Do spatial simulations change reference conditions?
Data files
Apr 20, 2026 version files 965.84 MB
-
Descriptions_of_Ecological_Systems_SouthernSnakeRange_09-27-2023.pdf
889.83 KB
-
GIS_Rasters.zip
42.87 MB
-
README.md
12.32 KB
-
Results_NRV.zip
2.22 MB
-
Results-8_Digit_Raster_500years.zip
209.78 MB
-
South_Snake_Range_NRV.zip
710.07 MB
Abstract
The reference condition is the expected proportions of reference vegetation classes per ecological system (hereafter, system) within an area and has been measured with non-spatial state-and-transition simulation models (STSM). We aimed to Demonstrate that reference conditions varied between non-spatial and spatial STSMs from an eastern Nevada USA landscape. The same STSM database was used to simulate the non-spatial and spatial reference conditions, except additional input were added for spatial simulations. The reference condition was obtained from averaged vegetation class proportions of replicated simulations. Mean fire return intervals (MFRI) were always shorter in spatial than non-spatial simulations. The dissimilarity between the spatial and non-spatial reference conditions was generally higher for smaller systems and those with long MFRIs (≥ 100 years). Non-spatial and spatial reference conditions were comparable for large systems with 50-100-year MFRI. The proportions of older vegetation classes for spatial systems were smaller for large systems with long MFRIs compared to non-spatial results, but those proportions were larger for smaller systems with shorter MFRIs likely because fires can spatially exit small narrow systems. National and local assessments that use departure from non-spatial reference conditions to determine vegetation treatments should recognize that spatial simulations can lead to revisions of treatments decisions.
There are one simulation database, one supporting simulation raster folder, one folder of tabular results of vegetation classes by systems and analyses, one folder of raster results of vegetation classes by systems for three years, and one PDF of the vegetation description:
(A) "South_Snake_Range_NRV.zip" zipped folder contains the simulation database "SouthernSnakeRange_Spatial-NRV.ssim," which requires the freeware Syncrosim (Syncrosim version 3.1.12 or later; downloaded from www.apexrms.com) and packages ST-Sim (package ST-Sim version 4.5.3 or later; obtained remotely from within Syncrosim), and the input and output database (SouthernSnakeRange_Spatial_NRV.ssim.data). The database of input and output data was created and managed by Syncrosim during simulations and should not be tampered with as all parameters and data to run the simulations are contained within. The Authors have never worked within the data folder. "SouthernSnakeRange_Spatial_NRV.ssim.data" contains simulation scenario results. (B) In addition to the simulation database, one folder contains all needed Geotiff rasters (GIS_Rasters.zip) that were uploaded in the Syncrosim database (i.e., already uploaded). These Geotiff rasters will be needed for upload if the simulation is conducted on a server with different directory pathway organization than the Cloud server used by this project. (C) Tabular results and reference conditions and dissimilarity analyses are contained in the folder "Results NRV." (D) All the spatial results of the combined ecological system by vegetation class in 14m Geotiff rasters for each of year 2022 initial conditions), 2023 (first year of simulation), and 2522 (last year of simulation) and all 20 replicates are found in the zipped folder "Results-8_Digit_Raster_500years.zip." (E) The vegetation description of all ecological systems and their vegetation classes (including non-reference classes) used in remote sensing and models are found in the PDF "Descriptions of Ecological Systems_SouthernSnakeRange_09-27-2023."
Description of the data and file structure
- Description of archived data:
(A) "South Snake Range NRV" Folder. While we cannot describe the content of the Syncrosim database and data folder (again, the SouthernSnakeRange_Spatial_NRV.ssim.data was entirely created and managed by Syncrosim during simulations and not the authors), we will describe the rasters within “GIS_Rasters.zip.” More is written below in the Code/Software section about the ST-Sim simulation software running "SouthernSnakeRange_Spatia-NRV.ssim."
