Você está na página 1de 11

Atmospheric Research 77 (2005) 367 377 www.elsevier.

com/locate/atmos

Design storm prediction and hydrologic modeling using a web-GIS approach on a free-software platform
E.M. Castrogiovannia,*, G. La Loggiab, L.V. Notob
a

Regione Siciliana-Ufficio Idrografico Regionale, via G. Bonsignore 1, I-90135, Palermo, Italy b Dipartimento di Ingegneria Idraulica ed Applicazioni Ambientali, University of Palermo, Viale delle Scienze, Palermo, Italy

Received 13 April 2004; received in revised form 15 December 2004; accepted 16 December 2004

Abstract The aim of this work has been to implement a set of procedures useful to automatise the evaluation, the design storm prediction and the flood discharge associated with a selected risk level. For this purpose a Geographic Information System has been implemented using Grass 5.0. One of the main topics of such a system is a georeferenced database of the highest intensity rainfalls and their assigned duration recorded in Sicily. This database contains the main characteristics for more than 250 raingauges, as well as the values of intense rainfall events recorded by these raingauges. These data are managed through the combined use of the PostgreSQL and GRASS-GIS 5.0 databases. Some of the best-known probability distributions have been implemented within the Geographical Information System in order to determine the point and/or areal rain values once duration and return period have been defined. The system also includes a hydrological module necessary to compute the probable flow, for a selected risk level, at points chosen by the user. A peculiarity of the system is the possibility to querying the model using a web-interface. The assumption is that the rising needs of geographic information, and dealing with the rising importance of peoples participation in the decision process, requires new forms for the diffusion of territorial data. Furthermore, technicians as well as public administrators needs to get customized and specialist data to support planning, particularly in emergencies. In this perspective a Web-interface has been developed

* Corresponding author. Tel.: +39 091 7079583. E-mail addresses: mcastrogiovanni@uirsicilia.it (E.M. Castrogiovanni), glal@idra.unipa.it (G. La Loggia), valerio@idra.unipa.it (L.V. Noto). 0169-8095/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.atmosres.2004.12.015

368

E.M. Castrogiovanni et al. / Atmospheric Research 77 (2005) 367377

for the hydrologic system. The aim is to allow remote users to access a centralized database and processing-power to serve the needs of knowledge without complex hardware/software infrastructures. D 2005 Elsevier B.V. All rights reserved.
Keywords: GIS; Hydrology; GRASS; Free-software; Flood

1. Introduction Owing to GIS development, a new challenge must be accepted by water resources engineers and hydrologists: to link GIS and hydrological models, or, better, to redefine hydrological models in a spatial context in order to build better models, based on GIS techniques. Nowadays two different approaches are commonly used: the first is to couple current models with GIS, using conversion programs, or user interfaces. In this way GIS data are converted to the specific input format needed by the model. The model is run, and the output is converted from the input format back into the GIS. In this case there is no need to change the model, but the procedure is time-consuming, the model input files may become extremely large, and errors in the model input file are difficult to find. The second approach goes one step further integrating of models in a GIS; it is also called embedded coupling. In this case, with current GIS, mathematical modelling is limited by the power of the user interface. Moreover, if the integration is aimed at giving a wider diffusion of technical information, this would enlarge the community (researchers, scientists, public administrators, the general public), thus spreading the knowledge. It is clear that such diffusion has to be achieved using low cost systems able to perform the computation required, without the direct intervention of the people looking for that information. For these reasons, the aim of this study has been to build up an embedded hydrologic modellingoriented web-GIS, implemented by a free-software platform. To be more specific, GRASS has been adopted as the GIS platform, and GNU/Linux as the operating system: the user (acting on a client browser) can acquire information on intense rainfall events of given duration and return periods, and/or a flood hydrograph for a given return period at a selected point by a simple click on a map shown by the browser. The root of this system is a georeferenced database of the highest intensity rainfall events and their assigned duration recorded in Sicily (Italy). This database contains the main characteristics of these stations, beside the values of the intense rainfalls (maximum annual rainfall depths for fixed duration) for more than 250 raingauges managed by the Regional Hydrographic Office (Ufficio Idrografico della Regione Siciliana). A 20 20 m DEM (Digital Elevation Model) of Sicily represents the source of information used to perform all the morphology-related computations aiming to obtain flow directions, subbasin areas, timearea curves, and so on. Some of the best-known probability distributions have been implemented within the Information System to determine the point and/or areal rain values, once duration and return time have been selected. In order to perform faster computations, the hydrologic input data are given by means of raster data related to the spatial distribution of coefficients for selected probability distributions.

