The Wildland Almanac - California: a 30 m Landsat-derived time series of California wildland ecosystem properties across forests, shrublands, and grasslands; annual water years 1985–2025 (41 years). 21 properties spanning six themes - vegetation cover, hydrology, fire hazard, carbon, drought-driven dieoff risk, and disturbance severity. The Fire_LCP landscape is delivered as eight single-band COGs (fuel model, canopy cover, canopy height, canopy base height, canopy bulk density, plus static elevation, slope, and aspect). Created using a unified methodology that emphasizes temporal and cross-property consistency, to track change, explore tradeoffs, and support decision-making and scientific discovery. Cloud-Optimized GeoTIFFs with a STAC catalog. CC BY 4.0.
The California release of the Wildland Almanac: an open, 30 m, Landsat-derived record of how California's forests, shrublands, and grasslands have changed over time. Twenty-one biophysical properties - vegetation cover and structure, hydrology, fuels and fire behavior, carbon, drought vulnerability, and disturbance - produced from the Landsat archive through a single cross-consistent pipeline, every water year from 1985 to 2025.
Current release: v2026.1. Annual water years 1985 through 2025, California, all 21 layers. Released under CC BY 4.0. This release stabilized on 29 August 2026 — if you pulled v2026.1 before that date, re-sync (see Releases and versioning).
This directory is the California half of the Wildland Almanac. The contiguous-U.S. release - decadal snapshots for water years 1990, 2000, 2010, 2020, and 2024 - is the sibling dataset at
source.coop/wildland-almanac/conus. The two share methods and conventions; the main differences are temporal cadence (California annual vs. CONUS decadal snapshots), the disturbance encoding (annual vs. cumulative), and the mask (California applies a full wildland-extent mask; CONUS is water-only).
v2026.1/ - the actual data files (1,024 COGs across 21 layer directories)WildlandAlmanac_CA_Documentation.pdf - the binding reference for units, methods, projections, the mask definition, and the UC disclaimer (§6)WildlandAlmanac_CA_QualityBenchmarks.pdf - how each layer compares against independent reference data: what was compared, against what, and how well it agreed. Read this before relying on a layer for a decision.v2026.1/catalog.json - machine-readable inventory, with 21 collections and 1,024 itemswildlandalmanac.org../conus/ - CONUS coverage (decadal snapshots, water years 1990-2024)Twenty-one properties, six themes, every water year 1985 through 2025 (October-September), 30 m resolution, EPSG:5070 (NAD83 / CONUS Albers). All files are LZW-compressed BigTIFF Cloud-Optimized GeoTIFFs with overviews (mode resampling for the categorical fuel-model band, nearest for aspect, average otherwise). Canopy cover, canopy base height, and canopy bulk density are provided as bands of Fire_LCP, not as standalone layers. Canopy height is provided twice, under two different definitions - see the note below.
Grid. California is 45,000 rows × 30,000 columns, EPSG:5070 extent −2,415,585, 1,214,805 to −1,515,585, 2,564,805 - the same continental grid as the CONUS release. All layers are co-registered to sub-pixel precision, so the 21 properties can be treated as a single stack.
Mask. Layers are masked to the California wildland extent. Set to no-data: open water, ocean and inland, from the LANDFIRE Existing Vegetation Type layer; everything outside California; urban, agricultural, water, barren and unclassified land (CALFIRE FVEG WHR types 0 and 46-49); and a set of non-target EPA Level-IV ecoregions (the Central Valley and specified deserts). This release drops a vegetation-richness threshold used in earlier CECS products, so slightly more pixels are retained. The mask is not necessarily complete - treat masked pixels as missing, not zero, in landscape-scale statistics.
