3 fireSense_dataPrepFit Module
3.1 Module Overview
Prepares the data needed to fit fireSense_ignitionFit (ignition and escape) and fireSense_spreadFit.
3.1.1 Module summary
The module combines historical fire records, vegetation, land cover and climate into the covariate tables used to fit the three fireSense processes (Marchal et al. 2017b, a; 2019): ignition, escape (an ignition that grows beyond one pixel) and spread.
Vegetation, land cover and stand age are snapshots, one per year in P(sim)$dataYears (default 2000, 2010, 2020).
Each year in P(sim)$fireYears uses the latest snapshot strictly before it (a year’s satellite vegetation already shows part of that year’s fires), so every fire year must be after the first data year, and every data year needs at least one fire year.
Climate is annual and is matched to the year of each fire.
Ignition and escape share climate variables (sim$climateVariablesForFire$ignition); spread can use others ($spread).
By default the module builds its own inputs:
- land cover (
rstLCCs) withfireSenseUtils::makeFireSenseLCC()and stand age withLandR::prepInputsStandAgeMap(), per data year; -
cohortDatasandpixelGroupMaps, by runningBiomass_borealDataPrep(andBiomass_speciesData, if it is in the project) in a nestedsimInitAndSpades()per data year. This is cached. Missing modules are downloaded to this module’ssubmodulesfolder; - fire polygons from the latest NBAC release, and ignition points (lightning- and natural-caused only) from the current NFDB release.
historicalClimateRasters has no default and must be supplied, e.g., by canClimateData.
It must cover every fire year; the module stops otherwise.
3.1.1.0.1 Previous SpreadFit results
At init, the module looks in a Google Drive ledger (P(sim)$spreadFitGoogleDriveFolder, P(sim)$spreadFitFilename) for SpreadFit results that overlap sim$studyArea.
If there are any, sppEquiv, the fuel classes, the non-forest groups and the spread climate variables are taken from that fit, and fuel classes are not estimated.
sim$climateVariables must then be supplied.
3.1.1.0.2 Fuel classes
If P(sim)$estimateFuelClasses is TRUE, and the user supplied none of nonForestedLCCGroups, fuelClassTable or a modified FuelClass column in sppEquiv, fuel classes are estimated with fireSenseUtils::assessFuelClasses().
Non-forest land cover classes are grouped by their historical propensity to burn, estimated with a GLM, into two groups.
For forest, each species starts as its own fuel class with an a priori class, by default an interpretation of the FBP fuel types (LandR::sppEquivalencies_CA$FuelClass).
A GLM relates the biomass of each species, in pixels where it is present, to whether the pixel burned.
Species are then merged until P(sim)$targetFuelClasses (default 5) remain:
- Abundance is the proportion of forested pixels where the species is more than 10% of the biomass.
- Rare species (abundance < 5%) are merged with another species of the same a priori class, least abundant first. This can leave fewer classes than the target.
- If there are still too many classes, the next least abundant species are merged, but only with species of the same a priori class and the same coefficient sign (or a non-significant one).
::: {.example #of semi-automated fuel classes-fireSense-dataPrepFit} 7 species were initially present on the landscape; 2 must be combined to achieve the target 5 in this case, Betu_pap and Popu_tre were combined due to their low abundance (above10PctRelB) Pice_gla and Pice_eng were combined due as Abie_las was the most abundant. Pice_mar and Pinu_con were not combined as there were no other species with similar FuelClass values
| species | coef | sign | FuelClass | Abundance | newFuelClass |
|---|---|---|---|---|---|
| Pice_mar | -0.01320 | negative | BlkSprc | 0.066 | Pice_mar |
| Pinu_con | 0.00154 | positive | LdJkPine | 0.625 | Pinu_con |
| Betu_pap | -0.00170 | negative | PopBrch | 0.051 | Bt_pa.Pp_tr |
| Pice_eng | -0.01910 | negative | SprcFrLrch | 0.095 | Pc_en.Pc_gl |
| Popu_tre | -0.02100 | negative | PopBrch | 0.223 | Bt_pa.Pp_tr |
| Pice_gla | -0.00920 | negative | SprcFrLrch | 0.352 | Pc_en.Pc_gl |
| Abie_las | -0.00630 | negative | SprcFrLrch | 0.560 | Abie_las |
:::
3.1.1.0.3 youngAge, missingLCCgroup and treed wetland
Pixels disturbed within P(sim)$cutoffForYoungAge years (default 15) are youngAge, a fuel class of its own.
