of concept experiment on the Iowa river. J Hydrol 2003;277:182–94. [23] Cummings RJ, Upton GJG, Holt AR, Kitchen M. Using microwave links to adjust the radar rainfall field. Adv Water Resour 2009;32:1003–10. [Corrected proof, Available online 31 August 2008, ISSN: 0309-1708]. [24] Dehotin J, Braud I. Which spatial discretization for distributed hydrological models? Proposition of a methodology and illustration for medium to large scale catchments. Hydrol Earth Sys Sci 2008;12:769–96. [25] Delrieu G, Ducrocq V, Gaume E, Nicol J, Payrastre O, Yates E, et al. The catastrophic flash-flood event of 8–9 September 2002 in the Gard region, France: a first case study for the Cévennes–Vivarais mediterranean hydrometeorological observatory. J Hydrometeorol 2005;6:34–52. [26] Delrieu G, Boudevillain B, Nicol J, Chapon B, Kirstetter PE, Andrieu H, et al. Bollène 2002 experiment: radar rainfall estimation in the Cévennes–Vivarais region, France. J Appl Meteorol Climatol, in press. [27] Diss S, Testud J, Lavabre J, Ribstein P, Moreau E, Parent du Chatelet J. Ability of dual polarized X-band radar to estimate rainfall. Adv Water Resour 2009;32:975–85. [28] Dunn SM, Freer J, Weiler M, Kirkby MJ, Seibert J, Quinn PF, et al. Conceptualization in catchment modelling: simply learning? Hydrol Proc 2008;22(13):2389–93. [29] Eder G, Sivapalan M, Nachtnebel HP. Modelling water balances in an Alpine catchment through exploitation of emergent properties over changing time scales. Hydrol Proc 2003;17(11):2125–49. doi:10.1002/hyp.1325. [30] Fabry F. On the determination of scale ranges for precipitation fields. J Geophys Res 1996;101(D8):12819–26. [31] Fabry F. Meteorological value of ground targets measurements by radar. J Atmos Oceanic Technol 2004;21:560–73. [32] Fabry F, Seed AW. Quantifying and predicting the accuracy of radar-based quantitative precipitation forecasts. Adv Water Resour 2009;32:1043–49. [Corrected proof, Available online 31 October 2008, ISSN: 0309-1708]. [33] Flügel WA. Delineating hydrological response units by geographical information system analyses for regional hydrological modelling using PRMS/MMS in the drainage basin of the river Bröl in Germany. Hydrol Proc 1995;9(3–4):423–36. [34] Franks SW, Sivapalan M, Takeuchi K, Tachikawa Y. Predictions in ungauged basins: international perspectives on the state of the art and pathways forward, book series, vol. 301. IAHS Publication; 2005. p. 1–348. [35] Freer JE, McMillan H, McDonnell JJ, Beven KJ. Constraining dynamic TOPMODEL responses for imprecise water table information using fuzzy rule based performance measures. J Hydrol 2004;291(3-4):254–77. [36] Kavetski D, Franks S, Kuczera G. In: Duan Q. et al., editor. Confronting input uncertainty in environmental modelling, in calibration of watershed models. AGU; 2002. p. 49–68. [37] Krajewski W, Smith JA. Radar hydrology – rainfall estimation. Adv Water Res 2002;25:1387–94. [38] Gaume E, Borga M. Post-flood field investigations in upland catchments after major flash floods: proposal of a methodology and illustrations. J Flood Risk Manage 2009;1(4):175–89. [39] Germann U, Galli G, Boscacci M, Bolliger M. Radar precipitation measurement in a mountainous region. Q J R Meteorol Soc 2006;132: 1669–92. [40] Gorgucci E, Chandrasekar V, Baldini L. Correction of X-band radar observation for propagation effects based on the self-consistency principle. J Atmos Oceanic Technol 2006;23(12):1668–81. [41] Gorgucci E, Baldini L. Attenuation and differential attenuation correction of Cband radar observations using a fully self-consistent methodology. IEEE Geosci Remote Sensing 2007;4(2):326–30. [42] Gourley JJ, Tabary P, Parent du Chatelet J. A fuzzy logic algorithm for the separation of precipitating from nonprecipitating echoes using polarimetric radar observations. J Atmos Oceanic Technol 2007;24:1439–51. [43] Grayson RB, Moore ID, McMahon TA. Physically based hydrologic modelling 2. Is the concept realistic. Wat Resour Res 1992;28:2659–66. [44] Ivanov VY, Vivoni ER, Bras RL, Entekhabi D. Catchment hydrologic response with a fully distributed triangulated irregular network model. Wat Resour Res 2004;40:W11102. [45] Joss J, Waldvogel A. Precipitation measurement and hydrology. In: Atlas D. editor. Radar in meteorology: Battan memorial and 40th anniversary radar meteorology conference. Am Meteor Soc; 1990. p. 577–606. [46] Jothityangkoon C, Sivapalan M, Farmer DL. Process controls of water balance variability in a large semi-arid catchment: downward approach to hydrological model development. J Hydrol 2001;254:174–98. [47] Kirchner JW. Getting the right answer for the right reasons: linking measurements, analyses, and models to advance the science of hydrology. Wat Resour Res 2006;42:W03S04. [48] Leavesley GH, Markstrom SL, Restrepo PJ, Viger RJ. A modular approach to addressing model design, scale, and parameter estimation issues in distributed hydrological modelling. Hydrol Proc 2002;16:173–87. [49] Leijnse H, Uijlenhoet R, Stricker JNM. Rainfall measurement using radio links from cellular communication networks. Water Resour Res 2007;43:W03201. doi:10.1029/2006WR005631. [50] Leijnse H, Uijlenhoet R, Stricker JNM. Microwave link rainfall estimation: effects of link length and frequency, temporal sampling, power resolution, and