E.M. Castrogiovanni et al. / Atmospheric Research 77 (2005) 367377

369

The web-interface is based on a server-side processing approach in which PHP is the main interpreter language. Requests received by the server from the client browser via internet are delivered to GRASS-GIS; requests instigate the writing of set of GRASScommands in the form of GRASS-script and executable BASH scripts, thus scripts are executed by GRASS-GIS and output is given back to the client.

2. Methodology The development of a model using an only-free-software platform implies a revision of the commonly used processing-tools and, with them, changes in previously developed algorithms. At first, the study was directed to reinvent an appropriate methodology: in terms of the type of basic input data (e.g. accuracy of DEM needed), number/type of steps needed, and verification of the consistency of results. Once steps were coded, validation of results was made by comparing results with real-data and with the ones obtained using different models. Developing a set of pre-processed support data, a series of steps was automatised, by grouping GIS-software commands through scripts. In this way, the model was fully defined for running in local mode. A further step was to adapt model turning it into a web-interrogable-model.

3. Characteristics of the GIS based hydrologic model The approach used in order to obtain the flood hydrograph for a given return period at a given section of a river basin is a typical indirect method approach: from this point of view, the first element deserving attention is the spatial distribution of maximum intensity rainfall events for a given duration and return period. The need to implement this distribution using GIS-techniques leads to the definition of a dedicated analysis module (Noto and La Loggia, 2001) (rainfall module shown in Fig. 1). This module can be used in two distinct ways: through its own interface or by using analysis routines in the embedded mode. By means of a specific panel it is possible to extrapolate the characteristics of extreme rainfall events for a given duration and return period with more than one representation scale. Particularly, this information is provided for each of the points of the selected subbasin in the form of IDF (IntensityDurationFrequency) or DDF (DepthDuration Frequency) curves, as mean of values in the same basin, or as the mean of regional values, using one of the probability distributions implemented (GEV, TCEV; EVI, EVII). All the prediction models provided by these probability distributions were implemented assuming scale invariance (Feder, 1988; Rosso and Burlando, 1990) of the rainfall process. This means that the physical property in which the ratio of rainfall depths, having kt and t duration respectively, is just equal to the k scale factor raised to the coefficient n that is a characteristic of the examined site. DDF curves, in the case of scale invariance, can be written as: hT d w T m l d n 1

370

E.M. Castrogiovanni et al. / Atmospheric Research 77 (2005) 367377

Fig. 1. Working scheme of hydrologic module.

where m l represents the scale coefficient of the DDF, which is given by the expected value of maximum annual rainfall with duration of 1 h, n is the scale power with which the analyzed phenomenon is transmitted from the basic temporal scale (hourly) to the other temporal scales, and w T is the frequency factor, since it depends on the return period T and the chosen distribution. In particular, Gumbel distribution (Extreme Value I, EVI) frequency factor w T is given by: &   !' T wT 1 CV 0:45 0:779ln ln 2 T 1

E.M. Castrogiovanni et al. / Atmospheric Research 77 (2005) 367377

371

where CV is the coefficient of variation of storm depths of any duration. In the case of Gibrat-Galton distribution, the frequency factor w T is given by: nq o exp ln 1 CV2 U1 T 1=T p wT 3 1 CV2 where U 1 is the inverse of the standard normal distribution function. Finally, in the case of Frechet distribution (Extreme Value II, EVII), the frequency factor w T is:    1=k H1 T wT C 4 ln H T 1 where C () is the Gamma function and the parameter k is connected to the coefficient of variation through following expression: s C1 2=k CV 1 5 C2 1 1=k The statistic analysis of intense rainfalls, developed by using the TCEV (Two Component Extreme Value) probability distribution for Sicily (Rossi et al., 1984; Cannarozzo et al., 1990, 1995) allows the determination of IDF/DDF curves at ungauged sites and also at gauged sites with short record lengths. The application of this probability distribution focused on three different and hierarchical regionalization procedures based on three different and sequential levels. Even in the case of TCEV, the DDF curves are obtained by Eq. (1) but the frequency factor depends on the rainfall duration besides on the return period; the frequency factor expression is given by: wT ;d b0 b1 d c2 d 2 c1 d c0 lnT 6 where b i and c i (i = 0, 1, 2) are parameters that are typical of homogenous zones defined by the same authors of the study. A procedure finalized to get the DDF curves from these regionalized variables was implemented within the Information System, by updating the authors original statistic analyses concerning the third regionalization level (fixing of coefficient m 1 and n ) with maximum annual rainfall depths until 2001. The system uses the L-moments method (Hosking, 1990) to estimate parameters (Rossi et al., 1984). The reason for this is that L-moments are less prone to adverse sampling effects (short historical series and measurement errors) than classic moments (Kottegoda and Rosso, 1998). The parameters needed by the rainfall module are essentially m 1, n and, for all the distributions except for TCEV, the coefficient of variation CV (needed to calculate the frequency factor w T ). These parameters are automatically obtained inside the database for all the raingauges. For the estimate of the parameters for the ungauged sites a spatial interpolation algorithm was implemented. This algorithm is built joining two different methods through a weighted average with the weight defined by user: the first is kriging (Matheron, 1971),

