How data is exported by Augur for visualisation in Auspice

This page is intended to describe which data underpins the different parts of an Auspice visualisation, and how this data can be produced and exported via Augur. It is not a comprehensive guide for how to generate the data in the first place, nor how to interpret them in Auspice, but it should help understand how to export your data and how to customize the visualization.

This guide is based on Auspice version 2.39 and Augur version 18.0.0. We will endeavour to keep this up-to-date however if you are using newer versions it may pay to check the Auspice changelog or the Augur changelog for any salient changes.

Auspice (visualization) components

It’s helpful to start in Auspice and then work backwards to Augur. The following (annotated) screenshot shows the components of a typical visualisation, which is based on a dataset JSON (sometimes called an Auspice JSON).

We’ll use the following annotated screenshot of the measles virus tree (built using this Augur pipeline) to introduce the various components:

Annotated screenshot of Auspice

Generally, data is defined on individual tree nodes in this dataset JSON via node.node_attrs (dictionary) or node.branch_attrs (dictionary). Dataset-wide metadata such as color scales are typically defined within the top-level meta (dictionary). The best place to really understand the data structures is the JSON schema however as an illustrative example we can look at the data structure behind the tip highlighted in the screenshot above:

  "name": "London.GBR/28.12/D8",
  "node_attrs": {
    "accession": "KT732232",
    "url": "",
    "author": {
      "author": "Penedos et al",
      "journal": "PLoS ONE 10 (11), E0143081 (2015)",
      "paper_url": "",
      "title": "Assessment of the Utility of Whole Genome Sequencing of Measles Virus in the Characterisation of Outbreaks",
      "value": "Penedos et al"
    "div": 0.03952195183861181,
    "num_date": {
      "value": 2012.4986338797814,
      "confidence": [2012.4986338797814, 2012.5721716251735]
    "country": {
      "value": "United Kingdom"
    "region": {
      "value": "Europe"
  "branch_attrs": {
    "labels": {
      "aa": "P: S103P; V: S103P"
    "mutations": {
      "P": ["S103P"],
      "V": ["S103P"],
      "nuc": ["A1184G", "T2113C", "G9914A", "C11201T", "A11522C", "C11843T", "A13151T"]

[a1] Colorings

The available colorings are defined by meta.colorings (array of dictionaries), where each coloring specifies a key which accesses the relevant data in the node_attrs. 1 As well as a key, a title can be specified (which is what the user will see), as well as information about the scale used.

[a2] sidebar data filtering

Currently the node_attrs available for filtering are the union of those defined in meta.filters (array of strings), strain names ( and mutations (branch_attrs.mutations).


We will hopefully change this shortly to include all data defined in a node.node_attrs; please see this GitHub issue to track progress here.

[a3] Temporal display of the tree

If num_date is set on all nodes (including internal nodes) then the TIME branch metric and the Date Range selector is available. Note that to colour the tree by this a corresponding entry in the colorings is required. Conversely, for DIVERGENCE then div (cumulative) must be set on all nodes in the tree.

[a4] Branch Labels

Any keys pair defined in node.branch_attr.labels (dictionary of strings) anywhere in the tree will be available in the branch labels dropdown.

[a5] Tip labels

This uses the same data as colorings, as well as (“Sample name”, which is the default labelling). Note that genotypes (key: “gt”) can’t be used as tip labels (future improvement).

[a6] Explode tree choices (experimental)

Generally, exploding a tree requires an attribute that is both discrete and defined across the entire tree. As this feature is still experimental, we currently use the defined colorings (see [a2], above) excluding any which use a continuous scale and excluding genotypes. Use this feature with caution — you are responsible for choosing sensible attributes on which to explode the tree!

[a7] Geographic resolutions

These are defined via metadata.geo_resolutions (array of dictionaries) in the same format as colorings, and thus an optional “title” may be specified. The key should correspond to both node_attrs (on some or all nodes). Each of the values in the tree (associated with this key) should have a corresponding latitude and longitude defined under demes in this dict.

The “Show transmission lines” toggle will only appear if the attribute is defined on internal nodes (as well as terminal nodes).


This has no effect on the tree, unlike all other blocks described here.

[a8] Legend Swatches

For a given coloring (see [a1], above) the set of values observed for this coloring across the tree is displayed as a swatch. This is responsive — only the currently in view part of the tree is considered, so as you filter the data and zoom the tree the swatches will update. For continuous and temporal scales, the values are grouped into bins automatically (unless specified in the scale).

[a9] What’s shown when clicking on a node?

For the selected node Auspice displays the following information, as applicable:

  1. Strain name, as defined by

  2. Vaccine information (node.node_attrs.vaccine), if applicable.

  3. Sample date (node.node_attrs.num_date) as well as any associated confidence values. For internal nodes this is always labelled as “Inferred date”; terminal nodes are labelled as “Inferred collection date” if a non-zero uncertainty range is provided, otherwise “Collection date”.