974
[51]
[52]
[53] [54]
[55]
[56] [57]
[58]
[59]
[60]
[61]
[62]
[63]
[64] [65] [66]
[67]
[68] [69]
[70]
[71]
Preface / Advances in Water Resources 32 (2009) 969–974 wet antenna attenuation. Adv Water Resour 2008;31:1481–93. doi:10.1016/ j.advwaterres.2008.03.004. Maki M, Iwanami K, Misumi R, Park SG, Moriwaki H, Maruyama K, et al. Semioperational rainfall observations with X-band multi-parameter radar. Atmos Sci Lett 2005;6:12–8. Mandapaka PV, Krajewski WF, Ciach GJ, Villarini G, Smith JA. Estimation of radar-rainfall error spatial correlation. Adv Water Resour 2009;32:1020–30. [Corrected proof, Available online 11 September 2008, ISSN: 0309-1708]. Matrosov SY, Clark KA, Martner BE, Tokay A. X-band polarimetric radar measurements of rainfall. J Appl Meteor 2002;41(9):941–52. McDonnell JJ, Sivapalan M, Vaché K, Dunn S, Grant G, Haggerty R, et al. Moving beyond heterogeneity and process complexity: a new vision for watershed hydrology. Wat Resour Res 2007;43:W07301. Mecklenburg S, Jurczyk A, Szturc J, Osrodka K. Use of radar observations in hydrological and NWP models. COST action 717: topic WG1-8: quantitative precipitation forecasts (QPF) based on radar data for hydrological models. 2002; Available on-line at:. Messer H, Zinevich A, Alpert P. Environmental monitoring by wireless communication networks. Science 2006;312:713. Moisseev DN, Chandrasekar V. Nonparametric estimation of raindrop size distributions from dual-polarization radar spectral observations. J Atmos Oceanic Technol 2007;24(6):1008–18. Moreau E, Testud J, Le Bouar E. Rainfall spatial variability observed by X-band weather radar and its implication for the accuracy of rainfall estimates. Adv Water Resour 2009;32:1011–19. [Corrected proof, Available online 3 December 2008, ISSN: 0309-1708]. Morin E, Y Jacoby, Navon S, Bet-Halachmi E. Towards flash-flood prediction in the dry Dead Sea region utilizing radar rainfall information. Adv Water Resour 2009;32:1066–76. [Corrected proof, Available online 7 December 2008, ISSN: 0309-1708]. Motovilov YG, Gottschalk L, Engeland K, Rodhe A. Validation of a distributed hydrological model against spatial observations. Agr Forest Meteorol 1999;98– 99:257–77. Reggiani P, Sivapalan M, Hassanizadeh SM. A unifying framework for watershed thermodynamics: balance equations for mass, momentum, energy and entropy, and the second law of thermodynamics. Adv Water Resour 1998;22:367–598. Ryzhkov AV, Schuur TJ, Burgess DW, Heinselman PL, Giangrande SE, Zrnic DS. The joint polarimetric experiment: polarimetric rainfall measurements and hydrometeor classification. Bull Am Meteor Soc 2005;86:809–24. Sangati M, Borga M, Rabuffetti D, Bechini R. Influence of rainfall and soil properties spatial aggregation on extreme flash flood response modelling: an evaluation based on the Sesia river basin, North Western Italy. Adv Water Resour 2009;32:1090–1106. [Accepted manuscript, Available online 1 January 2009, ISSN: 0309-1708]. Sivapalan M. Prediction in ungauged basins: a grand challenge for theoretical hydrology. Hydrol Proc 2003;17:3163–70. Sivapalan M. Process complexity at hillslope scale, process simplicity at the watershed scale: is there a connection? Hydrol Proc 2003;17:1037–41. Smith MB, Georgakakos KP, Liang X, editors. The distributed model intercomparison project (DMIP). J Hydrol 2004;298(1–4):1–334 [Special issue]. Straka JM, Zrnic´ DS, Ryzhkov AV. Bulk hydrometeor classification and quantification using polarimetric radar data: synthesis of relations. J Appl Meteor 2000;39:1341–72. Tabary P. The new French operational radar rainfall product Part 1: Methodology. Weather Forecast 2007;22(3):393–408. Tabary P, Desplats J, Do Khac K, Eideliman F, Guégen C, Heinrich JC. The new French operational radar rainfall product Part 2: Validation. Weather Forecast 2007;22(3):409–27. Testud J, Le Bouar E, Obligis, Ali-Mehenni A. The rain profiling algorithm applied to polarimetric weather radar. J Atmos Oceanic Technol 2000;17:332–56. Troch PA, van Loon E, Paniconi C. Hillslope-storage Boussinesq model for subsurface flow and variable source areas along complex hillslopes 1:
[72]
[73]
[74] [75]
[76]
formulation and characteristic response. Water Resour Res 2003;39:1316. doi:10.1029/2002WR001728. Uijlenhoet R, Steiner M, Smith JA. Variability of raindrop size distributions in a squall line and implications for radar rainfall estimation. J Hydrometeor 2003;4:43–61. Velasco-Forero CA, Sempere-Torres D, Cassiraga EF, Gomez-Hernandez JJ. A non-parametric automatic blending methodology to estimate rainfall fields from rain gauge and radar data. Adv Water Resour 2009;32:986–1002. [Corrected proof, Available online 18 October 2008, ISSN: 0309-1708]. Wood EF, Sivapalan M, Beven K, Band L. Effects of spatial variability and scale with implication to hydrologic modelling. J Hydrol 1988;102:29–47. Zawadzki I. Factors affecting the precision of radar measurements of rain. In: Preprints 22nd conference on radar meteorology, Am Meteor Soc. Zurich, Switzerland; 1984. p. 251–6. Zrnic DS, Ryzhkov AV. Polarimetry for weather surveillance radar. Bull Am Meteor Soc 1999;80:389–406.