372

E.M. Castrogiovanni et al. / Atmospheric Research 77 (2005) 367377

the second is based on the weighted average of known parameters for nearest sites. The weights used in the second method depend on the size of the available sample at the nearest sites (Durrans and Brown, 2002). Using the kriging method alone would be absolutely acceptable if the parameter values in gauged site were totally free from error and if the lengths of record of the available samples for all the sites was the same. Unfortunately, this does not happen, so information about the reliability and the size of the samples are needed. Thanks to this interpolation algorithm it is possible to arrange three grid layers with a resolution of 500 m showing the spatial distribution of the parameters m 1, n and (if necessary) CV that are necessary for point calculation of the DDF. When the system is used at a basin scale it is necessary to obtain mean areal rainfall depth; for this purpose an areal reduction factor obtained by World Meteolorological Organization (WMO, 1983) is used in order to consider problems involved in the spatial distribution of rainfall. Starting from the DDF curves for an assigned return period it is possible to get the hyetograph design using the Chicago method (Keifer and Chu, 1957). A second module (called flood module ), with a dedicated interface, has been developed by merging rainfall spatial distribution values with geomorphological basin data. For a given point (and thus a drainage basin), the model uses data from the rainfall module with an assigned return period (provided by web-user): the rainfall duration is equal to the concentration time of the watershed. With regard to the transformation of rainfall to runoff (hydroloss module shown in Fig. 1), two models have been included in the proposed methodology: the runoff coefficient W and the SCS-CN method (U.S. Dept. Agric., S.C.S., 1972). The first allows the abstraction of the design hyetograph by means of a constant coefficient which depends on the percent imperviousness, slope and ponding character of the watershed surface. The second, developed by the Soil Conservation Service, calculates the excess rainfall on the basis of the dimensionless parameter CN, called curve number, determined using the soil type and land use information contained within the system. The transformation of rainfall excess to direct runoff is carried out using the unit hydrograph (UH) (Sherman, 1932) derived by means of the topographic information contained in the DEM. This work, which starts with the automatic extraction of the drainage network and delineation of basin, preliminarily requires the correction of the DEM. Some authors (Band, 1986; Burrough, 1986) report that the results of these operations are often not completely satisfactory because of the poor quality of the DEM used; the presence of singularity zones like depressions, at areas and flow blockages causes a sure failure of any automated methodology for the extraction of hydraulic features. Once the sinks in the DEM are removed by breaching and filling, the resulting at surface can be interpreted to define the flow direction, using the singleflow algorithm (Jenson and Domingue, 1988) and it is possible to get drainage area and the drainage network by a specific GRASS module which requires the coordinates of the basin-outlet only. The determination of flow direction enables the realisation of a layer of accumulated flow for each cell, by accumulating all cells that flow into each downslope cell. The basin relative to the fixed outlet point can be easily determined using flow direction and flow accumulation layer previously created. The availability of the created layer enables the

E.M. Castrogiovanni et al. / Atmospheric Research 77 (2005) 367377

373

determination of a layer which contains, for each cell, the downstream distance from the outlet along a flow path. GRASS tools, combined with scripts, permit to obtain downstream distance as 3Dlength of water-flow from a generic point of the basin to the relative sub-basin outlet. This distance is linked to the watershed concentration time by means of the flow velocity within each cell. This parameter, depending only on local variables (contributing area, slope, etc.) that do not vary over time (i.e. discharge, rainfall, storage, etc.), is assumed to be constant over the whole watershed at 11.5 m/s for the sake of simplicity. Using map algebra operations characterising most common GIS packages, the isochrone layer can be obtained by dividing the flow path length by this velocity parameter. All of the steps described above, schematised in Fig. 1, give the timearea curve A (t ) as the end product; this shows the watershed area (in km2) contributing to runoff production at time t (in hours). The unit hydrograph (UH) U (t ) can be easily derived by the timearea curve using the following equation: U iDt AiDt Ai 1Dt : Dt 7