Time since disturbance comes from the stand age maps and the fire polygons (firePolysForAge), per data year.
Non-forest pixels can also be youngAge, unless P(sim)$nonForestCanBeYoungAge is FALSE.
sim$nonForestedLCCGroups takes precedence over P(sim)$forestedLCC: a land cover class listed in a non-forest group is a categorical fuel even if it is also in forestedLCC.
This lets treed wetland, for example, be simulated as forest by LandR but be a land cover fuel in fireSense.
A flammable class in neither gets no fuel covariate.
Pixels with forested land cover but no cohorts (usually because there was no species data) have no biomass, so they are assigned to one of the non-forest groups, sim$missingLCCgroup.
It is estimated along with the fuel classes; otherwise it defaults to the first group.
3.1.2 Usage
The module is normally run with canClimateData (for historicalClimateRasters and climateVariables) ahead of the three fit modules.
fireSense_spreadFit preparation needs the ignition preparation, so keep fireSense_ignitionFit in P(sim)$whichModulesToPrepare. Preparing fireSense_ignitionFit also prepares the escape data (prepEscapeFitData); the module fireSense_EscapeFit no longer exists and is rejected.
out <- SpaDES.project::setupProject(
paths = list(projectPath = "~/fireSenseFit"),
modules = c("PredictiveEcology/canClimateData@development", # needs its own parameters
"PredictiveEcology/fireSense_dataPrepFit@development"),
times = list(start = 2020, end = 2020),
studyArea = mySA, # buffered to limit edge effects
params = list(fireSense_dataPrepFit = list(
.useCache = c(".inputObjects", "dataPrepBuild", "prepIgnitionFitData",
"prepEscapeFitData", "prepSpreadFitData")))
)
sim <- do.call(SpaDES.core::simInitAndSpades, out)3.1.3 Module inputs and parameters
Table 3.1 shows the full list of module inputs.
| objectName | objectClass | desc | sourceURL |
|---|---|---|---|
| climateVariables | list |
Climate variable definitions for climateData::prepClimateLayers (canClimateData). Unless supplied, .inputObjects builds them from climateVariablesForFire. Declared as an input because .inputObjects sets it: a cached .inputObjects restores only declared inputs, so without this a cache hit returned no climateVariables and canClimateData failed.
|
NA |
| climateVariablesForFire | list |
List with elements ignition and spread, each a character vector of climate variable names, with or without underscores (e.g., CMD_sm or CMDsm). IgnitionFit uses all of ignition; SpreadFit uses spread. Default: ignition = c('CMD', 'cumMDC', 'CMD_sm', 'CMD_sp'), spread = 'auto': the ignition variable that best separates the study area’s worst fire years (see spreadClimateSelection). Unless supplied, climateVariables is built from these.
|
NA |
| cohortDatas | list |
List of cohortData data.tables, one per dataYears, named year<year>. If not supplied, built by running Biomass_borealDataPrep for each data year.
|
NA |
| spreadFirePoints | list |
List of spatial points, one per fire year, named year<year>; each point is the ignition location of one fire in firePolys. The default is the polygon centroids.
|
NA |
| spreadFitAdditionalColNames | character |
Names of the ledger (spreadFitPreRun) columns to read. The default is fireSenseUtils::spreadFitAdditionalColNamesTxt.
|
NA |
| firePolys | list |
List of fire polygons, one per year in fireYears, named year<year>. The default is the latest NBAC release.