(B) “GIS Rasters" contains all the Geotiff rasters at 14-m pixel resolution used in ST-Sim: - SNPLMA_14m_NaturalSYS_020724.tif was an initial condition raster of the ecological system identity whose numeric values are in the Definitions of ST-Sim/Syncrosim and in the PDF "Descriptions_of_Ecological_Systems_SouthernSnakeRange_09-27-2023.pdf";
- SNPLMA_14m_NaturalCLA_020724.tif was the initial condition for vegetation classes within ecological systems where each unique vegetation class corresponded to a numerical code found in the Definitions of ST-Sim/Syncrosim and in the PDF "Descriptions_of_Ecological_Systems_SouthernSnakeRange_09-27-2023.pdf";
- SNPLMA_Elevation_14m_092723.tif was used in Slope Multipliers option in Advanced menu to affect the spread of fire on different slopes (already defined in ST-Sim);
- SNPLMA_FireStarts_14m_113023.tif was used to control the probability of fire initiation caused by lightning strikes and human ignitions, and used in the menu Spatial Settings->Advanced->Transition Spatial->Transition Spatial Initiation Multipliers of ST-Sim;
(C) Tabular non-spatial and spatial scenario results of vegetation classes per system and dissimilarity analyses: The folder "Results NRV" contains all the tabular data used to produce Figures 1 and 2 in the manuscript. There are three types of data:
- Tables of vegetation class proportions per systems and dissimilarity analyses in folder "Systems Class data."
(a) Results of non-spatial (aka, aspatial) and spatial class proportions and dissimilarity values ("ED" for ecological departure). Each Excel file per systems was named in generic format such as "AntBitter_BlackSageNRV" where "AntBitter-BlackSage" is the abbreviation for "Antelope Bitterbrush-Black Sagebrush" (and similarly for all other systems) and "NRV" means Natural Range of Variation, which is the shorthand for reference class proportions. Two main spreadsheets are found in the Excel worksheet names by their simulation scenario: 2745 for non-spatial results and 2747 for spatial results. The MS Excel format was retained because it was created by the ST-Sim software in three spreadsheets that need to be maintained individually to keep results comprehensible. The "Scenario" spreadsheet defined the identity of the two other spreadsheets. A CSV format would destroy the spreadsheets. In the 2745 spreadsheet, the year 2523 (will be 2522 for spatial results due different, but inconsequential, end of simulation choices) final vegetation class proportion per system is found per each of five replicates per column. The average of the five replicates is presented below the five values. In some ecological system, the ST-Sim software was used to directly report the average without showing the time series. This simplified tables. This average value was copied and pasted into the 2747 spreadsheet for dissimilarity calculation. In the 2747 spreadsheet, the year 2522 final vegetation class proportion per system is found per each of 20 replicates per column. The spatial class proportion average is calculated below the 20 values. As explained for the non-spatial results, only the average was reported for some ecological systems by the ST-Sim software. The non-spatial average per class and system were pasted in the row just below the spatial average. Below the two rows of average class proportions, to the right of the term "ED" is the calculated dissimilarity between the two rows of class proportions. The ED value is used to make figures in the manuscript. - Tables of Fire Probability per System and Ratio of MFRI are found in folder "FRI Spatial and Non-spatial." The MS Excel format of the tables was retained because it was created by the ST-Sim software in three spreadsheets that need to be maintained individually to keep results comprehensible. The "Scenario" spreadsheet defined the identity of the two other spreadsheets. A CSV format would destroy the spreadsheets. There are four tables:
(a) "Aspatial_FRI_FirstBatch" contains the fire probability (1/MFRI) outputted per decade starting in 2323 (thus, also 2333, 2343 and so on). This wide table only includes in alphabetical order the systems from Antelope Bitterbrush-Black Sagebrush to Ponderosa Pine (thus the name First Batch). The zero values between the decadal years were outputted as zero although no data were outputted by ST-Sim. After the end of column, the fire probability was averaged per replicate by calculating the sum over all values, including the non-data zeros, and divided by 20 decades of real positive values. The average of the five replicates was calculated and copied and pasted in the "Spatial_FRI_FirstBatch" table.
(b) "Spatial_FRI_FirstBatch" is the spatial version of the previous tables (same systems), except the results are from spatial simulations and include 20 replicates. In this case, decadal results are only outputted from 2322, 2332, 2342, and so on until 2522. Here the grand average across all replicates are calculated across 20 decadal observations and 20 replicates, and reported at the bottom of each column. Below this average is the inverse value where the inverse of the average fire probability is the Mean Fire return Interval (MFRI) in years. The average fire probability from non-spatial simulations was pasted below the spatial MFRI. The non-spatial MFRI was also calculated as the inverse of the average fire probability. Below these numbers to "ratio of MFRIs" is calculated by dividing the non-spatial MFRI by the spatial MFRI.