Once the unit hydrograph has been determined, it can be applied to compute the streamflow hydrographs Q n given by the following discrete convolution equation between rainfall hyetograph (H m ) and unit hydrograph (U m ): n X Qn Hm Unm1 : 8
m 1

Stream network is introduced into model as raster files and geographic coordinates of the network cells are introduced in a database table. When the user clicks on a point of the map containing the stream network, the system computes distances from the point selected and each of the stream cells: the flood hydrograph is calculated for the closest stream cell. This method is used to permit significant values of accumulation to be obtained only and to avoid the flow accumulation calculation on hillslope.

4. Characteristics of the web-interface Currently, many cartographer-oriented internet sites are capable of showing clickable maps with interrogable points and/or areas. They mostly give results of queries performed by clicking, in the form of static data contained in a digital geographic archive. A further step for the diffusion of technical information is to allow the user to get map-data from customized processing, for instance allowing a certain value to be inserted in predefined HTML forms. Custom made maps will be queryable, and clicking will cause new processing or will extract data from a database. In other words, this kind of dynamic mapgeneration procedure allows users to interact with the map closely: making choices in the customising process, allows every user (decision-maker, as well as technicians or the general public) to take an active part in the communication of knowledge. The flexibility and power of GRASS-GIS is well-known to the scientific community. Starting from this assumption, it has been possible to try to investigate the possibility of

374

E.M. Castrogiovanni et al. / Atmospheric Research 77 (2005) 367377

developing a web-interface for an on-line modelling system, using GRASS as the core. To limit the bandwidth needed for the transaction of packets (in requestrespond phases) and to minimise the engagement of client processing resources, we use a typical server-side approach. No applet has to run, and no plug-in is required: the user is able to start calculation procedures and to visualize results using only a simple web-browser. In this configuration, requests are sent from client to the web-server, which can access both the GIS-data and the software needed to process calculations and requests. Once solutions are found, results return to the web-server, and then to clients allowing visualisation in the form of maps, tables and graphs. Furthermore, server-side strategies are platformindependent: each computer able to connect to the Internet (or any TCP/IP-based network) and to run a web-browser is capable of obtaining GIS-data (Hess, 2002). The development of our application presumes on one hand a user without GRASS skills and on the other hand the exclusive use of text-based GRASS commands. When a user browses the web-site, after a certain action (such as filling and submitting HTML forms or clicking on a map), that event is interpreted as a specific request of calculation by the PHP engine (loaded as a module in the Apache web-server). At this point, PHP writes a set of GRASS commands in one or more temporary files. Scripts are written using the BASH scripting language, where BASH (Bourne Again SHell) is the standard Linux shell: this is of enormous importance, because it gives great flexibility to the entire system, allowing the full integration of any software running on the server machine. Simultaneously, PHP writes executable files containing instructions for initialising and starting GRASS. Once these two types of file are written, UNIX scripts are executed and GRASS runs. Results of the processing are then formatted by PHP and sent to the web-browser that shows the data to the user (via web-browser) in a way depending on the type of processing and the type of data. A simplified scheme of the server-side approach used in web-GIS development is shown in Fig. 2. The software developed joins map-browsers category (it is not a simple map-generator), but the main features are oriented to calculation rather than visualization abilities (for

Fig. 2. GRASS web mapping by PHP client-server communication.

E.M. Castrogiovanni et al. / Atmospheric Research 77 (2005) 367377

375

Fig. 3. Web GIS/Modeluse schema.

instance the user is not able to zoom to a portion of a region). The structure of the software is shown in Fig. 3 in the form of site-map. All the software used was Free Software: a summary is shown in Table 1. Furthermore, two dedicated software were used for more specific tasks: jp-graph to produce on-line graphical representation and statistics of processing results, and Fluid Turtles (Rigon and Zanotti, 2002), a hydrological computation suite, to integrate GRASS modelling tools.