|
NA |
| firePolysForAge | list |
List of annual fire polygons used for time since disturbance; as firePolys, but starting cutoffForYoungAge years before the first of fireYears.
|
NA |
| historicalFireRaster | SpatRaster |
Optional raster of fire year, 1985-2020. If supplied it replaces firePolysForAge for time since disturbance. Only downloaded when useRasterizedFireForSpread = TRUE.
|
https://opendata.nfis.org/downloads/forest_change/CA_Forest_Fire_1985-2020.zip |
| historicalClimateRasters | list |
List of SpatRasters of historical climate, named by climate variable, with layers named year<year>. Must be supplied, and must cover fireYears.
|
NA |
| ignitionFirePoints | SpatVector |
Points of annual ignitions, of every fire size, with columns YEAR and SIZE_HA. The default is the fires of the current NFDB release with the fireCauses causes, without rows that repeat another row’s date and location (dropRepeatedFirePoints()).
|
NA |
| missingLCCgroup | character |
The nonForestedLCCGroups name given to forested pixels that are absent from cohortData. The default is the first name; it is replaced if fuel classes are estimated.
|
NA |
| nonForestedLCCGroups | list |
Named list of non-forest land cover classes, e.g. list(wetland = c(19, 23, 32)). Each group becomes a fuel covariate. The default is one group, nf, of every class that is neither forested nor non-flammable; it is replaced if fuel classes are estimated.
|
NA |
| pixelGroupMaps | list |
List of pixelGroupMap SpatRasters matching cohortDatas, named year<year>.
|
NA |
| propFlammables | list |
List of SpatRasters of the proportion of flammable land cover in a pixel, one per dataYears. Built with rstLCCs when that is not supplied.
|
NA |
| rasterToMatch | SpatRaster |
Template raster for studyArea. The default is 240 m, from SCANFI land cover.
|
NA |
| rasterToMatchLarge | SpatRaster |
Optional larger template. If supplied it defines studyArea_biomassParam; if not, .inputObjects sets it to rasterToMatch for Biomass_speciesData. Declared as an input because .inputObjects sets it (a cached .inputObjects restores only declared inputs).
|
NA |
| rasterToMatch_biomassParam | SpatRaster |
Template raster for studyArea_biomassParam, passed to Biomass_borealDataPrep. Expected to cover at least rasterToMatch (formerly rasterToMatchLarge).
|
NA |
| rstLCCs | list |
List of land cover SpatRasters, one per dataYears, named year<year>, on rasterToMatch_biomassParam. The default is from fireSenseUtils::makeFireSenseLCC.
|
NA |
| .ELFind | character |
The ELF this run is for (fireSenseUtils::polygonIDTxt in the SpreadFit ledger), set by fireSense_ELFs; its row’s species colours win when ledger rows disagree. Optional.
|
NA |
| sppEquiv | data.table |
Table of LandR species equivalencies. The default is from LandR::speciesInStudyArea.
|
NA |
| standAgeMaps | list |
List of stand age SpatRasters, one per dataYears, named year<year>; used to create cohortDatas and time since disturbance.
|
NA |
| spreadFirePolys | list |
Not needed from the user: firePolys in the CRS of rasterToMatch, declared as an input because a later event modifies it.
|
NA |
| studyArea | SpatVector | Study area for all data. Should be buffered to limit edge effects on fire spread. | NA |
| studyArea_biomassParam | SpatVector | study area passed to Biomass_borealDataPrep for vegetation calibration | NA |
Parameters are in Table 3.3.
| paramName | paramClass | default | min | max | paramDesc | cacheKey |
|---|---|---|---|---|---|---|
| areaMultiplier | numeric,…. | function…. | NA | NA |
Size of the unburned buffer sampled around each fire: a scalar (buffer area is areaMultiplier fireSize) or a quoted function of fireSize. The default is function(size) 10 fireSenseUtils::multiplier(size), 10 times fireSenseUtils::multiplier; smaller multipliers gave wrong fits. See ?fireSenseUtils::bufferToArea.
