The CAESAR model
Tom Coulthard Professor of Physical Geography, University of Hull, UK.
Website Themes by CoffeeCup Software
Copyright 2011 Tom Coulthard, University of Hull, UK.
T.Coulthard@hull.ac.uk
INTRODUCTION, HISTORY AND BACKGROUND OF CAESAR
CAESAR is a two dimensional flow and sediment transport model. It can simulate morphological changes in river catchments or reaches, on a flood by flood basis, over periods up to several thousands of years. It is free do download and use under a GNU licence. It was initially designed to simulate the geomorphic response of river catchments to changes in climate and/or land cover. Its purpose was to address the debate as to whether changes in climate or anthropogenic changes in land cover had led to changes in UK river behaviour over the Holocene. In order to answer these research questions the model had several requirements:

1. To simulate the relevant time scales (10's to 10 000 years - Holocene).
2. Cover suitable spatial scales, e.g. a long reach of river or a catchment
3. To incorporate the relevant processes and parameters that were operating over the above scales (e.g. hydrologicial, fluvial and slope processes)
4. To simulate the whole catchment. There are many feedbacks and interactions between processes operating within a river catchment, and it is important to try and include this within the model (CAESAR now also models river reaches).

This remit was initially part of a PhD project, and since then the model has grown, both in sophistication and application. CAESAR has proved to be very flexible, and to date has been applied to over 20 different catchments and reaches on scales ranging from 500m to 500km2, and over time scales from individual floods to 10 000 years.
One of the main early findings from CAESAR was that the behaviour of river catchments (in this example in the UK) was mainly controlled by shifts in climate, but was modulated (both increased and decreased) by changes in land cover. Coulthard et al., (2000) found that climate and land cover individually could lead to a c.150% increase in sediment yield, but when combined this could rise to 1300%.

This is also shown to the left in a figure from Coulthard and Macklin (2001) where simulations from the River Swale showed that as land cover decreases closer to present day, the size of sediment discharge from similar size 'wetter' periods increases. This work (Coulthard and Macklin, 2001) also allowed some validation of the model results, by comparing the simulated sediment discharge - driven by independent climate and land cover records - to frequency distributions of 14C dated flood deposits in Northern England.

This work also showed that the sediment output from river catchments was far from linear - something which would be a theme investigated by future work. Following on from this work in the Swale, the CAESAR model was altered to create the 'TRACER' model that allowed the tracking on heavy metal contaminated sediment down through the river system - this work was published in Geology (see images below and Coulthard and Macklin, 2003) and showed how contamination patterns were geomorphically controlled, and developed into hot spots of high concentrations.
Though initial applications of CAESAR were modelling river catchments - it became clear that the basic CAESAR model had the capability to simulate erosion and deposition over river reaches. This led to the development of 'reach mode' whereby water could be input at a point within the model (e.g. at the top of a reach) and erosion and deposition on a more detailed DEM of a reach could be simulated. In addition, catchments and reaches can be linked - so output from a coarser scale river catchment model can be fed directly into another higher resolution CAESAR model of a reach.

Technical developments in CAESAR have continued, with the recent addition of lateral erosion - allowing the channel to meander. This is carried out using a novel edge counting algorithm that counts the number of wet and dry cells next to a river bank and then uses this to calculate whether it is on the inside or outside of a bend.

Other developments from 2005 onwards included the addition of suspended sediment, how shear stress is calculated as well as how slope processes are determined. Recent work by Welsh et al., (in press) also relates the rate of slope processes within CAESAR to the hydrological model - so during wetter periods landslips and soil erosion increase.
Recent work (Van De Wiel and Coulthard 2010 - see left) has used CAESAR to explore the non-linear output of sediment from river catchments. The output of sediment from a river cathcment is hard to predict and seems to contain a large amount of noise or scatter. Using CAESAR in various studies we have found that this non-lienarity may be caused by the breaching of bed armour layers.

In most recent work, we suggest that sediment output may be controlled by Self Organised Criticality (SOC) which may mean that sediment delivery is effectively unpredictable, and that we may not be able to infer past environmental records (e.g. climate or land cover) from sedimentary records.
Case Studies and on-going research projects:
CAESAR has been applied to over 100 different reaches and catchments by researchers accross the world. These range from UK rivers, to those in Australia, New Zealand, Spain, Romania, France and Switzerland. It has been applied to whole catchments ranging from 1 to 1000 km2, and reaches up to 40km in length.

CAESAR is presently being used to simulate how changes in bed morphology will influence flood risk in Carlisle, as part of the EPSRC Flood Risk Management Research Consortium (FRMRC II) program. It is also in use as part of a Swiss project SED-RIVER to investigate the role of climate change on alpine catchment morphology and fish habitat. Furthermore, it is being used to establish rates of geomorphic change and the influence of extreme rainfall on rehabilitation plans for the Ranger uranium mine, NT, Australia.
CAESAR-Lisflood

The most recent developments (Spring 2011) bring a 2d hydrodynamic flow model to CAESAR. This is based on the Lisflood-ACC code (Bates et al., 2010) that adds an inertia term to the established lisflood flow model. This aids stability and greatly speeds up the code (especially on smaller grid cells) which means for the first time, there is a flow model that operates fast enough to realistically be used in a landscape evolution/development context.

This flow model was intially added to existing CAESAR erosion and depositon functions - but this afforded a chance to re-write 80% of the original code (some that dated back to 1996) and tidy up many of the operations. CAESAR-lisflood has the same feel and look as CAESAR but operates with a correct hydrodynamic non steady state flow. One further advantage of the lisflood-acc flow model is that it is readily run in parallell. The old scanning CAESAR flow routing algorithm was sequential, preventing this. As a result, CAESAR now makes full use of multi-core windows machines giving significant increases in operational speed. An early example of the flow model in operation can be seen in the video clip below: