Variability of the spatial structure of intense Mediterranean precipitation

https://doi.org/10.1016/j.advwatres.2008.11.008Get rights and content

Abstract

Intense Mediterranean precipitation can generate devastating flash floods. A better understanding of the spatial structure of intense rainfall is critical to better identify catchments that will produce strong hydrological responses. We focus on two intense Mediterranean rain events of different types that occured in 2002. Radar and rain gauge measurements are combined to have a data set with a high spatial (1 × 1 km2) and temporal (5 min) resolution. Two thresholds are determined using the quantiles of the rain rate values, corresponding to the precipitating system at large and to the intense rain cells. A method based on indicator variograms associated with the thresholds is proposed in order to automatically quantify the spatial structure at each time step during the entire rain events. Therefore, its variability within intense rain events can be investigated. The spatial structure is found to be homogeneous over periods that can be related to the dynamics of the events. Moreover, a decreasing time resolution (i.e., increasing accumulation period) of the rain rate data will stretch the spatial structure because of the advection of rain cells by the wind. These quantitative characteristics of the spatial structure of intense Mediterranean rainfall will be useful to improve our understanding of the dynamics of flash floods.

Introduction

Mediterranean coastal regions of France are regularly affected by intense rain events producing several hundreds of mm of rain in tens of hours. These extreme rainfall amounts lead to sudden and devastating flash floods that cause casualties and huge damage. During the last 20 years, such catastrophic flood events occured in Nîmes in 1988 [1], Vaison-la-Romaine in 1992 [2], Aude in 1999 [3] and Gard in 2002 [4]. The Cévennes–Vivarais region (see Section 2) in particular is prone to intense precipitation and flash floods.

During these intense Mediterranean rain events, the spatial and temporal variability of rainfall has been identified as the main controlling factor of the hydrological response of catchments (e.g. [5], [6]). It is therefore of primary importance to quantitatively characterize the variability and the structure of intense Mediterranean precipitation in order to improve the understanding, the modeling and the forecasting of rain events generating flash floods. The structure of precipitation has been a question of interest for many years and extensive work has been and is still carried on this topic (e.g. [7], [8], [9], [10], [11]).

In this paper, we use geostatistics to investigate the spatial structure of rainfall. The geostatistical framework has been proved to be suited to rainfall studies (e.g. [12], [13], [14], [15], [16]). In particular, geostatistics can be used to quantify the variability of precipitation (e.g. [17], [18], [19]). According to ([20] p. 3), geostatistical structural analysis aims at answering the following questions:

“What does an observation at a point tell us about the values at neighboring points? Can we expect continuity in a mathematical sense, or in a statistical sense, or no continuity at all? What is the signal-to-noise ratio? Are variations similar in all directions or is there anisotropy? Do the data exhibit any spatial trend? Are there characteristic scales and what do they represent? Is the histogram symmetric or skewed?”

All these questions are relevant for rainfall structure analysis, but we will focus on the characteristic scales and the anisotropy of intense Mediterranean precipitation, as well as on their variability during intense rain events. This information is crucial for a better understanding of the hydrological responses of catchments affected by intense rainfall.

The main objectives of the present paper are (1) to implement an automatic technique to quantify the structure of rainfall because of the large data set we use, and (2) to analyze the variability of the structure during a rain event and relate this variability to the dynamics of the event. Section 2 presents the study area and the data set. Section 3 describes the methodology developed to quantify the structure of rainfall and the results are discussed in Section 4. Finally, the conclusions are presented in Section 5.

Section snippets

Study area and data

The Cévennes–Vivarais region is located on the West bank of the river Rhône and close to the Mediterranean sea in France. It has hence a Mediterranean climate characterized by rainy autumns. The topography of the region is rugged with altitudes ranging from sea level to 1700 m at the top of Mont Lozère. The combination of Mediterranean influence with topography favors the occurence of intense rain events and consequently of devastating flash floods.

The main objective of the Cévennes–Vivarais

Methodology

In the present paper, we adopt a geostatistical view of precipitation: the rainfall field is considered as a realization of a random function. The key geostatistical tool for structural analysis is the variogram, which quantifies the continuity and the cofluctuation of the data. The variogram is a more general structural tool than the covariance: only intrisic stationarity (i.e., second order stationarity of the increments of the random function) is necessary for the variogram to be defined

Temporal variability of the spatial structure during rain events

Applying the methodology described in the previous section, the mean length is estimated in 28 directions and for a given threshold for each time step during the two studied rain events. This quantitatively characterizes the spatial structure of rainfall fields. The objective of this section is to analyze the evolution of the spatial structure during a rain event and to identify periods with distinct structures due to the dynamics of precipitation. In this section, we focus on the 5-min time

Conclusions

The structure of intense precipitation, and in particular the characteristic scale and the anisotropy, is of primary importance for a better understanding, modeling and forecasting of intense rainfall events and resulting flash floods. In this paper, we analyze the variability of the spatial structure during intense Mediterranean precipitation.