|
TRUE |
| bufferForFireRaster | numeric | 1000 | 0 | NA |
Buffer distance within which separate patches of burned pixels count as one fire. Only used when useRasterizedFireForSpread = TRUE.
|
TRUE |
| cutoffForYoungAge | numeric | 15 | NA | NA |
Age at and below which pixels are considered ‘young’ (young <- age <= cutoffForYoungAge)
|
TRUE |
| dataYears | integer | 1985, 19…. | NA | NA |
Two or more increasing years for which vegetation, land cover and stand age are built (cohortDatas, rstLCCs, standAgeMaps, …). Each fire year uses the latest data year strictly before it (a year’s satellite vegetation already shows part of that year’s fires), so every fireYears must be after the first, and every data year needs at least one fire year before the next data year.
|
TRUE |
| estimateFuelClasses | logical | TRUE | NA | NA |
Estimate fuel classes with fireSenseUtils::assessFuelClasses? Skipped if the user supplies nonForestedLCCGroups, fuelClassTable or a FuelClass column that differs from LandR’s, or if a previous SpreadFit exists for the study area.
|
TRUE |
| fireYears | integer | 1986, 19…. | NA | NA |
Years of fire records to use for fitting. All must be after the first of dataYears, and historicalClimateRasters must cover all of them. The default runs from 1986, the year after the first SCANFI V2 year, to the latest year with historical climate for every tile (fireSenseUtils::defaultFireYears()); climate is the last of the inputs to reach a year. fireSense_ELFs takes its fireYears default from the same function.
|
TRUE |
| flammabilityThreshold | numeric | 0.1 | 0 | 1 |
Minimum proportion of flammable fine-resolution land cover for a rasterToMatch pixel to be flammable. Only used when rstLCCs is built here.
|
TRUE |
| forestedLCC | numeric | 81, 210,…. | NA | NA | Forested land cover classes - these differ from non-forest because the biomass and composition of fuels are taken into account by fireSense, while non-forest classes are treated categorically | TRUE |
| igAggFactor | numeric | 4 | 1 | NA |
Aggregation factor (number of rasterToMatch cells per side) for the ignition and escape covariates.
|
TRUE |
| escapeSizeHa | numeric | 50 | 0 | NA |
Size (ha) a fire must reach to count as escaped, and the smallest fire the spread model is fitted to. The escape model’s response is the ignitions that became a fire polygon of at least this size (fireSenseUtils::escapedIgnitions()). Smaller fires’ sizes become nonEscapedFireSizesHa.
|
TRUE |
| fireCauses | character | natural,…. | NA | NA |
Cause groups of the fires all three models use: "natural", "unknown", "human", "other" (fireSenseUtils::fireCauseCodes() maps them to NFDB CAUSE and NBAC FIRECAUS values; a missing cause is "unknown"). Sets the ignition points, the polygons that add an ignition and an escape, and the polygons the spread model is fitted to. Checked against fireCauses of the other modules (SpaDES.core::paramCheckOtherMods()), which must match. Use a new fitFileSuffix when changing it.
|
TRUE |
| fuelClassCol | character | FuelClass | NA | NA |
the column in sppEquiv that defines unique fuel classes. A column named FuelClass exists in the LandR::sppEquivalencies_CA and will be used by default. To change the FuelClass classifications, add a column to that table, or to sim$sppEquiv and then modify this fuelClassCol parameter
|
TRUE |
| fuelCovariates | character | domSecWe…. | NA | NA |
How the spread-fit fuel covariates are represented. "domSecWetland" (default): dom_agb_<class> and sec_agb_<class> (the two fuel classes with the most total treed AGB over the fit study area) and treedWetland_agb (all tree AGB on treed-wetland pixels, removed from dom/sec there); other classes are not covariates; see fireSenseUtils::fireSenseCovariatesCreate(). "species": the previous one column per fuel class. fireSense_dataPrepPredict follows whichever a fit used; this is not a parameter there.