  4. Publication info, if provided. This considers (dictionary) and its associated keys value, title, paper_url, journal (only value is required).

  5. Strain information. This considers node_attrs.gisaid_epi_isl, node_attrs.genbank_accession, node_attrs.accession and node_attrs.url. 2

  6. Any other node_attrs provided are displayed. If the attribute is also a colouring then the colouring title will be used. Note that associated confidence values are not displayed.

  7. branch_attrs.mutations are displayed.

    • For terminal nodes, we traverse the tree to display the total mutations between the root node and this tip. This allows us to categorise mutations as “Changes” or “Reversions to root”. Note that (e.g.) a series of mutations such as “A -> T -> C” at position 100 will be shown as “A100C”.

    • For internal nodes we show the mutations ascribed to the branch but categorise them as “Unique”, “Homoplasies” and “Reversions to root”. Note that the first two categories are mutually exclusive however a mutation may appear again in the “Reversions to root” category, as applicable.

Exporting data via Augur

We now consider how information flows through Augur, specifically augur export v2 which produces the dataset (Auspice) JSON described above. This process combines data inputs with parameters configuring aspects of the visualisation and produces dataset files for Auspice to visualise.

digraph { graph [ rankdir=LR, ranksep=0.25, ]; node [ shape=box, style="rounded, filled", fontname="Lato, 'Helvetica Neue', sans-serif", fontsize=12, height=0.1, colorscheme=paired10, ]; edge [ arrowhead=open, arrowsize=0.75, ]; inputs [ shape=record, label="Data inputs|" +"- tree (Newick, required)\l" +"- metadata TSV\l" +"- node-data JSONs\l" +"- colors TSV\l" +"- lat-longs TSV\l", fillcolor=9, color=10, ]; config [ shape=record, label="Configuration|" +"- Auspice config JSON\l" +"- command-line arguments\l", fillcolor=7, color=8, ]; augur [label="augur export v2", fillcolor=1, color=2]; auspice [label="Auspice", fillcolor=3, color=4]; dataset [label="Dataset JSONs"]; inputs -> augur [headport=nw]; config -> augur [headport=sw]; augur -> dataset -> auspice; }

This guide will explain the functionality using an Auspice config JSON, usually named auspice_config.json, to control which data is exported and how it is presented in Auspice, however it is possible to use command-line arguments to specify a subset of this configuration instead; please run augur export v2 --help to see the available options here. The most comprehensive description of this file is via its schema, however to introduce this file here’s a snippet of the Auspice config JSON for the measles dataset presented above:

  "title": "Real-time tracking of measles virus evolution",
  "colorings": [
    {"key": "num_date", "title": "Date", "type": "continuous"},
    {"key": "country", "title": "Country", "type": "categorical"}
  "geo_resolutions": ["country", "region"],
  "display_defaults": {
    "map_triplicate": true
  "filters": ["country", "region"]

What metadata is exported

Augur has two main types of “metadata” which we’ll refer to as “TSV-data” (provided with the --metadata argument) and “node-data” which are JSONs provided via --node-data (see data formats for more). Typically, node-data information is automatically exported as a coloring whereas TSV-data must be explicitly exported by requesting it as a coloring in the auspice-config JSON. 3 Metadata is exported as a coloring if via the following process:

  • All key-value pairs specified in node-data JSONs, unless they match a hardcoded-list of exclusions and special-cases (next section). This has the effect that data provided in node-data JSONs will form a coloring and thus be exported in the dataset JSON.

  • Command-line provided colorings (not covered in this guide)

  • Colorings specified in the auspice_config.json (see below)

Metadata keys <X>_entropy (numeric) or <X>_confidence (dictionary), assuming key <X> exists, will be combined into the same node.node_attrs.X data structure. Auspice uses the entropy value to influence the saturation of the color and the confidence data is shown when hovering / clicking on the branch/node.

Special cases (keys which are excluded or behave differently)

To maintain backwards compatibility a number of keys are excluded or special-cased here and should be treated with caution. If you need to check a particular key you can search for it in the code, but currently these keys include:

  • any key which ends with _confidence or _entropy (see previous section)

  • annotations, aa_muts and muts (see How branch mutations are exported, below)

  • mutation_length and branch_length are converted to the tree’s divergence values 4 and exported as node.div (float); with the first matching key being used. This is optional — time only trees are just fine.

  • key hidden is only exported if its value is one of “always”, “divtree”, “timetree”. We don’t cover this behavior here, and it is somewhat experimental, so try it out first!