(c) Two other Excel worksheets are found in this folder: "FRI_SecondBatch" and "FRI_ThirdBatch." "SecondBatch" and "ThirdBatch" refer to different systems in the wide tables. Each of these worksheets combined both the non-spatial (2745) and spatial (2747) scenarios present in different tabs. Average fire probabilities from non-spatial simulations (2745) were pasted below the spatial fire probabilities and MFRI (2747). As for the "Spatial_FRI_FirstBatch" file, the ratio of the non-spatial to spatial MFRI was calculated below the pasted row. - Excel table of copied dissimilarity and ratio of MFRI data to make figures 1 and 2 in document "Tables to make spatial-aspatial_figure". In the table, the first column lists ecological systems and their areas (ha). The other columns in consecutive order are "NRV_Dissimilarity", and "ratio-aspatial/spatial_FRIs." Average data values from the dissimilarity and MFRI ratio were pasted in the appropriate columns. The two figures were submitted to the journal and published.
(D) Spatial vegetation raster results in folder "Results-8_Digit_Raster_500years." All rasters in Geotiff format contained pixels of 14-m resolution. While rasters are shown for year 2022 (initial conditions) and 2023 (first year of the simulation), analyses would be performed on rasters from year 2522. The paper does not directly use rasters for analyses, results are shown for readers that may want to observed spatial reference condition data in the GIS environment. All rasters have the same naming convention: scn2747.sa_8-Digit-Code-Name.it1.ts2022 where "scn2747" means scenario 2747 of ST-Sim; "8-Digit-Code-Name" means that the value of the pixel is an 8-digit code found in the vegetation description document and in the Definitions for Vegetation Classes and State Classes; "it1' is iteration #1 (there are 20 of them); and "ts2022" means timestep 2022.
(E) Vegetation Description file "Descriptions of Ecological Systems_SouthernSnakeRange_09-27-2023.pdf" contains all ecological systems and vegetation classes mapped by remote sensing and models. The description includes reference and non-reference classes mapped. The 8-digit codes associated with each vegetation classes is exactly used as such in the ST-Sim's definitions. All systems are listed in Table of Contents.
Sharing/Access information
Links to other publicly accessible locations of the data:
Not applicable. None of the original remote sensing imagery from the Spot 6/7 at 1.5m pixel resolution is available because the standard license agreement for purchase of any imagery with Airbus Defence and Space prevents us from sharing the imagery. This archival imagery can be purchased from Airbus Defence and Space or from a reseller.
Code/Software
Simulations require the freeware Syncrosim (Syncrosim version 3.1.12 or later; downloaded from www.apexrms.com) and packages ST-Sim (package ST-Sim version 4.5.3 or later; obtained remotely from within Syncrosim; downloaded from www.apexrms.com ). Follow instructions from ApexRMS to install the software. The software will open the simulation database SouthernSnakeRange_Spatial-NRV.ssim and associated folders (SouthernSnakeRange_Spatial-NRV.ssim.data) or by double-clicking the ssim file. Beyond this point, advanced training in the use of Syncrosim and the ST-Sim package will be required. Training can be arranged with ApexRMS at www.apexrms.com. Assuming training was completed, two folders will appear after opening SouthernSnakeRange_Spatial-NRV.ssim in the "Definitions" folder: (a) "NRV Scenario," which contains the "Non-spatial NRV" and "Spatial NRV scenarios" already loaded with results (i.e., no simulations are required to make figures and tables), and (b) all supporting scenarios are in folder "Sub-Scenarios." Should the reader choose to run simulations, the reader may need to upload all rasters in different menus and use a server with at least 24 processors running in parallel mode and 256 GB of memory. The simulations will require one week of run time.
Study Area. The south Snake Range Area of Interest (AOI) of 161,569 ha includes lands primarily managed by the Bureau of Land Management, Great Basin National Park, and US Forest Service, with private lands inholdings and larger tracts at the periphery of the AOI. The AOI is located in east-central Nevada adjacent to the Utah border (Wheeler Peak at 38°59’25.80”N, 114°20’09.48”W). Great Basin vegetation of the south Snake Range is highly diverse as the elevation gradient ranges from 1,219 m to over 3,900 m.