|
TRUE |
| minBufferSize | numeric | 5000 | NA | NA |
Minimum number of cells in each fire’s burned-plus-buffer sample, applied after areaMultiplier.
|
TRUE |
| minBurnableFireSizeHa | numeric | NA | 0 | NA |
A fire’s spread-fit target size is its burned area on burnable pixels (landcoverDTs), the only pixels the fit can burn, not its polygon area. A fire with less than this (ha) on burnable pixels is left out of the spread fit (fireBufferedListDT); it stays in the escape data. NA (default) is escapeSizeHa: the spread objective scores only fires of at least escapeSizeHa (fireSenseUtils::escapeSizePixels()), so smaller ones would only add unscored buffers.
|
TRUE |
| minCovariateProp | numeric | 0.05 | 0 | 1 |
A land-cover class covering less than this proportion of the ELF’s flammable pixels (most recent data year) is too rare to estimate a spread coefficient for. A non-forest class joins the non-forest group nearest its burn rate instead of being clustered on its own (fireSenseUtils::assessFuelClasses()); with less treed wetland than this there is no treedWetland_agb and its tree AGB stays in dom_agb_/sec_agb_ (fireSenseUtils::fireSenseCovariatesCreate()). 0 keeps every class.
|
TRUE |
| nonflammableLCC | numeric | 0, 20, 3…. | NA | NA |
Non-flammable classes in rstLCCs; the default, fireSenseUtils::fireSenseNonflammableLCC, is no data, water, rock, snow/ice and barren land.
|
TRUE |
| nonForestCanBeYoungAge | logical | TRUE | NA | NA |
if TRUE, burned non-forest will be treated as youngAge. Recommended to be TRUE as burned forest is often classified as non-forest. youngAge is resolved for each fire year from the data year’s time since disturbance, which covers forest and non-forest alike, so FALSE is not supported and stops the spread and ignition preparation.
|
TRUE |
| scanfiVersion | character | V3 | NA | NA |
SCANFI land-cover version for non-forest land cover, "V2" or "V3". Passed to fireSenseUtils::makeFireSenseLCC().
|
TRUE |
| sppEquivCol | character | LandR | NA | NA |
column name in sppEquiv object that defines unique species in cohortData
|
TRUE |
| heldOutFold | integer | NA | NA | NA |
NA (default): unchanged. 1 or 2: a cross-validation fold is being fitted (the same parameter as in fireSense_spreadFit), so the SpreadFit ledger is not relevant and Init() does not read it: sim$spreadFitPreRun stays NULL and this module derives the species, fuel and climate objects itself, as for an unfitted study area. Set it for every fireSense module at once with .globals = list(heldOutFold = 1L); Init() stops if fireSense_spreadFit has a different value. Any other value is an error.
|
TRUE |
| doFit | logical | FALSE | NA | NA |
The same parameter as in fireSense_spreadFit: FALSE (default) never fits, and Init() stops, before any data preparation, when the SpreadFit ledger (fitFileSuffix) has no fit for this ELF; NA fits only an ELF without a ledger fit; TRUE always refits. When this ELF’s ledger fit is reused (FALSE, or NA with a ledger row), the spread formula takes that fit’s climate variable(s); when it is fitted, climateVariablesForFire$spread as given ("auto": chosen afresh). Set it for every fireSense module at once with .globals = list(doFit = ...).
|
TRUE |
| fitFileSuffix | character | NA | NA |
Text put after the model tag in the ledger file names, e.g. "_boundfix_rep81", so a test run writes and reads ("latest") only its own files, never the production ones. "" (the default) writes the unsuffixed file and reads ("latest") every file of the model tag, newest by the date in its suffix (fireSenseUtils::ledgerFileOrder()); experiment fits belong in spreadFitLedgerFolder(testing = TRUE). The same parameter of fireSense_spreadFit, fireSense_ignitionFit and fireSense_ELFs; set it for all with .globals = list(fitFileSuffix = ...); modules that disagree stop (SpaDES.core::paramCheckOtherMods()). Ignored when a file name is given.