  • key clade_membership gets the title Clade (unless you provide one)

  • key num_date (numerical date) gets the title Sampling Date (unless you provide one) and type continuous

  • key author gets the title Authors (unless you provide one) and the exported data structure will incorporate any of the following metadata keys: journal, authors, title and paper_url. 5 In effect, this means that none of these keys are available to be a coloring themselves (apart from author).

  • Some keys are deprecated & changed on-the-fly: authorauthors and numdatenum_date

  • key gt gets the title Genotype; this is automatically created if annotations are defined (see How branch mutations are exported section below) but you can supply your own title if you like. Note that metadata provided under the key gt may be exported, but it won’t actually be used by Auspice!

  • clade_membership is converted to a branch label, see Branch Labels section below.

  • vaccine must be a dictionary with special keys (via a node-data JSON); see the the dataset schema for full details. Auspice uses this to render crosses over certain tips in the tree. This means that vaccine cannot be a coloring itself.

  • The following keys are typically ignored: date, raw_date, clock_length, sequence, aa_sequences, dTiter, dTiterSub. If you wish to export them then you must specify them in your auspice config.

How conflicting metadata is resolved

Node-data JSONs are combined using a deep-merge approach, where we recurse into dictionaries. For non-dictionary values we will overwrite previously seen values, so the order of JSONs provided matters! The information for each node is then added to the per-node (per-strain) TSV-data, with the node-data being used in the case of duplicate keys. As a contrived example:

// node-data file 1
nodes: {NODE_NAME: {country: "USA", region: "North America"}}
// node-data file 2
nodes: {NODE_NAME: {country: "Canada"}}
// metadata TSV
{NODE_NAME: {country: "Mexico", serotype: "ABC"}}
// resulting merged metadata
nodes: {NODE_NAME: {country: "Canada", region: "North America", serotype: "ABC"}}

Configuring coloring names and types

The colorings in the auspice-config JSON typically link the key (where the metadata is to be found) to a title (how the key is to be presented in Auspice) and a type (how should the color scale be constructed). Setting the type to one of “continuous”, “temporal”, “ordinal”, “categorical” or “boolean” lets Auspice use the appropriate color scale. If the type is not provided, it will be inferred in this order:

  1. “boolean” if all values are within ["false", "true", "1.0", "0.0", "1", "0", "yes", "no"].

  2. “continuous” if all values are numeric.

  3. “categorical” if none of the above are satisfied.

The provided colorings in the auspice-config JSON will result in corresponding metadata being exported. As mentioned above, for node-data provided metadata this is not necessary, but if in doubt it’s safest to provide this information in the auspice-config JSON.

Configuring color scales and legends

A colors TSV file may be provided to augur export v2 which is the most common way to associate (discrete) values with actual colors. Such a file has 3 tab-separated columns: the coloring key, the metadata value, and the color hex; no headers are necessary. As an example, here are (some of) the colors TSV used in the measles build:

region  oceania #88BB6C
region  west asia   #ADBD51
region  africa  #CEB541
region  europe  #E39B39

See the guide on custom trait colors for more details on how to use the colors TSV.

Augur itself provides a default set of colors, currently only for “region”. This means that if you don’t specify any colors for region then we’ll use the defaults, however if you specify any colors for “region” in the TSV then the defaults won’t be used (i.e. there is no merging).

This is all optional - if you don’t provide colors then Auspice will use the provided scale type and create a suitable scale for you. Remember that if you define any colors, then you should define them all (for that coloring) - any metadata values you don’t provide colors for will be shown in greyscale!

You can instead define which colors are used in the auspice-config JSON; this is a bit more involved but provides greater flexibility. For discrete scales, you can link individual values to color hex codes (similarly to the colors TSV approach, values not listed here will use a greyscale in Auspice) or for continuous scales by linking pivot points to colours which Auspice will interpolate between to create the final scale. Separate to the color definitions you can limit what values are present in the legend and how they are displayed. For full details please see the schema but the following example may illustrate what’s possible:

"colorings": [
    "key": "num_date",
    "title": "Date (custom colors)",
    "type": "continuous",
    "scale": [
      [2000, "#edf8b1"], [2020, "#7fcdbb"], [2022, "#2c7fb8"]
    "legend": [
      {"value": 2010, "display": "<2020", "bounds": [2000, 2020]},
      {"value": 2020.5, "display": "2020", "bounds": [2020, 2021]},
      {"value": 2021.5, "display": "2021", "bounds": [2021, 2022]},
      {"value": 2022.5, "display": "2022", "bounds": [2022, 2023]}
    "key": "serotype",
    "title": "Serotype",
    "type": "categorical",
    "scale": [
      ["DENV-1", "#781C86"], ["DENV-2", "#83BA70"], ["DENV-3", "#547BD3"], ["DENV-4", "#DF4327"]

Finally, nodes with missing or unknown data 6 are automatically set to gray in the tree and the “unknown” values will not be present in the color legend.

