v0.1.0

Unzip the archive for your computer and run COSMIC2-Explorer from the COSMIC2-Explorer folder (see Install). Your data, downloads and settings are kept in a separate folder; an installed copy updates itself once a week (Settings > Updates).

Changes since v0.0.1

• First start opens Explore with the examples and says the next three steps
• Prepare 0.1.0 as a public pre-release: CITATION.cff, issue forms, Feedback link, README note, changelog heading
• Changelog: time of the index re-read after an update
• Help: how many days the mission comparison needs, from a full day of COSMIC-2 and PAZ
• Refractivity-gradient diagnostics in N-units/km, not N/km
• Gravity-wave map settings explained in plain words
• Per-profile diagnostics computed once per file and shared by the map, tropopause table and scatter
• README quick start points to the tooltips and the word list in Help
• Averages, sections and maps say when they pool COSMIC-2 with PAZ or KOMPSAT-5
• Help: limb sounding in plain words; a rising signal crosses ever higher layers
• Help: how to read it, and Galileo among the tracked signals
• Summary takes the cache total from the index instead of walking the folder on every refresh
• Cache size from directory listings; performance checked on a month of data
• Profile plots name the products a mixed selection left out, and why
• Diagnostic map follows the searched product; sampling caption explains penetration only for neutral products
• Radiosonde CSV height column says geopotential height
• Glossary tooltips on the Statistics card headings; terms for sections, zonal means, weather models, three-cornered hat and ducts
• Changelog wording
• Plot hover values in significant digits so small quantities do not read as zero
• Messages say files, not granules; Get data names the example day
• First start says the examples are shown only when they loaded; otherwise the usual empty-data note
• IVM day drawn as points coloured by latitude; profile hover uses the vertical axis's unit
• Plain-language glossary: tooltips on headings and product codes, and a list in Help
• Diagnostic map leaves out profiles CDAAC flagged bad, like the other cards
• Error-message test uses the test server, not TestClient
• Errors explained: unexpected failures name the error and the log file, rejected values name the field
• README: plain-language introduction and a three-step quick start
• First start shows the bundled example day; plain-language opening in the guide
• podTc2 and scn1c2 indexed with satellite, transmitter and position; leoOrb said to be an orbit file the app does not read
• Profiles tab draws an IVM day against time along the orbit, local time or latitude
• IVM unit strings: GPS-time seconds and squared densities read plainly
• echPrf bending angle described as simulated from ECMWF, not ionosphere-corrected
• Profile CSV gives scnLv2 time units once, not one per file
• File unit strings N, N-unit, unitless and N/cc read unambiguously in titles and headers
• Profile overlays of several products name each product's quantity in the axis title
• Gravity-wave help: wetPf2 follows its ECMWF first guess above about 30 km; use atmPrf
• Scatter says how many occultations it draws twice as atmPrf and wetPf2
• Average groups in a fixed order so colours do not swap between runs
• Changelog: product-aware titles and the radiosonde selection option
• Single-product averages, sections and profile plots title the field as that product holds it
• Averages and sections split by the same quantity_label as the exports
• Exports name dry-derived and model fields per product through one quantity_label rule
• Radiosonde card: the selection option says it uses the chosen product
• Averages keep atmPrf's dry-derived fields apart from wetPf2's
• One once_per_occultation rule for the tropopause table and the diagnostic map
• Tests: remove unused imports and a variable
• Diagnostic map counts each occultation once when both its atmPrf and wetPf2 are selected
• Radiosonde comparison uses one product: dry atmPrf and wetPf2 temperatures are never pooled
• Tropopause table counts each occultation once when both its atmPrf and wetPf2 are selected
• Remove unused imports
• One column_stats helper instead of four copies; unused imports removed
• Remove unused selectionRequest from the frontend
• Remove unused Granule.to_columns, Granule.series and DownloadManager.active_count
• One local-solar-time helper instead of two
• ap_pot method text states only what the file says
• IVM ap_pot labelled as the aperture-plane potential, as its file names it
• Metadata CSV writes list values as semicolon-separated numbers
• Radiosonde station filter uses the same New Year window
• Radiosonde comparison reads the next or previous year's sondes when the time window crosses New Year
• netCDF export height coordinate says positive up, as CF requires
• ivmL2m described as it is: its drift fields are empty in every file checked
• Write the netCDF-3 magic bytes as escapes
• A cut-off netCDF file is refused instead of read as zeros through netCDF4
• Exports started in the same second no longer overwrite each other
• Remove the unused ZIP bundle export
• ionPrf hscale kept as CDAAC defines it, the scale height at orbit altitude (icalib = 0 only)
• Averages and sections keep observations apart from ECMWF and NCEP profiles of the same name
• scnLv2 fallback local time computed; CDAAC's lct attributes are not local solar time
• KOMPSAT-5 direction checked on occultation geometry: it records rising occultations too, and irs labels them right
• Profile outlier filters test each level against its neighbours, not the whole profile
• IVM drift texts follow the files' sign conventions (iv_mer is +up, not north-south); Azim is the occultation-plane azimuth