|
TRUE | |
| spreadFitGoogleDriveFolder | character | https://…. | NA | NA |
URL of the Google Drive folder holding the ledger of previous SpreadFit results (spreadFitFilename), read with reproducible::CacheGeo.
|
TRUE |
| spreadFitFilename | character | latest | NA | NA |
Name of the ledger file in spreadFitGoogleDriveFolder: study area polygons with their fitted SpreadFit parameters. "latest" (the default) takes each polygon’s fit from the most recent ledger file that has it (fireSenseUtils::latestSpreadFits()).
|
TRUE |
| spreadIntercept | logical | TRUE | NA | NA |
If TRUE, the spread formula has an intercept (~ 1 + ...) and fireSense_spreadFit fits it, with the covariates centred, so the coefficients describe variation and the intercept the level. Covariates are rescaled to 0-1 and so are all >= 0: without an intercept the level of the linear predictor is set by the coefficients alone, and a climate coefficient trades off against the fuel and non-forest ones (ELF 13.1). TRUE is the default; FALSE gives ~ 0 + .... Ignored if sim$fireSense_spreadFormula is supplied. Needs a fireSense_spreadFit and a fireSense_spreadPredict that know the intercept.
|
TRUE |
| targetFuelClasses | numeric | 5 | 1 | 7 | the target number of unique fuel classes when using semi-automated approach | TRUE |
| useRasterizedFireForSpread | logical | FALSE | NA | NA |
Use historicalFireRaster in place of fire polygons for spread? Not currently supported: TRUE stops with an error when preparing SpreadFit.
|
TRUE |
| wholeLandscapeSpreadCovariates | logical | FALSE | NA | NA |
If TRUE, fireSense_annualSpreadFitCovariates and fireSense_nonAnnualSpreadFitCovariates have every flammable pixel of the study area, not only the fire buffers (same names and columns), so that fireSense_spreadFit can re-simulate fires cut at a buffer edge at their true size. The buffer tables are then not made. fireBufferedListDT and the fire objects are unchanged. Costs minutes and 0.5-1.5 GB per ELF in the event cache. The parameter is in the prepSpreadFitData cache key, so that event runs once more after this module version even when it is FALSE.
|
TRUE |
| whichModulesToPrepare | character | fireSens…. | NA | NA |
Which fireSense fit modules to prep? defaults to both. Preparing fireSense_ignitionFit prepares its ignition and escape data.
|
TRUE |
| .studyAreaName | character | NA | NA |
studyArea name used in file names and cache tags; NULL derives it from sim$studyArea.
|
TRUE | |
| .useCache | logical | FALSE | NA | NA | Should this entire module be run with caching activated? This is intended for data-type modules, where stochasticity and time are not relevant | TRUE |
| .neverCache | character | init | NA | NA |
Events that are never cached, whatever .useCache says (read by SpaDES.core, 3.2.1.9037 or later): init asks Google Drive whether the spread-fit ledger already has parameters for this study area, and a cached init replays an older answer.
|
TRUE |
| .useCacheArgs | list | list(.ca…. | NA | NA |
Extra reproducible::Cache() arguments, by event. A cached event’s digest covers this module’s code but not the package functions it calls, so dataPrepBuild and prepSpreadFitData pass those in .cacheExtra: a changed function then re-runs the event. Every cached event after init also keys on sim$spreadFitPreRun, evaluated when the event runs: Init() has by then assigned it (this study area’s ledger rows, or NULL), so a new fit of this study area re-runs the events, while rows of other study areas do not.
|
TRUE |
3.1.4 Events
All events run once, at start(sim).
3.1.4.0.1 init
Checks Google Drive for a previous SpreadFit (see above). It cannot be cached, because that would freeze the answer.
3.1.4.0.2 dataPrepBuild
Runs straight after init, before other modules’ init.
Puts rstLCCs on rasterToMatch, and builds, per data year, flammableRTMs, landcoverDTs and nonForest_timeSinceDisturbances.
Fuel classes are estimated here.