Guy Delrieu Laboratoire d’étude des Transferts en Hydrologie et Environnement, UMR 5564 (CNRS, INPG, IRD, UJF), BP 53, F-38041 Grenoble Cedex 9, France E-mail address: [email protected] Isabelle Braud Cemagref, UR HHLY, Lyon, France Alexis Berne Ecole Polytechnique Fédérale de Lausanne, Switzerland Marco Borga Department of Land and Agroforest Environment, University of Padova, Legnaro, Italy Brice Boudevillain Laboratoire d’étude des Transferts en Hydrologie et Environnement, UMR 5564 (CNRS, INPG, IRD, UJF), BP 53, F-38041 Grenoble Cedex 9, France Frédéric Fabry McGill University, Montreal, Canada Jim Freer School of Geographical Sciences, University of Bristol, UK Eric Gaume Laboratoire Central des Ponts et Chaussées, Bouguenais, France Eiichi Nakakita Disaster Prevention Research Institute, University of Kyoto, Japan Alan Seed Center for Australian Weather and Climate Research, Melbourne, Australia Pierre Tabary Centre de Météorologie Radar, Météo-France, Toulouse, France Remko Uijlenhoet Wagenigen University, The Netherlands Available online 29 March 2009
Contents lists available at ScienceDirect
Advances in Water Resources journal homepage: www.elsevier.com/locate/advwatres
Preface
Weather radar and hydrology 1. Introduction The growing concerns about environmental and climate change issues, as well as the emergence of the concept of sustainable development enshrined in the European Water Framework Directive (WFD) legislation, has modified the requirements for future hydrological observation and modelling initiatives and techniques. Originally, the focus at larger catchment scales was on the water stream flow at a limited number of locations, essentially for operational purposes such as flood prediction, dam regulation and hydropower production as well as hydrological process understanding. Water is now considered under its multiple functions such as water supply, hydrological risk, hydropower, irrigation, tourism, ecological value, to name a few. The demand for better hydrological understanding and techniques has thus moved to a more integrated prediction of the water balance components (rainfall, runoff, water storage, transpiration, evaporation, groundwater levels, etc.) at every point within a region [21]. In addition, consideration of land-use and human-induced landscape modifications is now a major concern for water management problems (i.e. for flood forecasting) and impact studies of land use change on stream flow, pollutants and/or sediments transport. For most of these questions, knowledge of the general characteristics of water balance components is not sufficient, more detailed process understanding (i.e. storm responses) throughout the landscape are required. The development of distributed and/or semi-distributed hydrological models, where there is a greater representation of the land-surface heterogeneities and the characterization of distributed inputs, are thus increasingly necessary. The usefulness of distributed hydrological models has been questioned, mainly due to overparameterization problems, parameter estimation and validation limitations [9,43,8]. Such difficulties cannot be ignored but the continuous development of hydrological observation and landscape characterization techniques are likely to offer renewed possibilities in this field. We must recognise there is an inevitable compromise between the complexity of the model proposed, and the spatial data that is available to drive the model (inputs) and robustly evaluate its predictions. We suggest weather radar technology offers a unique means for characterizing the rainfall variability over the range of scales and with the space-time resolutions required for a large variety of hydrological problems. This is especially true for risk management in urban and mountain watersheds characterized by fast dynamics and for which rainfall monitoring and nowcasting are critical [21,37]. For the largest scales, continental rainfields derived from radar networks may also provide valuable information in terms of water budget in complement with climatologic networks and satellite imagery. 0309-1708/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.advwatres.2009.03.006
This article is a preface to a special issue journal that has been initiated in parallel to the International Symposium Weather Radar and Hydrology, held in Grenoble and Autrans, France, in March 2008. This event continues a series of symposia originally called Hydrological Applications of Weather Radar that started in 1989 at the University of Salford, UK. This first symposium was followed by the Hannover (1992), Sao Paulo (1995), San Diego (1998), Kyoto (2001) and Melbourne (2004) meetings. The main objective of the 2008 symposium was to promote the use of quantitative precipitation estimates (QPE) derived from weather radar technology in hydrological sciences and their applications. The symposium was composed of two events: (i) A 3-day conference held on 10–12 March 2008 in Grenoble. The programme was composed of 5 keynote speeches, 52 oral and 80 poster presentations covering the multiple aspects of radar physics, radar QPE and distributed hydrological modelling. This scientific material (abstracts, extended abstracts, oral presentations) can still be accessed on the symposium web site at the following address: http:// www.wrah-2008.com. (ii) A 2-day workshop held on 13–14 March 2008 in Autrans (Vercors). The workshop was dedicated to thematic discussions aimed at summarizing the symposium results and investigating future research directions. In the following, a presentation of the most promising techniques for radar rainfall QPE is proposed in Section 2 prior to a discussion about some of the challenges of distributed hydrological modelling (Section 3). A set of recommendations is made in Section 4 to foster the use of weather radar in hydrologic sciences. Section 5 briefly introduces the articles of the special issue. 2. The most promising techniques for radar rainfall estimation Real-time rainfall detection and estimation as well as rainfall nowcasting have been the main motivations for the deployment of operational weather radar networks over the past six decades. Although the qualitative value of radar networks is unquestioned, the complex technology of weather radar systems combined with the variety of uncertainty sources and the strong variability of precipitation at all scales [30] has limited the quality of radar precipitation estimates [37] and therefore the use of radar QPE in operational hydrology. Measurement artefacts (e.g., clear air echoes, ground and sea clutter, anthropogenic targets such as wind mills, airplanes, etc.) are a first very important limitation for quantitative use. In spite of automated identification techniques now being routinely
970
Preface / Advances in Water Resources 32 (2009) 969–974
implemented for conventional radars (i.e. single-polarization noncoherent Doppler radar in the following), residual artefacts may significantly alter rain reflectivity values and the subsequent radar data processing steps. An agreement exists now about the main radar error sources related to both instrumental and sampling properties [75,45,37]: calibration, attenuation due to wet radome, ground clutter and beam blockages, attenuation due to gas and rain, vertical heterogeneity of the hydrometeors (referred to as the vertical profile of reflectivity –VPR- in the following), reflectivity-rain rate conversion (Z–R relationship), rainfall advection. Volume-scanning protocols and physically-based data processing are now generally implemented operationally to cope with such error sources. Although some noticeable progress has thus been obtained [39,68,69,26] some inherent limitations will continue to exist due for instance to the difficulty in performing an absolute calibration, to the sampling problems (radar resolution varies as a function of range, non-uniform beam filling, non-observation of the low-level altitudes above ground level beyond 60–80 km), to entangled errors (e.g., combination of calibration, attenuation and VPR errors at C-band) and to the high intra-field variability of the Z–R relationship [72]. A major advent in the recent years is the implementation of dual polarization and Doppler techniques in a pre-operational mode for several weather radar networks [76,62,42]. In addition to the reflectivity Z, the differential reflectivity ZDR, the cross correlation qhv and the differential phase shift /DP between the horizontal and vertical polarizations can be measured for each radar resolution volume. It is worth noting that the potential for dualpolarisation techniques to eliminate non-meteorological echoes [42] has actually been the main reason why several countries have decided to deploy a network of polarimetric radars. Classification algorithms are used to distinguish several classes of liquid and frozen hydrometeors [67] prior to the application of specific rainrate estimators using various combinations of the radar measurables [62]. Polarimetry also offers some interesting possibilities such as self-calibration of reflectivity, attenuation correction [70,41] and DSD retrieval [16,57]. Of lesser direct value for radar QPE, the Doppler capability allows for a better description of the rain system dynamics [15]. Refractivity measurements from the phase variation of ground targets [31] also offer potential for the characterization of low-level humidity fields. There is an agreement about the critical importance of calibration and maintenance operations for quantitative applications. The parameters of the radar systems need to be monitored continuously and alarms have to be triggered when predefined thresholds are exceeded. Data from the monitoring system are critical metadata and need to be archived along with the radar observations and products. Polarization means more measured variables and thus more information, but also increased technical complexity and workload in terms of system maintenance. Despite the promises of polarimetry, there is evidence that the density of existing operational radar networks is insufficient for a proper hydrological coverage at the national/continental scale if one considers the 100-km range as a maximum practical limit. Two complementary solutions are currently being investigated: The X-band frequency offers attractive design characteristics compared to the S-band and C-band frequencies (reduced transmitted power and antenna size; reduced sensitivity to ground clutter) that are counterbalanced by a dramatic sensitivity to rain attenuation. In the recent years, a breakthrough has been achieved in radar QPE at X-band through the use of polarimetry and range-profiling algorithms [70,2,40,51,53]. Several industrial (e.g. (http://www.casa.umass.edu/) and research projects have been initiated to further develop this concept which has certainly a great future for radar QPE in rugged topography and/or in urban areas.