Vegetation cover and structure (5)
Veg_TreeFrac/ - fractional tree canopy coverVeg_ShrubFrac/ - fractional shrub coverVeg_HerbFrac/ - fractional herbaceous coverVeg_BareFrac/ - fractional bare/non-vegetated coverVeg_CanopyHt/ - areal-mean canopy height, in centimetres: a straight arithmetic mean across the 900 one-metre sub-pixels, gaps and bare ground included. A pixel that is half closed canopy at 20 m and half bare ground has an areal-mean height near 10 m.Hydrology (6)
WaterFlux_AET_FixedPrecip/ - evapotranspiration if water were never limitingWaterFlux_AET_ObservedPrecip/ - evapotranspiration that actually occurred, given both vegetation and the precipitation that fellWaterFlux_Soilmoisture/ - rooting-zone water at the end of the water yearWaterFlux_SoilmoistureFrac/ - the same, as a fraction of maximum rooting-zone storageWaterFlux_Runoff/ - runoff plus deep percolation, under observed precipitationWaterFlux_Runoff_FixedPrecip/ - runoff in a year of average precipitation (SPI-48 = 0), isolating the effect of vegetation on water yieldFire hazard and behavior (4)
Fire_LCP/ - the components of a FARSITE/FlamMap landscape, delivered as eight single-band Int32 COGs: five per-year (fuel model, canopy cover, canopy height, canopy base height, canopy bulk density) and three static (elevation, slope, aspect). Clip and stack the bands to assemble a landscape - see the Use It page for the carve-and-stack recipe. To convert to SI: divide CH and CBH by 10 for metres, and CBD by 100 for kg/m³. Band 6 (CH) is a stand height, not an areal mean.Fire_ELMFire_FL/ - characteristic head-fire flame length (metres × 100), frequency-weighted across 240 weather scenarios (ELMFIRE, pixel-local mode; no fire propagates between cells)Fire_ELMFire_ROS/ - characteristic head-fire rate of spread ((m/min) × 100), from the same runsFire_ELMFire_BurnProbabilityRelative/ - relative burn probability (relative BP × 1,000,000, int16; the 32,767 ceiling corresponds to 0.032767) from ELMFIRE level-set fire-growth simulation, ~500,000 ignitions per 30 km-buffered ARD tile. Natively 120 m, delivered on the 30 m grid. See the caveat below - this is not an annual probability.The fire properties are the fuels equivalent of
FixedPrecip. Flame length and rate of spread use the same ensemble of 240 weather scenarios in every year; burn probability uses that same ensemble and the same ignition locations, seeding and threading, so nothing random varies between years. The fire models are not tuned over time or space. These layers therefore change from year to year for one reason only: the fuels changed. That is what makes them usable for tracking the effect of management, fire and regrowth - and it is why they are not forecasts for any particular fire season. None of them is a record of where fire was actually observed or how severely it burned; theDisturbancelayers are the ones for that.
Carbon (2)
Carbon_AGB/ - aboveground live biomass, mostly trees (metric tons / hectare; total mass, not carbon). Caveat: comparison against plot inventories shows this may underestimate both absolute biomass and year-to-year increments in high-biomass stands - see the quality benchmarks before using it for regional accumulation accounting.Carbon_GPP/ - gross primary production (g C / m² / yr; mass of carbon, not total mass)Forest dieoff risk (2) - both a unitless index, 0-32,767 (int16); canopy height × the rectified multi-year water deficit
Vulner_TreeDieoff_ObservedPrecip/ - "is dieoff likely here, now?" Exposure to the drought that actually occurred, as a rolling four-water-year deficit. The layer for predicting which pixels are most at risk in a given year, and the one to compare against aerial dieoff mapping. Correspondingly poor at isolating the effect of vegetation density.Vulner_TreeDieoff_FixedPrecip/ - "how exposed is this stand if a severe drought arrives?" - the planning layer. The same calculation with precipitation held at the SPI-48 = −2 climatology, so all variation comes from the vegetation. The layer for isolating density and showing the benefit of management; correspondingly poor at predicting any particular year. Note this is a specified severe drought, not the worst possible one.Disturbance severity (2) - annual loss at disturbed pixels, water years 1986-2024 (boundary years 1985 and 2025 are not produced; see documentation §4.3)
Disturbance_TreeFrac/ - annual loss of tree fractional cover at disturbed pixels (Δ fraction × 10,000, positive = loss; 0 = no disturbance that year)Disturbance_AGB/ - annual loss of aboveground biomass at disturbed pixels (Δ tons/ha × 10, positive = loss). Watch the scaling: Carbon_AGB is stored in tons/ha unscaled, but Disturbance_AGB is stored at ten times that - divide by 10 before comparing it against a difference you computed yourself from Carbon_AGB.Filenames follow WildlandAlmanac_CA_{LayerName}_{WaterYear}.tif. The Fire_LCP bands follow WildlandAlmanac_CA_Fire_LCP_{Band}_{WaterYear}.tif (five per-year bands) and WildlandAlmanac_CA_Fire_LCP_{Band}.tif (three static topographic bands, no year). The version is the directory level (v2026.1/), not part of the filename. See the documentation PDF for full per-layer specifications, units, scaling conventions, methodology, and caveats.
