OR/14/011 Calibration: Difference between revisions

From MediaWiki
Jump to navigation Jump to search
 
(One intermediate revision by the same user not shown)
Line 42: Line 42:


==Debris flow==
==Debris flow==
The debris flow component may be calibrated against observed data, using an iterative technique similar in principal to the hydrological calibration. There are two main parts to the calibration of debris flows; matching the physical characteristics of previous flows (for example, lateral runoff length and total flow area/volume), and matching the triggering of such events in time with respect to the driving hydrology. Both require the parameter values that drive the simulation to be updated, improving the match between model and observation after each iteration. The triggering characteristics of debris flow are controlled by the friction angle parameter under the slope processes tab (Section 3.12.3). The remaining SCIDDICA parameters determine the flow characteristics. D’Ambrosio et al. (2003) provide an extended description of the component calibration.
The debris flow component may be calibrated against observed data, using an iterative technique similar in principal to the hydrological calibration. There are two main parts to the calibration of debris flows; matching the physical characteristics of previous flows (for example, lateral runoff length and total flow area/volume), and matching the triggering of such events in time with respect to the driving hydrology. Both require the parameter values that drive the simulation to be updated, improving the match between model and observation after each iteration. The triggering characteristics of debris flow are controlled by the friction angle parameter under the slope processes tab ([[OR/14/011 CLiDE Pre-processing#SCIDDICA|SCIDDICA]]). The remaining SCIDDICA parameters determine the flow characteristics. D’Ambrosio et al. (2003)<ref name="D’Ambrosio 2003">D’Ambrosio, D, Di Gregorio, S, and Iovine, G. 2003 Simulating debris flows through a hexagonal cellular automata model: SCIDDICA S3−hex. Nat. Haz. Earth Syst. Sci., 3, 545–559.</ref> provide an extended description of the component calibration.
==References==
<References/>
[[Category:OR/14/011 CLiDE version 1.0 user guide | 05]]
[[Category:OR/14/011 CLiDE version 1.0 user guide | 05]]

Latest revision as of 12:15, 22 April 2022

Barkwith, A* and Coulthard, T J**. 2011. CLiDE version 1.0 user guide. British Geological Survey Open Report, OR/14/011.

* British Geological Survey, Environmental Science Centre, Keyworth, Nottingham, NG12 5GG, UK
** Department of Geography, Environment and Earth Science, University of Hull, Cottingham Road, Hull, HU6 7RX, UK

Introduction

Calibration (platform validation) is an important step in the modelling process where simulations are to represent, or make predictions for, a ‘real’ environment. Within the CLiDE platform there are currently three main processes that can be calibrated to observed data; the hydrology, fluvial sediment transport and debris flow. For each of these system components the calibration process relies on modifying the input or initialisation data to fit the simulated output to observed data.

Hydrological

Calibrating the CLiDE platform hydrology consists of two distinct steps; the calibration of the surface runoff followed by the calibration of the groundwater module. Both of these steps may require a number of iterations before an acceptable match between observed and simulated data is found. Within the UK, gauged river data over a number of years is easily obtainable for major catchments.

Surface runoff

For this step in the calibration we are comparing the simulated surface runoff components to observed river gauging data. We therefore need to separate the runoff and baseflow components for both datasets before the comparison is made. This allows the separate influences on the different flow paths to be assessed and modified. The longer the calibration, the greater the certainty of the validation, both in terms of large scale climate trends and the response of the platform to short, intense events such as stroms. Uncertainty may also be reduced by calibrating the platform to multiple, spatially distributed locations.

Preparing the gauged flow data

A baseflow separation is required for the gauged river flow dataset. There are several methods for the separation of baseflow from gauged river data (Eckhardt, 2008[1]). The UKIH low flow method (Piggott et al., 2009[2]) is one such technique that can be applied to daily data and is described in the following paragraph.

The UKIH method is based on periods of flow that are assumed to be composed entirely of baseflow, which occur within a time series of gauged river data. To identify these periods, the gauged data is divided into 5 day segments and the minimum values within each segment are selected. The low point for each segment, yi, is compared to neighbouring (temporally) segments, and turning points along the sequence identified as low flows that satisfy the equation 5.1. The daily baseflow contribution to a river is estimated by linearly interpolating between these turning points.