Although much less established, microwave links rainfall measurement also offers interesting potential with the deployment of wireless communication networks all over the world [56,49]. A major advantage is related to the fact that such measurements are made close to the ground. First studies indicate that effects of rainfall variability are limited if the link length and frequency are chosen correctly [6,50]. For such settings, the main error sources are related to the wetting of antenna covers and to the resolution of power measurements.
3. Distributed hydrological modelling Recent and ongoing research initiatives such as DMIP (Distributed Model Intercomparison Project; [66]) and PUB (Prediction in ungauged basins; [34]) are focused on distributed hydrological modelling. For PUB the concept is to reduce the uncertainties in the modelling process by improving our understanding and representation in models. We briefly review hereafter some of the related challenges. One of the difficulties in hydrological modelling is that the spatial and temporal scales of hydrological processes span several orders of magnitude: from a few m2 and minutes for soil infiltration to thousands of km2 and several years for groundwater [13]. Furthermore, depending on the climate and the characteristics of the catchment (topography, land use, pedology, etc.), only one or several of these processes may be dominant. For instance, it is usually recognized that runoff (both surface and sub-surface runoff) is dominant in the case of flash-floods, but that evapotranspiration must be accounted for in long-term simulations. According to the scale of interest, the dominant processes might also change. For example while soil hydraulic characteristics are influential for runoff generation through infiltration excess and ponding at the plot scale, the presence of macropores and preferential flow might be the dominant factor when studying the runoff generated at the hillslope scale. Finally, there are several thresholds in hydrology (e.g., appearance of ponding, snow melting, saturated areas, flooding, etc.) which make the modelling more complex due to the associated high level of non-linearities in hydrological responses [65]. A second type of difficulty arises from the combined heterogeneities of the land surface and the sub-soil. The large spatial heterogeneity of the land surface is driven by topography, land use, geology and pedology and increasingly by human actions. To simplify the problem, hydrologists try to define so-called homogeneous zones, i.e. areas where the hydrological behaviour can be considered as homogeneous. This has led to the definition of Hydrological Representative Units (HRU) [33,10], Representative Elementary Areas (REA) [74], Representative Elementary Watersheds [61]. In terms of modelling, this range of concepts has led to various landscape discretizations. Models are mostly based on regular meshes that are usually derived from Digital Terrain Models (e.g. [1,11,60,18]. However, the continental surface heterogeneity is not well described by regular grids and other discretizations have been proposed such as triangular irregular networks [44], hillslopes [71] and hydrolandscapes [24]. For process representations, two approaches can be distinguished [64]. The first one is an upward approach where processes are modelled using equations established at very small scales (such as the Richards equation for water transfer within the saturated/ unsaturated zone, the St-Venant equation for water flow within a river channel, or the Boussinesq equation for water transfer within an aquifer) and extended to larger scales. Effective parameters, relevant to the scale of discretization must therefore be specified if not calibrated [13]. This approach is difficult to verify due to limitations with current hydrological observation techniques (e.g., for subsurface flow processes). Given the non-linearity in hydrological
Preface / Advances in Water Resources 32 (2009) 969–974
processes, it is also difficult to derive theoretical approaches describing how the model parameters might change across scales. The second approach is a downward approach where process conceptualization is derived from the available data, with a minimum of parameters, without representing what happens at smaller scales. The simplest approach to this is data based modelling techniques; other examples of differing complexity of representation can be found in [46,29]. But the establishment of such models requires that data are available at the modelling scale, which is not always the case, and the extrapolation to ungauged catchments and/or potentially more extreme observational events that are not yet characterised within the data is difficult. In terms of modelling practice, a change in modelling paradigm is certainly required as well as accounting for the uncertainties in model prediction. Modellers were trying to answer the following question: ‘‘Which model is suited to my problem?”. The question should move to ‘‘Which combination of processes is relevant to this problem?” [48]. Adopting this paradigm has two consequences; firstly the spatial and temporal discretization of the model should be adjusted so as to meet the main objectives of the modelling study and the data availability, and secondly the complexity and spatial heterogeneity of model process representation should be adapted to the model spatial and temporal discretization. Addressing these challenges is one way to reconcile the downward and upward approaches of model conceptualisation that provides pragmatic solutions to modelling problems [66]. Some authors have described this process as a learning framework for hydrological modelling [28]. Progress in computer science, such as object-oriented modelling, and the appearance of modular modelling platforms [3], which allow the user to couple various models in an efficient way, makes the exploration of different model representations more possible now and for the future. This new perspective of pragmatic model development to meet the needs of the application effectively in terms of spatial discretization and process conceptualization will benefit from the availability of new observational techniques that describe landscape heterogeneity and provide hydrological observations at various scales. Weather radar rainfall fields will contribute to progress in this direction. Understanding the variability of storm event rainfalls over catchment areas will improve the quantification of the uncertainties involved in the modelling process by better defining the input uncertainties to models, potentially moving away from the necessarily simplified representation of these estimates in current techniques [36]. Continuous increase in computing power makes uncertainty assessments possible now even for the more complex physically-based models [12]. This should be an important consideration in such studies, especially when using diverse data to assess model structures spatially [35]. 