How to accelerate century succession running speed

[myidream]

Hi, Is there any method to accelerate the running speed of Century-succession? I ran this extension to simulate a 400 million ha forest landscape for 300 years with 10 year time step. I estimated that it will take about 3 months. So, I want to know is there any way to accelerate the running speed.

[Melissa Lucash]

Hello,

I'm running a landscape with an 8 million ha landscape with 2.2 million active cells at 1ha resolution (2728 columns X 2100 rows) with Century.  Given my runs, I estimate that it would take 9 days to run 300y with 10y time steps on my computer (2400 CPU, 16GB) without any other extensions running and not enabling the maps of ANPP/NEE in the Century input file.  It seems like your landscape is running slower than I would predict, though I don't have any specific suggestions for improving the speed of Century (other than the obvious increase in RAM/CPU speed).

The new soon-to-be released version of Century runs much faster for me than the one currently available on the website.  It should be widely available soon (~1-2 months).

Best,
Melissa

-- 
Dr. Melissa S. Lucash
Research Faculty
Department of Environmental Science and Management
Science Teaching and Research Center, B1-24A
Portland State University
Portland, OR 97201
luc...@pdx.edu
503-725-3894
https://sites.google.com/a/pdx.edu/dynamic-ecosystems-landscape-lab/people/dr-melissa-lucash

On 6/24/2014 10:46 PM, myidream wrote:

Hi, Is there any method to accelerate the running speed of Century-succession? I ran this extension to simulate a 400 million ha forest landscape for 300 years with 10 year time step. I estimated that it will take about 3 months. So, I want to know is there any way to accelerate the running speed.

--
--
You received this message because you are subscribed to the Google
Groups "LANDIS-II Users" group.
 
To see the discussion, visit this group at
http://groups.google.com/group/landis-ii-users
 
To unsubscribe from this group, send email to
landis-ii-use...@googlegroups.com
---
You received this message because you are subscribed to the Google Groups "LANDIS-II Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to landis-ii-use...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/landis-ii-users/66c3baad-9f13-4346-8f2f-fcba4d8995fd%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Robert Scheller]

Myidream,

This is a general problem dating back to the origins of LANDIS.  It is always possible to push the model beyond its intended capacity.  And simulating 400 million ha for 300 years using Century is well beyond the reach of today's common computing platforms.

A few suggestions:

* Choose a smaller area.  Or choose a model without spatial interactions.  Are spatial interactions necessary and useful across 400 M ha?  How important is seed dispersal, in particular, to your answering your questions?/

* Choose a larger resolution.  You didn't mention your resolution but the CPU cost of seed dispersal increases exponentially as cell size decreases linearly (it is a Big-O-squared type algorithm).

* Use an extension that carries less information.  Century Succession uses the MOST memory of any extension as it is tracking many C and N pools.  If tracking N is not critical, there are other C-centric succession to choose from, including 'Forest Carbon' (available now) and 'PnET' (available soon).

Cheers,

Rob