Repository navigation
Release v3.2: California RC bridge fragilities, Hazus metadata clean-up, Explorer filters - #48
Merged
Merged
Conversation
The subtypes carry unit-count detail worth keeping in inventories, and the damage and loss models do not distinguish them. The auto-population script collapses them to RES3 where the occupancy enters the DL configuration, so the asset record keeps its subtype.
The ground failure models need a depth class (Shallow or Deep), while data sources like NSI describe the foundation system (slab, crawlspace, pile). The old name conflated the two. The Hazus manuals never name this input, so no FEMA vocabulary is affected, and the flood dataset's FoundationType is a genuine system type and stays as is. BREAKING CHANGE: the seismic Hazus v6.1 schema property FoundationType is now FoundationDepth. The auto-population script still accepts a Shallow or Deep value under the legacy key for backwards compatibility.
An absent GroundFailure or LifelineFacility flag now defaults to False; the script used to write None, which its own validation rejects. LifelineFacility selects the damage model family, so it is now a declared input. OccupancyClass is read unconditionally, so requiring it moves a missing value from a KeyError after validation to a proper validation failure. BREAKING CHANGE: the seismic Hazus v6.1 schema now requires OccupancyClass.
The PGA-based lifeline facility fragilities stop at High-Code, while the general building stock models extend to Very High-Code and Severe-Code. A lifeline evaluation of a building designed to one of the stronger levels previously requested a model that does not exist. Those levels now map to the High-Code lifeline models. The same inventory can now be assessed with either the lifeline or the general building stock models.
The new schema is the v6.1 schema without the two levels Hazus 6.1 introduced; a test asserts exactly that relation, so the copies cannot drift apart. The new config references this dataset for its damage and loss data instead of the alias that resolves to v6.1.
The generator builds two fragility datasets from the Chen et al. (2025) bridge database (DesignSafe PRJ-5910, v1): it applies the label and value corrections, mints one dotted ID per curve, writes the packaged CSV and JSON, and records every ID's source row in a crosswalk. The editorial text lives in metadata_text.json beside it.
Sa(1.0 s) fragilities for 26 groups of California RC bridges from Chen et al. (2025): 116 bridge-system models under portfolio/ and 801 component models under component/. Six duplicate system rows and 95 all-N/A rows are omitted; each model's Comments name its source study, ground-motion suite, and any correction applied to the published values. Tripwire counts in tests move to 13 datasets and 20 fragility pairs.
Datasets that are not FEMA, Hazus or SimCenter got an empty badge in the browse tree and search facets. They now share one category, RESEARCH, resolved by the same function in both places.
Note: pelicun copies EDP_TYPES at import, so these twelve names become valid pelicun demand types without a pelicun change.
The generator turns the 24 rows of EDP_Fragility Database.csv (DesignSafe PRJ-5910 v1) into 25 lognormal fragility models, each a function of the response of its component. The source file moves here from the Sa(1.0 s) authoring folder so the repository keeps one copy; the Sa(1.0 s) generator reads it from the new place for its capacity check. It reads models.csv, the published rows with one row per model and a Model ID of the form CRCB.<component>[.<qualifier>].<scope>.<source>, whose segments select each model's demand type, reference and text.
Demand-Directional is 1 because each EDP is defined in a direction; Column drift is stored as a ratio. Row 6 covers four bridge groups from two studies and becomes two models; it defines only extensive and complete unseating, so its slight and moderate medians are set equal to extensive.
The column was 0 in every row of the three California RC bridge fragility tables. pelicun skips its Incomplete handling when the column is absent, and the data format tests do not require it. The fragility and consequence figures of the documentation plotter and the consequence figure of the Explorer read the column unconditionally; they now treat a missing column as complete. The bridge datasets reach only the plotter's fragility figure, since they have no consequence data.
The generated dataset pages interpolate the description into a dedented template. A description with a line break added unindented lines, so dedent found no common margin and the page title turned into a block quote. Markdown links in the FEMA P-58 descriptions printed literally, since the build is RST only. The generator now indents continuation lines to the template margin and converts the links to RST. The plotter read the key DemageStateWeights, so no page ever showed damage-state weights. The key is spelled correctly now, and the table array is built as dtype object so that a missing weight stays NaN.
Three seismic building datasets shared the title "Hazus Earthquake Methodology - Buildings", and the v6.1 description said version 5.1. Titles now name the methodology and version, without the asset type, so datasets of the same source share a title across asset types. The story-level set is a "story-level prototype" with a description of how it was derived from the building-level models. The hurricane dataset says version 5.1. The bridge descriptions are organized in paragraphs. The Hazus building generators had pre-v3.0 input paths, and the v6.1 generator read the v5.1 inputs. They resolve their paths from the script location now, and the story-level generator applies its own general_information.json over the portfolio metadata it copies.
The four pipeline models shipped as a bare fragility.csv, so the Explorer and the documentation skipped the dataset. The new fragility.json describes the Hazus repair-rate models the curves come from.
The Browse and Search page filters by Hazard, Asset type, and Source, each list narrowed by the selections before it, and the tree groups sources under asset-type headings. The filters select whole datasets and no longer prune by component ID: datasets that share IDs (the Hazus building portfolios and the story-level set, the two hurricane portfolios) could not be told apart, and an ID filter rendered every model of every matching source on each rerun. Choosing a Source opens that source without expanding its models. The Component group filter and the source badges are removed; an empty EXCLUDED_DATASETS set gives the tree and the index one place to leave out a duplicate dataset.
Bump the version to 3.2, refresh uv.lock, date the 3.2 changelog entry, and add the release notes. Raise the Streamlit floor of the explorer extra to 1.58 on Python 3.10 and later. Streamlit 1.58 needs Python 3.10, so a bare floor does not lock against requires-python >=3.9; the second entry keeps 1.48 on 3.9, where the extra is already uninstallable.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This pull request brings
mainto version 3.2.reinforced concrete bridges from the Chen et al. (2025) database:
116 bridge system models, 801 bridge component models, and 25
EDP-based component models. Twelve new demand types support the
EDP-based models.
appear in the Explorer and the documentation.
and hurricane descriptions name the right manual versions, and the
bridge descriptions are organized in paragraphs.
and Source, with the tree grouped under asset-type headings.
links, and damage-state weights.
explorerextra requires Streamlit 1.58 or later on Python 3.10and later.