4. Recommendations Besides the necessary development of operational radar networks for rainfall detection and nowcasting purposes, two main recommendations can be formulated to foster the use of weather radar data in hydrological sciences (see also [37]). 4.1. Rainfall re-analyses for hydrological sciences Huge amounts of radar data have been collected for the last 30 years in almost all types of climate and for many regions all over the world. This wealth of data has not yet been fully exploited. Inspired by what has been done for atmospheric data sets (e.g., NCAR or ECMWF meteorological re-analyses), efforts should be encouraged for using archived radar and rain gauge data to produce long-term rainfall space-time data sets with relevant quantification of the associated errors. However, rainfall re-analysis, mainly
971
consisting in gathering large data sets, performing extensive quality checks and combining them to produce rainfall estimates, is a very demanding task and requires a community wide international effort. A first issue that needs to be addressed concerns the data that are available to perform re-analyses. Concerning rain gauge data, rain gauge networks date back to the 1950s. However, the diversity in data quality, type, consistency and accessibility is a serious issue. The effort and cost required to develop and maintain rain gauge networks is significant and there seems to be a trend of reducing the density of rain gauge networks as the radar network is improved. Concerning radar data, there has been a strong evolution of the radar networks during the previous two decades both in terms of density and operating protocols. Moreover radar systems collect huge amounts of numerical data, which raises the issue of efficient, safe and fast access archiving solutions. This issue will become even more acute as polarimetry becomes a standard for operational weather radar networks. Although radar networks date back to the 1980s, exploitable radar data will probably cover a much shorter period due to unreliable numerical archiving. In order to produce data sets that are consistent at the continental scale, the question of the homogeneity of different radar networks is important. In Europe, the OPERA project (http://www.knmi.nl/opera), which aims at harmonizing radar data from 28 countries and at promoting data exchange, is an important initiative in this respect. The second issue concerns the choice of methodologies to apply when performing a rainfall re-analyses from radar and other sources of rainfall data.
Depending on the archived data (raw volume data versus already processed QPE products), the possibility to apply physically-based processing algorithms may vary significantly. Pragmatic and standardised approaches should be implemented to control the stability of the radar signal, to determine the radar detection domain, to identify and remove residual artefacts, to quantify and correct for attenuation effects, and to characterize the vertical profile of reflectivity. Radar–radar merging techniques should receive more attention especially for relatively dense networks operating in rugged topography: an earlymerging of volume data may be preferable to merging the QPE products of various radars. As already mentioned, conventional radar technology does not guarantee an accurate electronic calibration and the Z–R relationship is highly variable. Therefore recourse to other sources of rainfall data is needed to optimize the radar parameters and/or to control the radar QPE quality. Rather than a simple ‘‘radar calibration”, such an operation should be seen as a multi-sensor merging requiring a detailed knowledge of the space-time structure of rainfall and of the instrumental/sampling errors associated with the various sensors used [37,5,73]. The lack of absolute rainfall reference is often invoked; rain gauge networks with adequate densities remain the only practical solution for obtaining such reference values for past data. The issue of representativeness and discrepancies in sampling scales is significant; time integration is a possible way to limit this influence. Other types of rainfall data such as disdrometers, microwave links, and satellite data (in particular TRMM and GPM that offer suitable resolutions) could be progressively incorporated in such merging procedures. A variety of applications can benefit from the possibility of accessing a standardized long-term rainfall re-analyses (in a similar way to the extensive use of atmospheric re-analyses for many
972
Preface / Advances in Water Resources 32 (2009) 969–974
research or operational applications). For example, rainfall re-analyses will be useful for research and engineering applications like the analysis of extremes, the derivation of space-time ‘‘design rainfall”, the detection of climatic trends, the evaluation of meteorological numerical models, the compilation of multiscale water budgets, the forcing of distributed hydrological models. Critical information that has to be provided in conjunction with the reanalyses is the quantification of the uncertainties associated with precipitation estimates to enable ensemble and probabilistic approaches.