ObservedPrecipandFixedPrecip, and how to choose. The water-related properties - both hydrology and dieoff risk - come in two forms. TheObservedPreciplayers use the precipitation that actually fell, and answer what conditions were in a particular year; they are the ones to compare against independent observations such as stream gauges or dieoff mapping. TheFixedPreciplayers hold precipitation at a reference, so that variation in space and time comes from the vegetation alone; they are the ones for tracking change caused by management or vegetation shifts. The reference differs by layer:WaterFlux_AET_FixedPrecipholds precipitation non-limiting,WaterFlux_Runoff_FixedPrecipholds it at the long-term mean (SPI-48 = 0), andVulner_TreeDieoff_FixedPrecipholds it at severe drought (SPI-48 = −2). The observed-precipitation properties areWaterFlux_AET_ObservedPrecip,WaterFlux_Soilmoisture,WaterFlux_SoilmoistureFrac,WaterFlux_Runoff, andVulner_TreeDieoff_ObservedPrecip.
Two canopy heights, and why.
Veg_CanopyHtis the areal mean - the average height over the entire 30 m pixel, including canopy gaps and bare ground, in centimeters. TheCHband ofFire_LCPis a stand height - an estimate of the height of the taller trees within the pixel, in decimeters, which is the quantity a fire model needs: fire models carry canopy openness separately (through canopy cover), so supplying a gap-diluted mean height would count openness twice and under-shelter surface fuels. The two differ by roughly 2-6× on the same ground, with the largest divergence at low canopy cover. UseVeg_CanopyHtfor vegetation structure, biomass, and change analysis; use theFire_LCPCHband for fire modeling. They are not interchangeable and should not be compared directly.
On
Fire_ELMFire_BurnProbabilityRelative. This layer is the fraction of simulated fires that reached each pixel, and its magnitude is set by the number of simulated ignitions - a modeling choice, not a property of the landscape. It is comparable across space and across years within this dataset, and it is not an annual probability of burning. Do not read a value of 0.0005 as a 0.05% chance of burning in a given year. The layer is calibrated so that simulated burned-area shares by fuel class match the FPA-FOD observed record; the absolute rate is not calibrated.
Note on base names.
Disturbance_TreeFracandDisturbance_AGBare annual here and cumulative in the CONUS release (where they carry a_cumulativesuffix). Do not assume the two are directly comparable without accounting for that difference.
Dieoff spin-up years.
Vulner_TreeDieoff_ObservedPrecipuses a rolling four-water-year window that cannot be filled at the start of the record. Water years 1985-1988 sum the available years and scale to a four-year equivalent (1985 is a single year scaled ×4). They are retained rather than nulled, but they are not comparable with 1989 onward and should be treated as questionable.
The Almanac uses an archival versioning model:
v2026.1/, v2027.1/, …). Files at a published version URL are stable: once a version has been used in published work, its contents are preserved.v2026.2/) with its own DOI, and the prior version will be retained.Next release. v2027 will cover 1985 through the end of water year 2026 and is expected on Source Cooperative in November 2026.
v2026.1stabilized on 29 August 2026 — re-sync if you pulled it earlier. This is the first complete and stable release of the Almanac. Reaching it took several in-place revisions through June and August 2026, the last of which corrected the fire-behavior layers (Fire_ELMFire_FL,Fire_ELMFire_ROS,Fire_ELMFire_BurnProbabilityRelative). Those corrections changed values materially, so a copy pulled before 29 August 2026 is not the same data asv2026.1today. Copies pulled on or after that date are current. From this point the archival model above applies as written: any change that alters a quantitative result will bev2026.2, published as a sibling, andv2026.1will remain as it now stands.
Cite the version you used. Each release receives its own DOI. When citing the Almanac in a paper, EIR, plan, or other document where future readers may need to verify the exact values you relied on, cite the specific version (and DOI), not the dataset as a whole. This is what makes the reference chain reproducible.
