CREEP  - Changing Relief and Evolving Ecosystem Project
Nan Rosenbloom
Terrestrial Sciences Section
Climate and Global Dynamics Division
National Center of Atmospheric Research
Boulder, CO  80305

David Schimel (NCAR)
Jennifer Harden (USGS-Menlo Park)
Jason Neff (Univ of Colorado at Boulder)

The CREEP model is a theoretical, process-response model that links ecosystem dynamics with geomorphology at the landscape scale.  CREEP  focuses on the interaction of soil aggregation and organic matter with downslope particle size distribution (percent sand, silt, and clay) and the formation of catena sequences, defined by Milne [1935] as a lateral distribution of chemical, physical, and biological soil properties along a hillslope.  The CREEP model simulates edaphic controls on ecosystem dynamics by simulating the downslope movement of soil and nutrients while monitoring the effect on soil carbon accumulation.


The CREEP model simulates the sediment transport processes that fractionate soil particles by texture into a soil catena.  A soil catena is the cumulative result of natural downslope interactions between pedogenic and geomorphic processes, including drainage conditions, differential transport of eroded material, and leaching, translocation, and redeposition of mobile chemical constituents. Within a classically defined catena, finer particles, particularly clays, are winnowed from the upper slopes and redeposited at the foot and toe slopes by hillslope transport processes (e.g., Riecken and Poetsch, 1960).  Thus a catena implies both a lateral soil sequence and the geomorphological and pedological processes instrumental to its formation.

We developed the CREEP model to track the evolution of surface transport and the subsequent evolution of a soil profile.  The model is diffusion-based; as such it ignores event driven transport by specific rainfall events.  Rather, the decadal time step of the model presumes a time-averaging of surface transport that, in gentle, vegetated landscapes can be assumed to be roughly proportional to slope.

Sediment Transport

Sediment transport is simulated numerically as texturally dependent diffusive fluxes.  Mass transport is parameterized by a linear diffusion model analog, transporting mass downslope in response to gradients in the local surface topography.  However, CREEP differs from previous hillslope models by partitioning mass transport into texturally dependent diffusive fluxes where each textural component (sand/silt/clay) diffuses independently, responding to gradients in the local surface topography with a texturally dependent rate.  To accomodate this textural dependence we take advantafge of the finite difference format found in the alternating direction implicit (ADI) numerical scheme but write separate differential equations for each texture component.

Isotopic Tracers

We introduced a stable isotope (10Be) to trace the movement of soil layers.  Using a conservative isotopic tracer allows us to observe the motion of surface layers, and contrast this movement with the signature of soil carbon, which has a more complex and transient soil residence history.   10Be is introduced into the system through dust deposition.  Once deposited, it is assigned to the clay fraction, where it is assumed to advect downslope by adhering to clay particles.  Model-data comparisons require the addition of an independent downhill transport mechanism of 10Be, in addition to that assigned to the mineral fraction of eolian dust.


The CREEP model is written in C++, running on a Dell PC under Cygwin, a Linux-like environment for Windows..  I compile and run CREEP with gcc3.2.2.  The current model uses the Blitz++ version 0.6 Library.  Blitz++ is a C++ class library for scienfic computing which provides performance on par with Fortran 77/90 by using template techniques to achieve high performance.  Blitz++ provides dense arrays and vectors, random number generators, and small vectors.  Blitz++ is now available being served through sourceforge:
Because of  the embedded Blitz++ array structures, CREEP is not currently configured to run under UNIX or Linux.

Algorithm Development:

Grid Configuration:
CREEP runs on a uniformly grid, with two lateral spatial dimensions (rows and columns) and one vertical dimension representing depth within the soil column.

Dust Deposition:

Dust is deposited uniformly over the grid as a 'blanket' at each timestep.  The deposition rate is controlled by the user, and may vary through time.

10Be Deposition:

 The delivery of 10Be with the dust is also variable, but assumed to be nearly constant at 2.2e8 atoms10Be / gDust.  


The landscape is classified as an ‘open’ system which is continually evolving and contains no unique or finite solution.  As such, the model output is completely dependent on initial conditions, which are truly unknowable from the present landscape.  

Soil Texture Redistribution:
Running CREEP as a forward model requires assigning initial conditions that are, in fact, unknowable, including soil texture, landscape form, and climate conditions.  We look to the data to make educated guesses about past history, looking deep into the soil layers of the relatively stable ridgetops, where presumably we have the best chance of finding "parent" material.  We use the basal layer soil texture distributions from the ridge profiles to uniformly  initialize the landscape, assigning these sand/silt/clay percentages to all soil columns and layers.  As the model runs forward in time, the range of assigned diffusivities for each soil texture produces differential particle movement and a downslope fractionation of the soil partitions.

We assign soil texture to be that of observations of the basal layer at the ridge top, and let the model differentiate the soil texture through time.

We then ‘rejuvenate’ the model by increasing the distance between the upper and lower slopes.  We then run the model forward in time, comparing the mean 2D curvature of the model surface.  The two surfaces approach each other at between 3500 – 4500 years model run time.  We chose 4000 years and a representative model run time.

Topographic rejuvenatio
Soil Carbon Dynamics/ Soil Decomposition (Jason Neff Algorithm)

Soil carbon dynamics are now simulated by CENTURY-4.0 based parameter estimates of total carbon production.  These parameterizations will eventually be replaced by explicit estimation of multiple carbon pools by the Neff soil carbon model.  The Neff soil decomposition algorithm implements a system of differential equations, associated utilities and an ODE solver to simulate the changes in carbon content and isotopic composition in various forms (pools) in soil.  The soil structure evolved from two primary sources - Tom Hilinski's soil base class used in the implementation of erosion in the Century 5 ecosystem model, and Jason Neff's differential equations which quantify changes in content and composition of soil pools.


Riecken, F.F., and E. Poetsch (1960) Genesis and classification considerations of some prairie-formed soil profiles from local alluvium in Adair County, Iowa.  Iowa Academy of Sciences. 67. p. 268-276.

Rosenbloom, N.A., S. C. Doney, and D. S. Schimel (2001) Geomorphic evolution of soil texture and organic matter in eroding landscapes. Global Biogeochemical Cycles. vol. 15, no. 2, p. 365-381.

Rosenbloom, N.A., S. C. Doney and D. S. Schimel (1998) Hillslope mass transport, catenary sequences, and soil organic matter: Numerical simulations and model-data comparisons of the CREEP model for Great Plains grassland environments. Presented at the 1998 Fall Meeting of the American Geophysical Union. San Francisco, CA. December 1998. EOS Abstract. 79(45):F264.

Rosenbloom, N.A., R. McKeown and D. S. Schimel (1996) Predicting the interaction between long-term hillslope mass transport and vegetation dynamics. Presented at the 1996 Fall Meeting of the American Geophysical Union. San Francisco, CA. December 1996. EOS Abstract. 77(46):F253.

Rosenbloom, N.A. (1996) A model to predict the interactions between long-term hillslope mass transport and vegetation dynamics. Presented at the 1996 Fall Meeting of the Geological Society of America. Denver, CO. October 1996.

Rosenbloom, N.A., and R. S. Anderson (1994) Hillslope and channel evolution in a marine terraced landscape, Santa Cruz, California.  Journal of Geophysical Research.  99. p. 14,013-14,029.