Mapping: One 161,569 ha image from WorldView 2/3 pan-sharpened 0.65 m resolution multispectral satellite imagery (Maxar Space Systems, acquired by Lanteris Space Systems in 2025, https://lanterisspace.com/) was captured on June 22, 2022. Remote sensing analysis was conducted with the software ERDAS Imagine® from Hexagon AB (https://hexagon.com) applied to WorldView imagery.
Remote sensing methodology is explained in Provencher et al. (2021: Climate 9, 79, 2024: International Journal of Wildland Fire 33), albeit for different landscapes. The method started with one of several iterative unsupervised classifications of satellite imagery to differentiate pixel clusters characterized by unique combinations of red, blue, green, and near infrared. At least five examples of each unique combination were visited in the field during the summer and fall leading to >3,000 observations. Months of analysis of field observations applied to parts of the AOI not visited resulted in a geotiff raster of the combined systems and vegetation classes map.
Remote sensing mapped systems (e.g., ponderosa pine) and their vegetation classes (e.g., late-succession open canopy). Systems are stable vegetation types expected in the physical environment under natural disturbance regimes and climate usually named for the dominant upper-layer vegetation (Provencher et al. 2021: Climate 9, 79). Vegetation class is the system’s current status defined by canopy structure, cover, and species composition that can change with ecological disturbances.
State-and-Transition Simulation Modeling of Reference Conditions: The reference condition’s expected proportions were obtained by simulating all 46 systems for 500 years to achieve equilibrium proportions using the ST-Sim package in Syncrosim (version 2.3.12; www.apexrms.com). Spatial modeling methodology for reference conditions is found in Provencher et al. (2024; International Journal of Wildland Fire 33), including replacing all remotely sensed current non-reference vegetation classes with the closest successional reference class from the same systems. Importantly, the non-spatial and spatial simulations shared the same ST-Sim database inputs, except that the spatial simulations required additional spatial input such as vegetation initial condition rasters, size distributions of disturbances, a fire initiation probability raster based on lightning strike data, land use and topographic disturbance constraints, and fire spread constraints (Provencher et al. 2021: Climate 9, 79). The non-spatial simulations were replicates five times, whereas the spatial simulations were replicated 20 times by assigning stochasticity to each ecological disturbance as described by Blankenship et al. (2015). The differences in the number of replicates reflected the use of reference conditions for different components of a larger project and the fact that non-spatial simulations are often substantially less variable than spatial ones, especially when averaged over different temporal observations; thus, the spatial simualtions requires more iterations.
Fire was the most important disturbance in the AOI. Spatial fire ignition probabilities were calculated from lightning strike locations obtained from the Western Regional Climate Center (https://www.dri.edu/western-regional-climate-center/). Strikes were converted to a frequency map using a trial-and-error 12.5 km2 moving window. The frequency values were standardized from 0 (no ignition) to 1 (highest probability of ignition) and converted into a raster of lightning strike density to model natural fire starts. Once ignited, fire spread was modeled using two principles: (1) Prevailing winds elongated each fire predominantly from the southwest to the northeast, while allowing other directions; and (2) fire were spread more rapidly upslope than downslope relative to wind directions (Provencher et al. 2021: Climate 9, 79).
Disturbances other than fire included native herbivory (grazing and browsing) in systems with successional phases critically affected by herbivory, plant mortality from severe drought, short droughts, avalanches, native conifer invasion of shrublands, flooding, insect and disease outbreaks, wet years, very wet years, and succession (Provencher et al. 2021: Climate 9, 79).
We estimated two metrics to show spatial effects on the reference conditions. First, we divided the spatially-simulated by the non-spatially simulated probability of fire per year by system. The hypothesis was that fire spread would increase the probability of fire per year in many systems, especially systems with longer MFRI. The total probability of fire per year was a result of simulations averaged over 20 decadal observations of each replicate simulation. The second metric was the dissimilarity between the distributions of reference class proportions between non-spatial and spatial simulations on the last year of the simulation. Each system contains 2 to 5 vegetation classes representing linear or more complicated succession pathways. The hypothesis was that fire spread to systems with longer fire return intervals would at least reduce the proportion of older classes and increase dissimilarity. We used the ecological departure metric to measure dissimilarity.