Can be cached: add "dataPrepBuild" to .useCache.
3.1.4.0.3 prepIgnitionFitData
Fuel, climate and lightning rasters are aggregated by P(sim)$igAggFactor, and ignitions are counted per coarse pixel and year.
sim$fireSense_ignitionCovariates has one row per coarse pixel and year.
Fitted probabilities only apply at that resolution, so the template sim$ignitionFitRTM is also output, with attributes nonNAs (the number of rows) and meanForestB.
No ignition formula is built: fireSense_ignitionFit fits with xgboost, which does not use one.
3.1.4.0.4 prepEscapeFitData
Adds an escapes column to the ignition covariates: the ignitions that grew beyond one rasterToMatch pixel (not one coarse pixel).
Builds sim$fireSense_escapeFormula if it was not supplied. Needs prepIgnitionFitData.
3.1.4.0.5 prepSpreadFitData
Spread uses fires of every cause, where ignition uses lightning- and natural-caused ones only, and only fires larger than one pixel.
Each fire needs an ignition point in a flammable pixel inside its polygon (sim$spreadFirePoints; default the polygon centroid); fireSenseUtils::harmonizeFireData() enforces this.
Each fire polygon is buffered, with the buffer size set by P(sim)$areaMultiplier and P(sim)$minBufferSize, and only burned and buffer pixels are kept (sim$fireBufferedListDT).
The module stops if buffer pixels are fewer than five times the burned pixels.
To keep the objects small for the optimizer, the covariates are split in two:
sim$fireSense_annualSpreadFitCovariates, one table of climate per fire year, and sim$fireSense_nonAnnualSpreadFitCovariates, one table of fuels per data year.
Fuel columns that are all zero are dropped.
Builds sim$fireSense_spreadFormula if it was not supplied.
P(sim)$useRasterizedFireForSpread = TRUE is not currently supported.
3.1.5 Module outputs
Description of the module outputs (Table 3.5).
| objectName | objectClass | desc |
|---|---|---|
| climateVariablesForFire | list |
As the input. If previous SpreadFits exist, spread becomes the climate variables of all of them (prediction blends neighbour ELFs) plus this ELF’s own, and they are added to ignition. The spread formula has only this ELF’s own (see doFit).
|
| spreadFitPreRun | data.frame |
Ledger rows of previous SpreadFit results that overlap studyArea, from CacheGeo: a geometry column (convert with sf::st_as_sf), plus polygonID and the columns in spreadFitAdditionalColNames. NULL if there is no previous fit. Init() always assigns it (fireSense_ELFs may have put the whole ledger there before). It is not an input: as one, it would be in the .inputObjects cache key and any ELF’s new fit would redo the data preparation. The cached events after init key on it through .useCacheArgs, so a cached event cannot restore an old row over the one Init() just read.
|
| studyAreaWithSpreadParams | sf |
Same as spreadFitPreRun; not created if there is no previous fit.
|
| sppColorVect | character |
Named vector of hex colours, one per species. Only created if a previous SpreadFit exists, from the sppEquiv stored with it.
|
| sppNameVector | character |
Sorted species names (sppEquivCol) from the sppEquiv stored with a previous SpreadFit; only created if one exists.
|
| spreadClimateSelection | data.table |
With climateVariablesForFire$spread = 'auto': each candidate’s AUC for separating the worst quarter of fire years (by area burned), its Spearman correlation, and which was chosen.
|
| climateVariables | list |
Climate variable definitions, as used by climateData::prepClimateLayers (canClimateData). Unless supplied, built from climateVariablesForFire for fireYears (and canClimateData’s projectedClimateYears, if any). If a previous SpreadFit exists, the variables of that fit are added.
|
| fireBufferedListDT | list |
list of data.tables with fire id, pixelID, and buffer status, of the fires the spread model is fitted to: those with at least minBurnableFireSizeHa on burnable pixels.
|
| fuelClassTable | data.table | Fuel class assigned to each tree species; only created if fuel classes are estimated. |
| spreadFirePolys | list |
List of annual fire polygons used for SpreadFit: larger than one pixel and matched to spreadFirePoints.
