Benchmark of hydrological models applied to the Var

Benchmark of hydrological models applied to the Var catchment: quality and availability of results for different hydrological situations and different models Team 8

Structure Introduction • The Var catchment and 1994 flood • What is benchmarking? • Efficiency criteria • Model types and selected models Results • NAM • HEC-HMS • Mike. SHE • Comparison of model results Conclusion 2

The Var Catchment and 1994 flood The Var Catchment: • Area of 2, 822 km², 114 km long. • Source at an altitude of 1800 m - catchment is highly sloping, resulting in fast runoff responses to rainfall. • Main river system is constituted of four large valleys in addition to the Var River. November 1994 Flood ● Peak flow: approx. 3770 m³/s at Pont Napoleon III (However, as the flow gauge stopped working at 800 m³/s this is uncertain). ● 350 mm of rainfall in 72 hours. Most fell over highland areas. ● Soils were already saturated due to 600 mm of precipitation in the previous two months. 3

What is benchmarking? • • Benchmarking is a standard or a set of standards used as a point of reference for evaluating model performance. In the context of hydrological models it refers to the assessment of a model’s ability to simulate • an observed event. Hydrological model performance can be assessed using a range of efficiency criteria. 4

Efficiency criteria Nash-Sutcliffe efficiency: is used to describe the accuracy of a model output compared to observed data. NSE values range between minus infinity to one. 1 corresponds to a perfect match between modelled and observed results. It is calculated as follows: R² (Coefficient of determination): is a number that indicates the proportion of variance in the dependant variable that is predictable from the independent variables. An R² value of 1 indicates that the regression predictions perfectly fit the data. It is calculated as follows: 5

Model types 6

Selected Models Three models of differing complexity were selected NAM Lumped conceptual rainfall-runoff model HEC-HMS Semi-distributed conceptual model Mike. SHE Physical distributed model 7

1. NAM (Nedbør-Afstrømnings Model) Model type: Lumped, conceptual model based on considerations of the physical processes. It takes into account the laws of conservation in the governing equations. Inputs: • Catchment area • Model parameters • Initial conditions • Meteorological data (rainfall and PET) • Streamflow data (observed runoff at outlet) for calibration and validation Outputs: Single hydrograph at the outlet 8

Advantages and disadvantages of NAM Advantages : • Simple to use and understand • Requires little data input • Easy and quick to calibrate • Very short simulation time Limitations : • Very limited output data • No representation for topographic influence on flood 9

NAM Results NSE 0. 96 R² 0. 97 ΔV 9% ΔTPeak 10 min ΔQPeak 15 % 10

2. HEC-HMS Model type: Conceptual semi-distributed Model, It Includes semi empirical equations with a physical basis. The physical process of the Catchment represented by a number of interconnected reservaiors. Inputs: • Basin model - Main parameter for loss - SCS curve number - Transform Method - SCS unit hydrograph - Base flow method- constant monthly • Meteorological model (Thiessen polygone) - Hourly time-series of rainfall from 6 gauges (gauge weights) • Observed discharge data at Napoleon bridge for calibration Outputs: Hydrographs at each sub-catchment and at the outlet 11

Advantages and disadvantages of HEC-HMS Advantages : • Due to its graphical user interface the program is easy to apply and results are easy to visualize • HEC-Geo. HMS can process geospatial data in Arc. GIS • Compatible with other HEC-programs • Short simulation time • Open source Limitations : • • Requires a large amount of hydrological and meteorological data Filed investigation needed in order to get best representation of the model. 12

HEC-HMS Results NSE 0. 82 R² 0. 97 ΔV 15 % ΔTPeak 1 h ΔQPeak 1% 13

3. Mike. SHE Model type: Physically based, distributed model Overland flow is calculated using the diffusive wave approximation of Saint-Venant equations. Ignores momentum. Inputs: • Model domain grid • Spatial and temporal precipitation for 6 different gauging stations • Topography (75 m resolution DEM) • Manning’s ‘n’ (Strickler coefficient) Outputs: Possibility of creating a hydrograph for each grid square, not just the outlet. 14

Advantages and disadvantages of Mike. SHE Advantages • Spatially distributed (i. e. internal catchment response) • The impact of land use change event can be predicted • Less dependent on existing records Limitations: • Complex model with long run times • Evaluation of many parameters necessary • Many uncertainties in the result of the model • The spatial scale can affect the predictions 15

Mike. SHE Results NSE 0. 48 R² 0. 69 ΔV 23 % ΔTPeak 1. 5 h ΔQPeak 1% 16

Comparison of model results Efficiency Criterias Mike. SHE NAM HEC-HMS NSE 0. 48 0. 96 0. 82 R 2 0. 69 0. 97 ΔVolume 23 % 9% 15% ΔTPeak 1. 5 h 10 min 1 h ΔQPeak 23 % 15 % 1% 17

3. Conclusion • In our case, HEC-HMS, NAM perform best according to different benchmark criteria • Multiple efficency criteria should be taken into account when benchmarking a model • Decision for a certain model should not only be based on benchmarking results • Benchmarking of hydrological models is an important tool to quantify model performance and should be an integral part of the modeling process 18

Thank you for listening Questions? 19

References • • • Agrawal, N. and Desmukh, T. (2016). Rainfall Runoff Modeling using MIKE 11 Nam –A Review. International Journal of Innovative Science, Engineering & Technology , [online] 3(6), pp. 659 -667. Available at: http: //ijiset. com/vol 3/v 3 s 6/IJISET_V 3_I 6_87. pdf [Accessed 14 Feb. 2019]. Gourbesville, P. (2009). Le bassin versant du Var et la crue de 1994. Fleuves, territoires et infrastructures Regards croisés sur la Plaine du Var , pp. 1 -15. Krause, Peter, D. P. Boyle, and Frank Bäse. "Comparison of different efficiency criteria for hydrological model assessment. " Advances in geosciences 5 (2005): 89 -97. Le Portail Numérique des Savoirs des Alpes Maritimes (2019). La Crue du Var du 5 novembre 1994. Nice: Département des Alpes Maritimes, pp. 1 -5. Ministère de l'écologie, de l'énergie du développement durable et de la mer (1995). Les Crues du 5 au 7 Novembre 1994 en Provence Alpes Côte d'Azur. Nice: Direction Regional de l'Environnement, pp. 1 -41. EWEN, John. Hydrograph matching method for measuring model performance. Journal of Hydrology , 2011, 408. Jg. , Nr. 1 -2, S. 178 -187. 20