Table 1 Software used in the development Type of software Operative system GIS-software Geographic/Web DB server Scripting-lang Web-server Web-lang for interface-development Name of the application Debian GNU/Linux 3.0 (Woody) GRASS 5.0 PostgreSQL 7.2.1 BASH Apache 1.3.26 PHP 4.1.2

376

E.M. Castrogiovanni et al. / Atmospheric Research 77 (2005) 367377

5. Conclusions In the recent last years GIS has played an ever-increasing role in the development of both distributed and lumped hydrologic models. GIS capabilities in the management and analysis of a great amount of data and information has given an indispensable role to this type of analysis. In this work, a methodology is presented for GIS-based hydrologic modelling to determine the flood hydrographs for a fixed return period at any point of the Sicilian region chosen by the user. A web-interface has been developed for the same model in order to allow remote users to obtain information about flood discharge in Sicily in a very intuitive mode. Methodology used in the development (free-software platform, standardization of processing routines) allows a great flexibility for expansion and the addition of features in the future with low-cost/high-efficiency technology. The results of this work show that, although the scarce availability of topographic and morphologic data with an appropriate resolution can limit the application of the proposed methodology, this is a powerful tool to represent the spatial variability of watershed features, integrating hydrologic modelling processes by means of use of GIS techniques.

References
Band, L.E., 1986. Topographic partition of watershed with digital elevation model. Water Resour. Res. 22 (1), 15 24. Burrough, P.A., 1986. Principles of geographical information systems for land resources assessment. Monographs on Soil and Resources Survey Clarendon Press, Oxford. Cannarozzo, M., DAsaro, F., Ferro, V., 1990. Analisi regionale dei massimi annuali delle piogge di breve durata per il territorio siciliano. Idrotecnica 6. Cannarozzo, M., DAsaro, F., Ferro, V., 1995. Regional rainfall and flood frequency analysis for Sicily using the two component extreme value distribution. J. Hydrol. Sci. 40 (1), 19 42. Durrans, S.R., Brown, P.A., 2002. Development of an internet-based rainfall atlas for Alabama. Water Sci. Technol. 45 (2). Feder, J., 1988. Fractals. Plenum Press, New York. Hess, S., 2002. GRASS on the web. Proc. of Open Source GRASS-GIS Users Conference. Faculty of Engineering, University of Trento, Italy. Hosking, J.R.M., 1990. L-moments: analysis and estimation of distribution using linear combination of order statistics. J. R. Stat. Soc., Ser. B Stat. Methodol. 52 (2). Jenson, S.K., Domingue, J.O., 1988. Extracting topographic structure from digital elevation data for geographic information system analysis. Photogramm. Eng. Remote Sens. 54 (11), 1593 1600. Keifer, C.J., Chu, H.H., 1957. Synthetic storm pattern for drainage design. J. Hydraul. Div., ASCE HY4 83, 1332 1352. Kottegoda, N.T., Rosso, R., 1998. Statistical, Probability and Reliability Methods for Civil and Environmental Engineers. McGraw-Hill, New York. Matheron, G., 1971. The theory of regionalized variables and its applications. Les Cahiers du centre de morphologie mathematique de Fontainebleu. Ecole Nationale Superieure des Mines de Paris. Noto, L.V., La Loggia, G., 2001. Sistema Informativo Territoriale per le precipitazioni intense in Sicilia. Atti della V Conferenza Nazionale ASITA. Rimini. Rigon, R., Zanotti, F., 2002. The Fluid Turtle Library, Users and Programmers Guide. University of Trento, Italy. Rossi, F., Fiorentino, M., Versace, P., 1984. Two-components extreme value distribution for flood frequency analysis. Water Resour. Res. 20 (7).

E.M. Castrogiovanni et al. / Atmospheric Research 77 (2005) 367377

377

Rosso, R., Burlando, P., 1990. Scale invariance in temporal and spatial rainfall (abstract). Proceedings XV General Assembly European Geophysical Society, Copenhagen. Sherman, L.K., 1932. Stream flow from rainfall by unit-graph method. Eng. News-Rec. 108, 501 505. U.S. Dept. Agric., Soil Conservation Service, 1972. SCS National Engineering Handbook, Sec. 4, Hydrology. World Meteorological Organization, 1983. Guide to hydrological practices: data acquisition and processing, analysis, forecasting and other applications, 4th ed. WMO, vol. 168. Geneve, Switzerland.

Você também pode gostar