Forum Replies Created
-
Posts
-
Hi crosbya1,
The approach you show above should work just fine.
With respect to your question about what happens when probabilities exceed 1, in your case case lets say that for the second table for transition pathways you provided values that add up to 1.15 instead of 1.0:
| Class | To Class | Transition Type | Probability |
| coniferous:all | coniferous:all | Fire[type] | 0.9000 |
| coniferous:all | deciduous:all | Fire[type] | 0.2500 |Instead of a realized transition probability for Fire[type] in coniferous:all of 0.0050 you would see a realized probability of 0.0050 * 1.15 = 0.00575
Hope that helps!
LeonardoThanks Miles, If you apply the checkbox to “[] Calculate from distribution” and specify an integer number of cells for each record, there will be no variation at time 0.
One additional note is that under non-spatial initial conditions, if you select “[] Calculate from distribution” and specify the “Relative Amount” values in the Non Spatial Initial Condition “Distribution” datasheet as the number of cells to be used for each record, then the landscape is initialized identically for each Monte Carlo realization. See: https://docs.stsim.net/reference/prop_initial_conditions_non_spatial.html
Hi Ayn
I have seen this kind of behaviour in a model once before where the were very different transition probabilities for timber harvest across state classes. The way attribute targets work is by multiplying base probabilities. If some state classes have much higher probabilities than others, the resulting probabilities may exceed one.in that case the model will not achieve the attribute target. Can you review the transition probabilities to see if that is the case?
Hi Ayn, I am aware of folks at USFS Regions 1, 6, and 3 using FVS to “calibrate” their STSMs. A key aspect of this process is to include the state resolution and classification in order to relate state changes from FVS to ST-Sim. These folks have developed a state class classification based on variables such as diameter, number of canopy stories, and percent canopy cover that can be related back to transitions in ST-Sim when certain thresholds are crossed for these variables. I think “calibration” is a more appropriate word for this rather than “validation”.
Similarly others have used retrospective analyses over time periods with known landscape conditions to “calibrate” transition rates in their models such that the projected landscape over the historic time period matches observed changes on the real landscape, for example Frid et al. 2013.
Oreskes et al. 1994 provide a thoughtful perspective on whether it is ever really possible to validate numerical models.
Hi Kori,
I am curious as to why you want to sample from a Weibull distribution for transition probabilities? A Weibull is typically used to model time to failure, with an example of this being fire return intervals, not probabilities. Outputs of single cell ST-Sim runs using fixed probabilities that are either constant or vary with age will produce different shapes of a Weibull distribution for the times between transition events on that cell. A constant probability that does not vary with age will produce an exponential distribution for times between transition events. An exponential distribution is just a special case of the Weibull.
Built in distribution support in SyncroSim currently includes (Beta, Gamma, Normal, Uniform and Uniform Integer), but not Weibull. This is something we could consider adding.
In the meantime, if you still want to sample from a Weibull, you could do your own sampling outside of SyncroSim to generate a datasheet for multipliers where the following fields are specified:
1. Iteration [only required if you want every iteration to be different]
2. Timestep
3. Transition Group [the transition group that you want to vary probabilities for]
4. Multiplier [sampled from your Weibull distribution]To sample from a Weibull you could use R:
https://search.r-project.org/CRAN/refmans/distributions3/html/random.Weibull.htmlCheers,
LeonardoHi Miles,
It has been a long time since we supported VDDT legacy models. However, I was able to find some old materials on making the transition:
Forum posts on the move from VDDT to ST-Sim: https://syncrosim.com/forums/topic/importing-vddt-projects-to-st-sim/
Path 4.2.2: https://syncrosim.com/download-id/path-4-2-2/?tmstv=1682118322
Tutorial on moving from Path to ST-Sim: https://syncrosim.com/download-id/path-to-st-sim-tutorial/?tmstv=1682118521A caveat is that we have not tested these tools and tutorials for almost a decade. You might be better off writing your own scripts using R and rsyncrosim to first extract the data you need from the MS Access database and then save it to a new ST-Sim library.
Regards,
LeonardoIf I understand your question correctly, your flow multipliers are now stochastic and are sampled from a distribution in SyncroSim. You would like to know the sampled rate values and pull them into your rsyncrosim external program. As I understand it, these values vary across years so your sampling frequency should be “Iteration and Timestep”.
stsimsf_OutputFlow will get you the total flow “Amount” broken down by source state class, strata, etc… Using this datasheet, you would then need to back calculate the rate taking into account total area. For example if a particular row has a value of x that represents a flow amount for that combination of strata, your rate would be x divided by the total area on the landscape associated with that combination of strata.
Note that you could also get the rate associated with each individual cell on your landscape by getting the raster you need from stsimsf_OutputSpatialFlowGroup. These values are already rates specified as flow amount per unit area.
One final note: if you prefer to continue using State Attributes to specify your rates, this datasheet also allows for specifying distributions rather than single values.
Hope that helps!
Just a follow-up: when expanding the details of the message there was an additional note “The given key was not present in the dictionary.” My apologies that this is not the most informative error. In the case of your library it appears that there are scenarios referencing a project that does not exist in the library. Somehow it looks like the project got deleted but the scenarios did not. Is this a library that was accessible on a network via multiple users? One possible use case leading to this problem is multiple users making edits to the library at the same time.
Hi Kevin, Please send us the library and let us know what versions of SyncroSim and packages you are using.
The steps to run an External Program using a Python script are similar to the above example, with a few minor changes outlined below.
The example was created using the following software and packages:
- SyncroSim version 2.4.10
- ST-Sim version 3.3.9
- Python version 3.10.6
- Python packages (Note these need to be pre-installed using conda or pip):
- pysyncrosim version 1.0.19
- pandas version 1.5.1
- numpy version 1.23.4
- rasterio version 1.3.3
You can use the same ST-Sim Library as the above example. However, you will also need to download the external program Python script example here.
Update the External Program Datasheet within the Scenario Properties to include your Python executable file and Python script.
Follow the remaining instructions above to retrieve the desired outputs.
Hi David,
There is a Beta version (0.3.4) now available. You can find more info, and a Getting Started tutorial here:
https://burnp3.github.io/BurnP3Plus/getting_started.htmlCheers,
LeonardoThanks for reaching out David. We are planning for a Beta release by October 28th, 2022. Will let you know when it is ready.
Hi Anthony,
Sarah is correct, spread distance is accumulated over time for each possible source and destination cell combination. ST-Sim uses the age of the source cell as a proxy for how much time has elapsed and hence how many draws to make from the spread distribution for each source/destination vector. For this reason, the initial age of the source cells is important. In your simulations you will see spread happening soonest to neighbors of the oldest source cells. Spread distance is calculated from center to center. When this distance threshold is exceeded, the destination cell will transition.
Regards,
Leonardo