For analysis within one release: use one version end-to-end. Do not splice years from different versions - every release reprocesses the full series, so values for a given year may differ between versions.
The data are Cloud-Optimized GeoTIFFs served over both S3 and HTTPS. Three access patterns, in increasing order of effort:
Stream it - most common, no download. GIS tools (ArcGIS Pro, QGIS, rasterio, terra) can read COGs directly from cloud storage, transferring only the bytes needed for your current window.
s3://us-west-2.opendata.source.coop/wildland-almanac/california/v2026.1/{Layer}/WildlandAlmanac_CA_{Layer}_{Year}.tif/vsicurl/https://data.source.coop/wildland-almanac/california/v2026.1/{Layer}/WildlandAlmanac_CA_{Layer}_{Year}.tifDownload one file - curl, wget, browser, or PowerShell Invoke-WebRequest against the HTTPS URL above. To clip a lat/long box without downloading the whole file, use gdalwarp -te ... -te_srs EPSG:4326 /vsicurl/<url> clip.tif - GDAL fetches only the tiles overlapping your box.
Bulk download - AWS CLI with --no-sign-request (no account needed):
aws s3 sync s3://us-west-2.opendata.source.coop/wildland-almanac/california/v2026.1/{Layer}/ ./{Layer}/ --no-sign-request
Building a fire-behavior landscape from Fire_LCP - the eight bands are single-band COGs; clip each to your area of interest, then stack them in LCP band order (elevation, slope, aspect, fuel model, canopy cover, canopy height, canopy base height, canopy bulk density). Modern FlamMap reads the GeoTIFF stack directly. The worked recipe is on the Use It page.
Full how-to with worked examples - ArcGIS Pro, QGIS, Python (rasterio), R (terra), and the cloud-native workflow background - is on the website's Use It page. That page is the authoritative how-to reference.
Start with the quality benchmarks PDF. It sets out, layer by layer, what each was compared against and how well it agreed — the independent reference data, the comparison method, and the result. It is the right place to judge whether a given layer is good enough for what you intend to do with it.
The Wildland Almanac is built to be as useful as the underlying observations allow, and the layers compare favorably with comparable products and with the published literature. But much of this work remains very difficult, and the dataset has real limits. Some pixels are wrong in ways that are known and documented; others are wrong in ways that have not yet been identified. Some layers are better than others; some years are more or less constrained than others; some regions and vegetation types are better characterized than others.
Anyone using these data for a specific decision should look critically at the values for their area of interest, compare against on-the-ground knowledge and independent observations when possible, and consider the documentation's notes on methods and caveats. The California layers are masked to the California wildland extent - water, areas outside California, urban/agricultural/barren land, and a set of non-target ecoregions (Central Valley and specified deserts) are excluded (see documentation §4.4). The decision to release the data publicly reflects our judgment that it uses the best information of its kind available, that we believe it has reached the point where it can aid planning and research, and that use-with-feedback is the path to improvement. Reports of what looks wrong - pixel-level errors, structural issues, or systematic problems - are the most valuable contribution a user can make.
Released under Creative Commons Attribution 4.0 (CC BY). Free to use, share, and adapt with attribution.
University of California disclaimer. These data are a University of California product, provided "as is" with no warranty; the user assumes all risk of use. Use of the data implies consent to the full University of California disclaimer, reproduced in §6 of the documentation PDF.
Goulden, M.L. (2026). The Wildland Almanac - California (Version v2026.1). Source Cooperative. DOI: [pending - EZID]. Released under CC BY.
When citing analyses based on these data, cite the specific version (above). Different versions reprocess the full time series and may differ in detail; the version DOI is what makes your analysis reproducible.
Found something wrong, or have a use case to share? Please file an issue on GitHub: github.com/wildland-almanac/wildland-almanac-site/issues. Reports of errors and notes on real-world use are tracked there and folded into future releases.
Contact: mgoulden@uci.edu · Department of Earth System Science, University of California, Irvine. The data are offered with no promise of technical support, but feedback on errors and use cases is welcome and shapes future releases.
The Wildland Almanac is an outgrowth of the Center for Ecosystem Climate Solutions (CECS), a multi-year University of California research program supported by California's Strategic Growth Council. Earlier work and publications referencing CECS share the same project lineage.