To create a baseflow separated time series of river flow, whereby only surface runoff contribution is included, the baseflow value at each point must be removed from the total gauged flow.

Simulation

Calibrating the simulated surface runoff requires the groundwater module to be disabled (Groundwater), so that no baseflow is returned into river flows. For the calibration period, it is important that the platform has reached a quasi-steady (balanced) state, whereby soil moisture and surface water stores are no longer trying to reach equilibrium from their initialised state. This may be checked by monitoring the SMD and NSSS output from the water balance. Upon equilibrium both of these stores should show an annual cycle of variation, but no significant trend inter-annually.

Following confirmation of steady state, the platform is ready to undertake a calibration run. River flow at the location corresponding to that of the flow gauging station must be output during the simulation. With baseflow return disabled the simulated river flow is directly comparable to the baseflow separated river gauge data. If the two datasets do not match within acceptable limits, the initialisation and input parameters for the simulation must be re-assessed and the calibration repeated.

It is suggested that the following parameters are modified in order to attain a better match between simulated and gauged data:

SLiM surface water partitioning (Hydrology tab)

  • Modifying the rooting depth imparts a change in water take up be the vegetation. Seasonal changes to this value represent annual changes in the growth cycle of vegetation. Increasing these values reduces the available water in the soil store and increases evapotranspiration rate. Again the response will not be related linearly to the change in forcing and is catchment specific.
  • The rain interception coefficient represents a loss of rainfall (as a percentage) to unknown sinks.

Lisflood surface water routing (Flow Model tab)

Groundwater

Once the surface water has been calibrated, the groundwater component may be calibrated to either; total gauged river flow (i.e., without the baseflow separation) or to groundwater head levels. The procedure for calibrating to total flow is similar to the surface water calibration. The key differences are, that the groundwater component is enabled during simulation and that the hydraulic conductivity and specific yield of the groundwater component are modified instead of the surface water components. If calibrating to groundwater head levels, the groundwater level output from the simulation must be saved for the required time period (Gridded). As with the surface water component, it is important the simulated system is in steady state for the calibration period. In the case of groundwater, this involves monitoring the groundwater head levels in the water balance to ensure they have reached an equilibrium condition that shows seasonal changes, but little inter-annual change.

When calibrating the groundwater module it is advised that only the multiplier values in the Hydrology tab are modified and not the distributed values. This avoids having to modify all of the values in the distributed groundwater input files after each iterative step of the calibration process.

  • Increasing the hydraulic conductivity allows water to flow at a faster rate through the subsurface, creating a less steep gradient below high and low heads and reducing the time lag from recharge to baseflow return.
  • The specific yield determines the groundwater head response to an incoming or outgoing flux of water. A groundwater system with a high specific yield will show a smaller response to the influx or removal of water in comparison to a catchment with a low specific yield.

Debris flow

The debris flow component may be calibrated against observed data, using an iterative technique similar in principal to the hydrological calibration. There are two main parts to the calibration of debris flows; matching the physical characteristics of previous flows (for example, lateral runoff length and total flow area/volume), and matching the triggering of such events in time with respect to the driving hydrology. Both require the parameter values that drive the simulation to be updated, improving the match between model and observation after each iteration. The triggering characteristics of debris flow are controlled by the friction angle parameter under the slope processes tab (SCIDDICA). The remaining SCIDDICA parameters determine the flow characteristics. D’Ambrosio et al. (2003)[3] provide an extended description of the component calibration.

References

  1. Eckhardt, K. 2008. A comparison of baseflow indices, which were calculated with seven different baseflow separation methods. J. Hydrol., 352, 168–173.
  2. Piggott, A R, Moin, S, and Southam, C. 2009. A revised approach to the UKIH method for the calculation of baseflow. Hydrological Sciences Journal, 50(5), 911–920.
  3. D’Ambrosio, D, Di Gregorio, S, and Iovine, G. 2003 Simulating debris flows through a hexagonal cellular automata model: SCIDDICA S3−hex. Nat. Haz. Earth Syst. Sci., 3, 545–559.