4.2. Hydrometeorologic observatories A synergy between various disciplines including hydrology and meteorology, but also neighbouring disciplines such as geochemistry, ecology, geography, applied mathematics is required to achieve progress in process understanding and modelling in hydrology. Better techniques to regionalise the most appropriate model structures and to define in the landscape what drives the dominant modes of hydrological behaviour are critically needed. Hydrometeorologic observatories (CUASHI initiative in the USA – http://www.cuashi.org; HYDRATE project in Europe – http:// www.hydrate.tesaf.unipd.it/; [25]) are now proposed worldwide to provide the required datasets. International scientific experiments like AMMA (http://amma-international.org/) and HyMeX (http://www.cnrm.meteo.fr/hymex) offer additional opportunities for coupled studies on the water cycle including its oceanic, atmospheric and continental components. Operational and research radars, together with the other instruments evoked in Section 2, have the potential to dramatically improve the description of rainfall patterns. Dense raingauge networks are required to further validate radar QPE and characterize their error structure. In addition, hydrologic experimental design should be based on nested instrumented catchments, to address process understanding and model evaluation at various scales as well as understanding the uncertainties in our observational data: Detailed hillslope instrumentation is required to advance the understanding of factors controlling soil moisture redistribution and for upscaling from local to the hillslope scales. Besides conventional instrumentation, hydrogeophysics offers now a variety of new instruments to characterize the soil profiles and the water dynamics. Small- and medium-scale catchments (1–100 km2) instrumentation should include the distributed monitoring of soil moisture, evapotranspiration and the other terms of the water and energy budget. Distributed hydrometry could be achieved using new remote sensing techniques such as large-scale particle imagery velocimetry [22]. Very high resolution satellite images and lidar also offer new opportunities for a better description of surface water pathways. For regional scale catchments (100–10,000 km2), the collection and organisation of all the existing operational data sets is required. Improving the accuracy of flood discharge measurements using 1D/2D hydraulic modelling and image analysis is necessary. Satellite data providing estimate of soil moisture should also be acquired and processed in order to provide a regional view of soil moisture variability. As a complement to detailed instrumentation, post-flood event field campaigns have to be further developed to gather and analysis the large number of evidence left by extreme flood events even on ungauged catchments. Such surveys appear to be an efficient way of progressing the hydrology of extremes and testing distrib-
uted hydrological models beyond their common calibration limits [38,20,14]. The availability of accurate radar QPE is again a key element for successful post-flood surveys. In the next years, one challenge for distributed hydrological models is thus to be able to integrate all these new sources of data together with their error characteristics in a consistent way and to progress in the knowledge of water pathways and fluxes. A second challenge is to tackle the problem of process conceptualisation at various scales (change of scale problem). This must be based on the recognition that when moving to larger scales, the form of the equations should perhaps be modified [47]. McDonnell et al. [54] go one step further in this direction by promoting the adaptation of methodologies borrowed from ecology, focusing on the hydrological functions which can be associated to patterns of heterogeneity. In situ measurements and models will benefit from a mutual enrichment. One of the promising ways to reach this goal is by extending the technique of sensitivity analysis and data assimilation, used in meteorology and oceanography to hydrology. Recent results [17] are very promising and show that these techniques can be useful in improving measurement networks by identifying the places and time where additional observations can be most informative. 5. Overview of the special issue This special issue of Advances in Water Resources contains twelve articles, which represent only a partial sample of the scientific material presented during the International Symposium Weather Radar and Hydrology (http://www.wrah-2008.com/). About radar QPE, Diss et al. [27] offer a convincing assessment of a dual-polarization Doppler X-band radar in mountainous terrain with reference to raingauge measurements and the estimates of an operational S-band radar. With regard to merging techniques, Velasco-Forero et al. [73] propose a non-parametric blending methodology intended to be used in real-time operation for the radar-raingauge estimation while Cummings et al. [23] utilize microwave link measurements for the mean field bias adjustment of an operational radar. Three papers address more fundamental aspects related to the rainfall spatial variability. With X-band polarimetric Doppler measurements, Moreau et al. [58] analyze the rainfall spatial correlation function in order to assess the representativeness error of rain gauges with respect to areal rainfall and to more precisely characterize the radar instrumental error. In the same line, Mandakapa et al. [52] present a theoretical framework for estimating the spatial correlation of the radar rainfall error. Berne et al. [7] develop a method based on indicator variograms to automatically determine the anisotropic rainfall spatial structure from radar data. Two articles are about quantitative precipitation forecast (QPF). This important subject is not specifically addressed in the present preface and the interested readers are referred to [55] for a recent review. Herein, Fabry and Seed [32] analyze the accuracy of rainfall nowcasts from the US 3D radar mosaic and consider weatherdependent predictors to try improving the forecast and/or reducing its uncertainty. Barillec and Cornford [4] introduce a new variational Bayesian assimilation method for precipitation nowcasting using a spatial rainfall model based on radar data. Four articles are about hydrologic modelling with radar rainfall input. Focus is given to extremes for which weather radar offers unprecedented observations. Morin et al. [59] present a robust hydrological model utilizing radar data intended to be used for flash-flood warning in a semi-arid region. Bonnifait et al. [14] discuss the hydrologic and hydraulic modelling exercise that was conducted to put in coherence the various sources of data (including