How branch mutations are exported

Nucleotide and amino-acid mutations are provided via node-data JSONs under the keys muts (list) and aa_muts (dictionary), along with annotations (dictionary). These are typically produced by augur ancestral or augur translate and so aren’t detailed here. These metadata are exported as node.branch_attrs.mutations and meta.genome_annotations.

If these metadata are available, then a special coloring is created in the exported dataset JSON: {"key": "gt", "title": "Genotype", "type": "categorical"} which is used by Auspice to allow coloring by genotype. You can also define this in the auspice-config JSON if you wish to use a different title. The presence of these metadata will also enable the entropy panel in Auspice.

Branch Labels

Currently branch labels are unable to be specified beyond two special cases, and there is no way to customize their appearance via the auspice-config JSON. We are in the process of improving this.

  1. If the key clade_annotation is present, these are converted to the “Clade” branch label (node.branch_attrs.labels.clade).

  2. If amino acid mutations are defined, then the aa branch label is automatically created.

Geographic resolutions and lat-longs

The map (geographic) panel in Auspice works by associating certain metadata with lat-long values. The metadata keys used for this are called “geographic resolutions”, and are exported if the certain criteria are met.

The auspice-config JSON should define the geo_resolutions as a list of either dictionaries or strings; the former allows more customization but the latter is much simpler.

Example auspice-config JSON using strings:

"geo_resolutions": [

Example auspice-config JSON using dictionaries:

"geo_resolutions": [
  {"key": "country", "title": "Custom Country Title"},
  {"key": "region", "title": "Custom Region Title"}

The specified keys will be exported from the metadata inputs (“country” and “region”, in the above examples). Finally we need a mapping of the metadata values, called “demes” (e.g. the individual countries or regions) to their lat-long values. This is typically provided via a lat-longs TSV input file, such as:

region  africa  4.070194    21.824559
region  china   35.000074   104.999927
region  europe  49.646237   10.799454

Augur has some defaults configured for “region”, “country” and “division” which are merged in with any user-provided data (the latter takes precedence). We also maintain a very large set of these for our ncov build which may be a helpful reference.

If a deme is defined on the tree but no lat-long value is provided then it will not be shown in Auspice. In these cases augur export v2 will print a warning.

Specifying which colorings are available as filters

Select (discrete) colorings are available for filtering in Auspice (both via the sidebar UI and listed in the footer) if they are defined in the auspice-config JSONs filters list:

"filters": ["country", "region", ...]

Additionally, each mutation and strain name will be automatically available in Auspice’s sidebar UI for filtering.


This behavior may change in the future; ideally all exported metadata should be searchable via Auspice’s sidebar filtering UI.

Customizing the Auspice defaults and available panels

There are a number of other sections of the auspice-config JSON which can configure how Auspice presents the exported data.

The panels (array, with entries of “tree”, “map”, “frequencies”, “entropy”, “measurements”) restricts which panels are available to be rendered in Auspice. By default all panels the dataset supports are available.

The display_defaults (dictionary) configures the defaults for many of Auspice’s UI settings, such as which coloring to use. Please see the Auspice documentation for a list of these.

The data_provenance (list of dictionaries) defines the source of the data. This is exported as meta.data_provenance and is rendered by Auspice at the top of the page.

The build_url (string) specifies URL with instructions to reproduce the build, which is typically a GitHub repo URL. This is used by Auspice in the page header.

The maintainers (array of dictionaries) is used in the Auspice header to identify who created or maintains the dataset.

As an example, here’s how the measles auspice-config uses these keys and you can see here how Auspice renders these:

"title": "Tutorial Nextstrain build for Zika virus",
"maintainers": [
  {"name": "Trevor Bedford", "url": ""}
"build_url": "",
"panels": [ "tree", "map", "entropy" ],
"display_defaults": {
  "map_triplicate": true



Mutations defined here should have an associated entry in the meta.genome_annotations (dictionary of dictionaries). Auspice will still try to display datasets if this is missing, but functionality will be missing.


If both genbank_accession and gisaid_epi_isl are provided, then Auspice will display both. Only if neither are provided do we consider node_attrs.accession and node_attrs.url.


We treat these two sources of information differently because (in our experience) users often have sensitive / private data in metadata TSVs which they may not wish to export, and conversely node-data only exists if it has been computed by a previous step in the workflow and therefore most probably wants to be part of the resulting dataset.


The input is per-branch, however the exported value is cumulative.


The key of each author is generated by grouping authors and adding alphabetical suffixes as necessary. This produces authors “Surname et al A” and “Surname et al B” etc.


Values which aren’t valid are those in the set ["undefined", "unknown", "?", "nan", "na", "n/a", 'none', '', 'not known'] (case-insensitive) .