We focus on two intense rain events that occured on the 8–9 September 2002 and on the 11 November 2002. The 20020908 rain event was a mesoscale

Acknowledgements

The authors acknowledge the financial support from the French Project CYPRIM and the EU Project FLOODsite (GOCE-CT-2004-505420).

References (30)

  • P.M. Austin et al.

    Analysis of the structure of precipitation patterns in New England

    J Appl Meteor

    (1972)
  • P.V. Hobbs et al.

    The mesoscale and microscale structure and organization of clouds and precipitation in midlatitude cyclones. I: a case study of a cold front

    J Atmos Sci

    (1980)
  • R.K. Crane

    Space-time structure of rain rate fields

    J Geophys Res

    (1990)
  • M. Steiner et al.

    Climatological characterization of three-dimensional storm structure from operational radar and rain gauge data

    J Appl Meteor

    (1995)
  • W. Berg et al.

    Differences between East and West Pacific rainfall systems

    J Climate

    (2002)
  • Cited by (26)

    • Statistical characterization of rainfall fields based upon a 12-year high-resolution radar archive of Belgium

      2023, Atmospheric Research
      Citation Excerpt :

      Two main types of approaches have been employed: Eulerian approach, where the statistical analysis is performed to the rainfall time series obtained in fixed grid points or areas; and Lagrange approach, where rainfall fields are segmented in different objects/entities whose physical properties are identified and studied. Amongst the different Eulerian methods, the use of spatial distribution occurrence frequency statistics (Chen et al., 2014; Goudenhoofdt and Delobbe, 2013; Overeem et al., 2009), spatial intensity–duration–frequency curves (Marra and Morin, 2015; Paixao et al., 2015), areal reduction factors (Svensson and Jones, 2010; Thorndahl et al., 2019) and variograms (Berne et al., 2009; Emmanuel et al., 2012), are one the most used approaches. They can be utilized to extract and provide the necessary information for applications such as identification of vulnerable areas (Davini et al., 2011), analysis of diurnal patterns (Chen et al., 2014; Fairman et al., 2017), and construction of design storms for urban drainage calculations (Bruni et al., 2015; Svensson and Jones, 2010).

    • Meteorological systems producing rainfall

      2022, Rainfall: Modeling, Measurement and Applications
    • Impact of temporal resolution of inputs on hydrological model performance: An analysis based on 2400 flood events

      2016, Journal of Hydrology
      Citation Excerpt :

      The highest values of the daily precipitation intensity coefficient indicate the occurrence of intense convective precipitation events and are concentrated mainly above the Cévennes region and in the south-eastern Mediterranean area. In the same area, the lowest values of the flow auto-correlation are found, corresponding to catchments presenting the fastest response to precipitation with frequent flash floods (Berne et al., 2009; Braud et al., 2014; Delrieu et al., 2005; Javelle et al., 2010; Saulnier and Le Lay, 2009). The catchment set includes as many catchments as possible respecting the selection criteria discussed above, with the objective to get general results, i.e. not too much dependent on specific catchment types or climate conditions.

    • Empirical study of extreme rainfall intensity in a semi-arid environment at different time scales

      2014, Journal of Arid Environments
      Citation Excerpt :

      As a result flash-floods of ephemeral streams are highly dependent on rainfall intensity (Bracken et al., 2008; Camarasa and Tilford, 2002; Cudennec et al., 2007; Kokkonen et al., 2004). Despite the importance of intensity, it is not easy to identify reference thresholds (Montesarchio et al., 2009) because the characteristics of rainfall differ according to the time scale used for observation (Berne et al., 2009; Dunkerley, 2008; Jebari et al., 2007; Valdés et al., 1985; Waymire and Gupta, 1981). Nonetheless, it is by no means clear what the most appropriate time interval for measuring rainfall intensity is.

    • Temporal and spatial variability of rainfall at the urban hydrological scale

      2012, Journal of Hydrology
      Citation Excerpt :

      Moreau et al. (2009) summarized the spatial variability of rainfall of a set of observed rain events by a mean correlogram derived from both rain gauge data (25 gauges over a 16 km × 16 km surface area) and radar data with short-time resolutions (ranging from 6 min to 1 h). Berne et al. (2004, 2009) applied a geostatistical method to study the spatial and temporal structure of Mediterranean rainfall events from rain gauge and radar data. Berne et al. (2004) also investigated the temporal and spatial scales of different urban catchments and assessed, based on this analysis, the required temporal and spatial resolution of rain measurements for urban hydrological applications.

    • Weather radar and hydrology

      2009, Advances in Water Resources
    View all citing articles on Scopus
    View full text