Preface / Advances in Water Resources 32 (2009) 969–974
operational data and post-event field estimates) collected for an extreme event. Sangati et al. [63] utilize high resolution rainfall fields and a distributed hydrologic model to test the sensitivity of flash-flood simulations to spatial aggregation of rainfall and soil properties over a range of scales. Finally, Cole and Moore [19] perform an ungauged modelling investigation using a distributed hydrologic model and radar rainfall inputs for instrumented nested catchments. Acknowledgements First author would like to thank the members of the scientific and organizing committees for their invaluable help for the organization of the WRaH 2008 international symposium. This event has been sponsored by the Centre National de la Recherche Scientifique, the Institut National des Sciences de l’Univers, Météo France and the Cluster Environnement of the Rhône-Alpes region. A contribution in time was funded by the UK Natural Environment Research Council (NERC), Flood Risk from Extreme Events (FREE) programme with the grant number NE/E002242/1. The guest editors wish to recognize the scientific contributions of the anonymous reviewers to the special issue. References [1] Abbott MB, Bathurst JC, Cunge JA, O’Connell PE, Rasmussen J. An introduction to the European hydrological system – système hydrologique European, (SHE), 1: history and philosophy of a physically-based distributed modelling system. J Hydrol 1986;87:45–59. [2] Anagnostou EM, Anagnostou MN, Krajewski WF, Kruger A, Miriovsky BJ. Highresolution rainfall estimation from X-band polarimetric radar measurements. J Hydrometeorol 2004;5(1):110–28. [3] Argent RM. An overview of model integration for environmental applicationscomponents, frameworks and semantics. Environ Model Soft 2004;19:219–34. [4] Barillec R, Cornford D. Data assimilation for precipitation nowcasting using Bayesian inference, Adv Water Resour 2009;32:1050–65. [Corrected proof, Available online 8 October 2008, ISSN: 0309-1708]. [5] Berenguer M, I Zawadzki. A study of the error covariance matrix of radar rainfall estimates in stratiform rain. Weather Forecast 2008;23:1085–101. [6] Berne A, Uijlenhoet R. Path-averaged rainfall estimation using microwave links: uncertainty due to spatial rainfall variability. Geophys Res Lett 2007;34:L07403. doi:10.1029/2007GL029409. [7] Berne A, Delrieu G, Boudevillain B. Variability of the spatial structure of intense mediterranean precipitation. Adv Water Resour 2009;32:1031–42. [Corrected proof, Available online 3 December 2008, ISSN: 0309-1708]. [8] Beven K. How far can we go in distributed hydrological modelling. Hydrol Earth Syst Sci 2000;5:1–12. [9] Beven K, Binley A. The future of distributed models: model calibration and uncertainty prediction. Hydro Proc 1992;6:279–98. [10] Beven KJ, Freer J. A dynamic TOPMODEL. Hydro Proc 2001;15(10):1993–2011. [11] Beven K, Kirkby MJ. A physically-based variable contributing area model of basin hydrology. Hydrol Sci Bull 1979;24:43–69. [12] Blasone RS, Madsen H, Rosbjerg D. Uncertainty assessment of integrated distributed hydrological models using GLUE with Markov chain Monte Carlo sampling. J Hydrol 2008;353(1-2):18–32. [13] Blöschl G, Sivapalan M. Scales issues in hydrological modelling – a review. Hydrol Proc 1995;9:251–90. [14] Bonnifait L, Delrieu G, Le Lay M, Boudevillain B, Masson A, Belleudy P, et al. Hydrologic and hydraulic distributed modelling with radar rainfall input – reconstruction of the 8–9 September 2002 catastrophic flood event in the Gard region, France. Adv Water Resour 2009;32:1077–89. [15] Bousquet O, Tabary P, Parent du Châtelet J. On the value of operationally synthesized multiple-Doppler wind fields. Geophys Res Lett 2007;34:L22813. doi:10.1029/2007GL030464. [16] Bringi V, Chandrasekar V, Hubbert J, Gorgucci E, Randeu WL, Schoenhuber M. Raindrop size distribution in different climatic regimes from disdrometer and dual-polarized radar analysis. J Atmos Sci 2003;60(2):354–65. [17] Castaings W, Dartus D, Le Dimet FX, Saulnier GM. Sensitivity analysis and parameter estimation for the distributed modelling of infiltration excess overland flow. Hydrol Earth Sys Sci 2007;4:363–405. [18] Ciarapica L, Todini E. TOPKAPI: a model for the representation of the rainfallrunoff process at different scales. Hydrol Proc 2002;16:207–29. [19] Cole SJ, Moore RJ. Distributed hydrological modelling using weather radar in gauged and ungauged basins. Adv Water Resour 2009;32:1107–20. [20] Costa JE, Jarrett RD. An evaluation of selected extraordinary floods in the United States reported by the U.S. geological survey and implications for future advancement of flood science. USGS Scientific Investigations Report 2008–5164, Reston, Virginia; 2008.
973
[21] Creutin JD, Borga M. Radar hydrology modifies the monitoring of flash flood hazard. Hydrol Proc 2003;17:1453–6. [22] Creutin JD, Muste M, Bradley AA, Kim SC, Kruger A. River gauging using PIV technique: proof of concept experiment on the Iowa river. J Hydrol 2003;277:182–94. [23] Cummings RJ, Upton GJG, Holt AR, Kitchen M. Using microwave links to adjust the radar rainfall field. Adv Water Resour 2009;32:1003–10. [Corrected proof, Available online 31 August 2008, ISSN: 0309-1708]. [24] Dehotin J, Braud I. Which spatial discretization for distributed hydrological models? Proposition of a methodology and illustration for medium to large scale catchments. Hydrol Earth Sys Sci 2008;12:769–96. [25] Delrieu G, Ducrocq V, Gaume E, Nicol J, Payrastre O, Yates E, et al. The catastrophic flash-flood event of 8–9 September 2002 in the Gard region, France: a first case study for the Cévennes–Vivarais mediterranean hydrometeorological observatory. J Hydrometeorol 2005;6:34–52. [26] Delrieu G, Boudevillain B, Nicol J, Chapon B, Kirstetter PE, Andrieu H, et al. Bollène 2002 experiment: radar rainfall estimation in the Cévennes–Vivarais region, France. J Appl Meteorol Climatol, in press. [27] Diss S, Testud J, Lavabre J, Ribstein P, Moreau E, Parent du Chatelet J. Ability of dual polarized X-band radar to estimate rainfall. Adv Water Resour 2009;32:975–85. [28] Dunn SM, Freer J, Weiler M, Kirkby MJ, Seibert J, Quinn PF, et al. Conceptualization in catchment modelling: simply learning? Hydrol Proc 2008;22(13):2389–93. [29] Eder G, Sivapalan M, Nachtnebel HP. Modelling water balances in an Alpine catchment through exploitation of emergent properties over changing time scales. Hydrol Proc 2003;17(11):2125–49. doi:10.1002/hyp.1325. [30] Fabry F. On the determination of scale ranges for precipitation fields. J Geophys Res 1996;101(D8):12819–26. [31] Fabry F. Meteorological value of ground targets measurements by radar. J Atmos Oceanic Technol 2004;21:560–73. [32] Fabry F, Seed AW. Quantifying