|
| fireSense_annualSpreadFitCovariates | list |
List of data.tables, one per fire year, of pixelID and the spread climate covariates in the fire buffers (every flammable pixel with wholeLandscapeSpreadCovariates).
|
| fireSense_escapeCovariates | data.table | ignition covariates with added column of escapes |
| nonEscapedFireSizesHa | numeric |
Sizes (ha) of this study area’s natural-cause fires (ignitionFirePoints) below escapeSizeHa, for giving a forecast’s non-escaped ignitions a size.
|
| fireSense_escapeFormula | character | formula for escape, using fuel classes and landcover, as character |
| fireSense_ignitionCovariates | data.table | table of aggregated ignition covariates with annual ignitions |
| fireSense_nonAnnualSpreadFitCovariates | list |
List of data.tables, one per dataYears, of pixelID and the fuel covariates in the fire buffers (every flammable pixel with wholeLandscapeSpreadCovariates).
|
| fireSense_spreadFormula | character | formula for spread, using climate and vegetation covariates, as character |
| fuelClassRoles | list |
Only when fuelCovariates = "domSecWetland": list(domClass =, secClass =), the fuel classes chosen once for this ELF by fireSenseUtils::chooseDomSecFuelClasses(). Both NA with fuelCovariates = "species" or when the ELF has no tree fuel class.
|
| ignitionFirePoints | SpatVector |
The input, in the CRS of rasterToMatch and clipped to studyArea; prepIgnitionFitData adds an ignition for each natural-cause fire polygon that no ignition point matches (fireSenseUtils::polygonIgnitions()).
|
| ignitionFitRTM | SpatRaster |
Template raster with the resolution and extent of fireSense_ignitionCovariates. Attributes: nonNAs, the number of rows in that table, and meanForestB, the mean forest biomass per pixel.
|
| landcoverDTs | list |
List of data.tables, one per dataYears, of pixelID and a 0/1 column per non-forest group, for flammable pixels.
|
| lightningMaps | SpatRaster | A 4-layer SpatRaster of lightning: lightningDays, lightningDensity, positiveCG, positiveCGdensity |
| missingLCCgroup | character | As the input, or the estimated group if fuel classes are estimated. |
| nonForestedLCCGroups | list | As the input, or the estimated groups if fuel classes are estimated. |
| nonForest_timeSinceDisturbances | list |
List of SpatRasters, one per dataYears, of years since disturbance in flammable pixels, forested or not.
|
| rstLCCs | list |
The input, on rasterToMatch.
|
| flammableRTMs | list |
List of binary SpatRasters of flammable land cover on rasterToMatch, one per dataYears.
|
| sppEquiv | data.table | sppEquiv table potentially modified with new or overwritten fuel class |
| spreadFirePoints | list |
List of ignition points, one per fire year, for fires larger than one pixel, harmonized with spreadFirePolys. Column fireSenseUtils::burnableHaTxt is each fire’s burned area (ha) on burnable pixels, the spread fit’s target size.
|
| propFlammable | SpatRaster |
Last element of propFlammables.
|
| standAgeMap | SpatRaster |
Last element of standAgeMaps.
|
| rstLCC_RTM | SpatRaster |
Last element of the output rstLCCs, i.e., on rasterToMatch.
|
| rstLCC | SpatRaster |
Last element of the input rstLCCs, i.e., on rasterToMatch_biomassParam.
|
| flammableRTM | SpatRaster |
Last element of flammableRTMs.
|
| landcoverDT | data.table |
Last element of landcoverDTs.
|
| nonForest_timeSinceDisturbance | SpatRaster |
Last element of nonForest_timeSinceDisturbances.
|
3.1.6 Links to other modules
This module links to fireSense_ignitionFit and fireSense_spreadFit
3.1.7 Getting help
Contact the authors for help.
https://github.com/PredictiveEcology/fireSense_dataPrepFit/issues