Dipartimento di Ingegneria Civile Ambientale ed Aerospaziale DICA

Dipartimento di Ingegneria Civile, Ambientale ed Aerospaziale (DICA), Università di Palermo, 90128 Palermo, ITALY Influence of raingauge network characteristics on hydrological response at catchment scale Domenico CARACCIOLO, Elisa ARNONE, Leonardo Valerio NOTO 4 th International Workshop on Hydrological Extremes AMHY-FRIEND group 15 -17 September 2011, Cosenza, Italy

Precipitation data is one of the most important inputs required in hydrological modeling and forecasting. In an hydrological model, accurate knowledge of precipitation is essential for an acceptable estimation of hydrograph flood The uniformity of precipitation monitoring network, in terms of spatial scale (network density and location of raingauges) raingauges and resolution time, time allows the reproduction, with acceptable accuracy, of the characteristics of the flood phenomenon

Previous studies In this context, over the last thirty years, several studies concerning the influence of rainfall point measurement for the estimation of total runoff volume have been carried out In particular, some studies have been focused on the analysis of the influence of the spatial distribution of raingauges, raingauges others on the influence of the number of raingauges; raingauges however the two issues have never been analyzed simultaneously

Wilson et al. (1979): ü Use 1 or 20 fictitious raingauges to record rainfall concerning to 15 events ü The spatial distribution of rainfall has a strong influence on the runoff. The number of raingauges has an important role for the correct estimation of the hydrograph peak Krajewski et al. (1991): The work is based on the determination of the appropriate raingauges network for the estimation of flood hydrograph, using a physically based distributedparameter hydrologic model The cases considered were: ücase 1: 87 raingauges, temporal interval: 5 minutes ("real") ücase 2: 1 raingauge, temporal interval: 1 hour ücase 3: 5 raingauges, temporal interval: 1 hour ücase 4: 87 raingauges, temporal interval: 1 hour ücase 5: use of the lumped model Higher sensitivity of basin response with respect to the temporal resolution than to the spatial resolution of the rainfall data

Obled et al. (1994) TOPMODEL The use of 21 instead of 5 raingauges is irrelevant to the estimation of the precipitation The small differences that we have in terms of estimation of the precipitation become important when the precipitation is transformed to runoff Goodrich et al. (1995) ü Uncertainty of measuring rainfall due to the number and location of gauges ü Existence of sufficient spatial and temporal variability in rainfall In this paper they show the influence of the different positions of the raingauges for the estimation of the runoff

Purpose The aim of this work is to use a physically based distributed-parameter hydrologic model (t. RIBS) to investigate the influence of the raingauges network configuration in terms of number and spatial distribution, on the estimation of : ü discharge hydrograph ü hydrograph peak ü time-to-peak ü total runoff volume This has been done considering the spatial distribution of soil types in the basin as well

Hydrologic model t. RIBS (TIN Real-Time Integrated Basin Simulator) üDeveloped at MIT (2003) üPhysically based distributed-parameter hydrologic model üRepresentation of the surface with TIN (Triangular Irregular Network) Voronoi Cells Triangle

Catchment The hydrologic model has been applied to the Baron Fork at Eldon watershed, a catchment of Oklahoma (800 km 2)

Experimental part Assumptions : üThe radar measurements, available in the area (NEXRAD), have been assumed as representative of the "real" distribution of precipitation üThe "real" hydrological response of the catchment was considered as obtained from the model t. RIBS using as meteoric input the real precipitation (NEXRAD) The position of 8 raingauges was generated randomly. Precipitation value is set equal to the corresponding NEXRAD raster cell value

9 events of precipitation occurred during 1998 were taken into account. The nine events were chosen according to the average intensity of precipitation (I) classified as high (I> 2. 5 mm/h), medium (1. 5 mm/h <I <2. 5 mm/h) and low (I<1. 5 mm/h) and the coefficient of variation (CV) of average precipitation classified as high (CV> 0. 6), medium (0. 25 <CV <0. 6) and low (CV <0. 25). CV, for each event, is calculated from the raster obtained by adding the hourly precipitation raster (NEXRAD)

The analysis has been carried out assuming five different soil spatial distributions: üa simplified fictitious spatial distribution of soil characteristics: üa single soil type: silty-clay (c) (Ks=1 mm/h) or sandy-clayloam (s) (Ks=235 mm/h) ü two soil types: silty-clay and sandy-clay-loam (cs and sc) üthe real (r) spatial distribution of soil types c s cs sc r

Simulations ü Simulations considering "uniform" precipitation in space and measured by the 8 raingauges (interpolated with the Thiessen polygons). After we have combined the raingauges in pairs, three by three, four by four, five by five, six by six, seven by seven and the complete network ü The hydrographs flood obtained for each combination of raingauges are compared with the "real" hydrological response calculating performance indices

Performance Index statistical correlation index: index RMSE (Root Mean Squared Error) Qi, PLUV : flow obtained with the precipitation misured by raingauges Qi, RAD : flow obtained with the precipitation misured by RADAR N: event hours number RMSE is calculated for each event For each combination of raingauges and for each soil distribution the network of raingauges with the smallest RMSE (RMSEmin) has been chosen

Event 1: 36 houres, CV=medium, I=high Spatial pattern For high intensity, the raingauges are placed in the less permeable soil, but also where the precipitation is high

Event 3: 7 houres, CV=high, I=low Spatial pattern For low intensity, the raingauges are placed in the less permeable soil

Event 4: 67 houres, CV=low, I=medium Spatial pattern For CV=low, varying the distribution of soil, the raingauges network is almost the same

In order to summarize all results in a single table, the average flow was calculated from the flood hydrograph and each value of RMSE is divided for the corresponding average flow. Normalized values obtained for each event, were added together and divided by the number of events in order to calculate the average value of RMSE/QM minimum

ss c cs sc r üUsing only a raingauge, it is placed in the less permeable soil üWith a network of two raingauges follows the same pattern ü With a network of three, four, …. raingauges there is not a clear criterion for the best position of the i-th gauge

… in conclusion. . . ü There is not an optimal raingauges network finalized to the estimation of all the considered flood events. ü The network finalized to the best reconstruction of rainfall field does not coincide with the network finalized to the best flood hydrograph estimation. ü For a fixed event, the best raingauges configuration is strongly dependent on the soil types distribution. ü The best raingauges configurations depend on the precipitation events (in terms of intensity and spatial distribution) and on the soil types distribution (general trend to locate the raingauges where the soil is less permeable): § in case of high average rainfall intensity, the influence of precipitation pattern is greater than that of soil types distribution; § in case of medium or low average rainfall intensity, intensity the effect of precipitation is lower than the effect of soil types distribution; § if the rainfall spatial variation is medium or low the distribution of raingauges varies little with the change of the distribution of soils.

Thank you for your attention