and predicting the accuracy of radar-based quantitative precipitation forecasts. Adv Water Resour 2009;32:1043–49. [Corrected proof, Available online 31 October 2008, ISSN: 0309-1708]. [33] Flügel WA. Delineating hydrological response units by geographical information system analyses for regional hydrological modelling using PRMS/MMS in the drainage basin of the river Bröl in Germany. Hydrol Proc 1995;9(3–4):423–36. [34] Franks SW, Sivapalan M, Takeuchi K, Tachikawa Y. Predictions in ungauged basins: international perspectives on the state of the art and pathways forward, book series, vol. 301. IAHS Publication; 2005. p. 1–348. [35] Freer JE, McMillan H, McDonnell JJ, Beven KJ. Constraining dynamic TOPMODEL responses for imprecise water table information using fuzzy rule based performance measures. J Hydrol 2004;291(3-4):254–77. [36] Kavetski D, Franks S, Kuczera G. In: Duan Q. et al., editor. Confronting input uncertainty in environmental modelling, in calibration of watershed models. AGU; 2002. p. 49–68. [37] Krajewski W, Smith JA. Radar hydrology – rainfall estimation. Adv Water Res 2002;25:1387–94. [38] Gaume E, Borga M. Post-flood field investigations in upland catchments after major flash floods: proposal of a methodology and illustrations. J Flood Risk Manage 2009;1(4):175–89. [39] Germann U, Galli G, Boscacci M, Bolliger M. Radar precipitation measurement in a mountainous region. Q J R Meteorol Soc 2006;132: 1669–92. [40] Gorgucci E, Chandrasekar V, Baldini L. Correction of X-band radar observation for propagation effects based on the self-consistency principle. J Atmos Oceanic Technol 2006;23(12):1668–81. [41] Gorgucci E, Baldini L. Attenuation and differential attenuation correction of Cband radar observations using a fully self-consistent methodology. IEEE Geosci Remote Sensing 2007;4(2):326–30. [42] Gourley JJ, Tabary P, Parent du Chatelet J. A fuzzy logic algorithm for the separation of precipitating from nonprecipitating echoes using polarimetric radar observations. J Atmos Oceanic Technol 2007;24:1439–51. [43] Grayson RB, Moore ID, McMahon TA. Physically based hydrologic modelling 2. Is the concept realistic. Wat Resour Res 1992;28:2659–66. [44] Ivanov VY, Vivoni ER, Bras RL, Entekhabi D. Catchment hydrologic response with a fully distributed triangulated irregular network model. Wat Resour Res 2004;40:W11102. [45] Joss J, Waldvogel A. Precipitation measurement and hydrology. In: Atlas D. editor. Radar in meteorology: Battan memorial and 40th anniversary radar meteorology conference. Am Meteor Soc; 1990. p. 577–606. [46] Jothityangkoon C, Sivapalan M, Farmer DL. Process controls of water balance variability in a large semi-arid catchment: downward approach to hydrological model development. J Hydrol 2001;254:174–98. [47] Kirchner JW. Getting the right answer for the right reasons: linking measurements, analyses, and models to advance the science of hydrology. Wat Resour Res 2006;42:W03S04. [48] Leavesley GH, Markstrom SL, Restrepo PJ, Viger RJ. A modular approach to addressing model design, scale, and parameter estimation issues in distributed hydrological modelling. Hydrol Proc 2002;16:173–87. [49] Leijnse H, Uijlenhoet R, Stricker JNM. Rainfall measurement using radio links from cellular communication networks. Water Resour Res 2007;43:W03201. doi:10.1029/2006WR005631. [50] Leijnse H, Uijlenhoet R, Stricker JNM. Microwave link rainfall estimation: effects of link length and frequency, temporal sampling, power resolution, and
974
[51]
[52]
[53] [54]
[55]
[56] [57]
[58]
[59]
[60]
[61]
[62]
[63]
[64] [65] [66]
[67]
[68] [69]
[70]
[71]
Preface / Advances in Water Resources 32 (2009) 969–974 wet antenna attenuation. Adv Water Resour 2008;31:1481–93. doi:10.1016/ j.advwaterres.2008.03.004. Maki M, Iwanami K, Misumi R, Park SG, Moriwaki H, Maruyama K, et al. Semioperational rainfall observations with X-band multi-parameter radar. Atmos Sci Lett 2005;6:12–8. Mandapaka PV, Krajewski WF, Ciach GJ, Villarini G, Smith JA. Estimation of radar-rainfall error spatial correlation. Adv Water Resour 2009;32:1020–30. [Corrected proof, Available online 11 September 2008, ISSN: 0309-1708]. Matrosov SY, Clark KA, Martner BE, Tokay A. X-band polarimetric radar measurements of rainfall. J Appl Meteor 2002;41(9):941–52. McDonnell JJ, Sivapalan M, Vaché K, Dunn S, Grant G, Haggerty R, et al. Moving beyond heterogeneity and process complexity: a new vision for watershed hydrology. Wat Resour Res 2007;43:W07301. Mecklenburg S, Jurczyk A, Szturc J, Osrodka K. Use of radar observations in hydrological and NWP models. COST action 717: topic WG1-8: quantitative precipitation forecasts (QPF) based on radar data for hydrological models. 2002; Available on-line at:
[72]
[73]
[74] [75]
[76]
formulation and characteristic response. Water Resour Res 2003;39:1316. doi:10.1029/2002WR001728. Uijlenhoet R, Steiner M, Smith JA. Variability of raindrop size distributions in a squall line and implications for radar rainfall estimation. J Hydrometeor 2003;4:43–61. Velasco-Forero CA, Sempere-Torres D, Cassiraga EF, Gomez-Hernandez JJ. A non-parametric automatic blending methodology to estimate rainfall fields from rain gauge and radar data. Adv Water Resour 2009;32:986–1002. [Corrected proof, Available online 18 October 2008, ISSN: 0309-1708]. Wood EF, Sivapalan M, Beven K, Band L. Effects of spatial variability and scale with implication to hydrologic modelling. J Hydrol 1988;102:29–47. Zawadzki I. Factors affecting the precision of radar measurements of rain. In: Preprints 22nd conference on radar meteorology, Am Meteor Soc. Zurich, Switzerland; 1984. p. 251–6. Zrnic DS, Ryzhkov AV. Polarimetry for weather surveillance radar. Bull Am Meteor Soc 1999;80:389–406.
Guy Delrieu Laboratoire d’étude des Transferts en Hydrologie et Environnement, UMR 5564 (CNRS, INPG, IRD, UJF), BP 53, F-38041 Grenoble Cedex 9, France E-mail address: [email protected] Isabelle Braud Cemagref, UR HHLY, Lyon, France Alexis Berne Ecole Polytechnique Fédérale de Lausanne, Switzerland Marco Borga Department of Land and Agroforest Environment, University of Padova, Legnaro, Italy Brice Boudevillain Laboratoire d’étude des Transferts en Hydrologie et Environnement, UMR 5564 (CNRS, INPG, IRD, UJF), BP 53, F-38041 Grenoble Cedex 9, France Frédéric Fabry McGill University, Montreal, Canada Jim Freer School of Geographical Sciences, University of Bristol, UK Eric Gaume Laboratoire Central des Ponts et Chaussées, Bouguenais, France Eiichi Nakakita Disaster Prevention Research Institute, University of Kyoto, Japan Alan Seed Center for Australian Weather and Climate Research, Melbourne, Australia Pierre Tabary Centre de Météorologie Radar, Météo-France, Toulouse, France Remko Uijlenhoet Wagenigen University, The Netherlands Available online 29 March 2009