A Framework for the Quantitative Testing of Landform Evolution Models

  1. Peter R. Wilcock and
  2. Richard M. Iverson
  1. Garry R. Willgoose,
  2. Gregory R. Hancock and
  3. George Kuczera

Published Online: 29 MAR 2013

DOI: 10.1029/135GM14

Prediction in Geomorphology

Prediction in Geomorphology

How to Cite

Willgoose, G. R., Hancock, G. R. and Kuczera, G. (2003) A Framework for the Quantitative Testing of Landform Evolution Models, in Prediction in Geomorphology (eds P. R. Wilcock and R. M. Iverson), American Geophysical Union, Washington, D. C.. doi: 10.1029/135GM14

Author Information

  1. Centre of Environmental Dynamics, The University of Newcastle, Callaghan, Australia

Publication History

  1. Published Online: 29 MAR 2013
  2. Published Print: 1 JAN 2003

ISBN Information

Print ISBN: 9780875909936

Online ISBN: 9781118668559



  • Geomorphology—Mathematical models


Recent years have seen the development of a number of physically based computer models simulating the evolution of landforms under the action of erosion. To date, comparisons with field data have been largely qualitative so that only subjective assessments of their adequacy have been performed. To be able to rely on the quantitative predictions of these models methodologies for assessing their quantitative reliability are needed. Key problems are (a) the lack of repeatability of field measurements, (b) the sensitive dependence of models on initial conditions combined with the inherent unknowability of initial conditions in the field, (c) identification of measures for assessing model adequacy that can distinguish differences arising out of the physics, from random effects and unknown inputs, and (d) development of an objective, statistical methodology that can reject an inadequate model. This paper addresses these problems by proposing a statistical methodology based on Monte-Carlo simulation using the landform evolution model being tested. The principles are presented in a series of examples that compare the SIBERIA catchment evolution model with a natural undisturbed site in Arnhem Land, Northern Territory, Australia. One model calibration, based on an eye fit of field data, is shown to be deficient. An improved method for determining drainage density, based on fitting to the slope of the cumulative area diagram, is proposed. While the examples presented are by no means comprehensive, it is concluded that SIBERIA model does an adequate job of simulating the field landform. A formal probabilistic framework for model testing is developed, together with a methodology for objectively assessing the value of data for model testing. This methodology allows for model and input uncertainty, and correlation of statistics.