• Imp_height_corr described as the oblateness correction it is, not a corrected height
• Regridded CSV names atmPrf Temp as the dry temperature, per product
• atmPrf and ionPrf local times computed from UTC; CDAAC's are cut to the minute or in GPS time
• Guide and README: 12 bundled granules of 7 products; help lists every log-interpolated field
• Rapid stream text: scnLv2 and ivmL2m are there, ionPrf is in provisional
• postProc described as it is on CDAAC: IVM files of 2019-2022 and podTc2 of 2020, no profiles
• Bending angle interpolated in log space; three unused functions removed
• Bean and Dutton cited as the 1966 NBS Monograph 92 their DOI resolves to
• IVM ion densities of zero or below are missing (the TEC rescaling drives low densities negative)
• irs = 1 is a setting occultation, as CDAAC's format and same-occultation ionPrf show; no PAZ exception needed
• Times are UTC: CDAAC's date attributes are GPS time, 18 s ahead
• ionPrf TEC integrated from 80 km like CDAAC's tec0; below it the profile is inversion residue
• foF2 from the exact plasma-frequency coefficient (8.98e-3 was 0.015 % high)
• An index from an older version is re-read once at start-up, imported files included
• ionPrf with an E-region archive peak is placed at its F2 peak, up to 1600 km away
• Profile CSV header lists the units of every column; C and mb read as °C and hPa
• README: refractivity and bending angle are interpolated in log space; setting or rising and PAZ; archive-link test
• Profile CSV of mixed products keeps each value under its own column
• CDAAC archive links open the level folders that exist, not per-product folders (404)
• Bending-angle texts name the Galileo and GLONASS frequencies, not only GPS L1/L2
• atmPrf geoid undulation shown in metres (the file gives it in km)
• Dash check skips the local graphify-out folder
• Setting or rising read correctly: ionPrf's setting flag is 1 for setting, PAZ records setting occultations only
• Get data fetches a range of days, one download per day
• Scintillation diagnostics show the tangent height of the S4 peak and where the file is placed
• README: where scintillation files are placed, that they are triggered, and the new checks
• Scintillation files name their satellite C2E1-C2E6 like every other product
• Numbers with a power of ten use a superscript exponent instead of a caret
• Radiosonde comparison reports stations left out by its limit, and the limit can be raised
• S4 texts match the files: 10 s sliding windows, triggered files, one set of class thresholds
• Radiosonde comparison also compares boundary-layer tops of atmPrf and sonde
• Boundary-layer top: steepest fall of refractivity with published acceptance tests, from atmPrf, checked against radiosondes
• Diagnostic map shows the poles when PAZ or KOMPSAT-5 profiles are there
• Place scintillation files at the tangent point of the S4 maximum, not at the satellite
• Ducting texts say a retrieved RO profile cannot show a duct; docs describe the ducting views
• RO minus ECMWF refractivity with and without an ECMWF duct, and around the duct
• One-folder build on Windows and Linux, so a killed app leaves nothing in TEMP
• Tutorial, README and catalog describe the comparison cards as they now work
• Help: tropopause by latitude, weather models and other missions
• Weather-model caption: wetPf2's 1D-Var first guess is the same ECMWF profile it is compared with
• Model comparison and tropopause captions say when the selection mixes missions
• Explore satellite filter can choose COSMIC-2 as a whole when another mission is on disk
• Mission caption: COSMIC-2 minus PAZ temperature on 14 days of 2026, small real difference at 28-30 km
• Mission plot: pair spread hidden until its legend entry is clicked, so the means are readable
• Mission pairs also shown corrected with ECMWF's difference between the two occultations (Gilpin et al. 2018)
• COSMIC-2 vs PAZ/KOMPSAT-5 through ECMWF over the latitudes both missions cover
• Changelog: drop an entry for a change not yet committed (picked up by 940f380)
• Tropopause table fits 1024 px; CDAAC columns only when the selection has atmPrf
• Table number columns keep their trailing zeros (tropopause table showed 11.7 for 11.70)
• Compare COSMIC-2 with PAZ and KOMPSAT-5: collocated pairs and the double difference through ECMWF
• After an update that skipped releases, the panel says how many and links the full comparison
• Release workflow keeps only the two newest releases after publishing, and all tags
• Test that the update installs the newest release even when the installed one is no longer listed
• Download PAZ and KOMPSAT-5 profiles through mission streams, kept apart from COSMIC-2
• Tropopause and cold point by latitude band with 95% intervals, beside CDAAC's own values
• Bending angle against ECMWF by impact height, with the impact-height placement of the model values verified
• Compare radio occultation with the ECMWF and NCEP profiles of the same occultations; refractivity errors by the three-cornered hat
• Refractivity is interpolated in log space, like pressure and density
• Read, download and plot the ECMWF and NCEP model profiles CDAAC colocates with each occultation
• Dropdown labels fit their controls column; the zonal-mean interval has a visible legend marker
• Analysis worker processes exit when the app is killed, so they cannot block an update
• Optional robust outlier screen for Compare averages, their figure and latitude-height sections
• Compare each atmPrf with the wetPf2 retrieved from it: refractivity, dry retrieval and moisture effect by height
• Latitude-height sections show the standard error, zonal means their 95% confidence interval
• Tutorial Guide no longer scrolls sideways: grid columns may shrink below a long citation block
• Statistics table says why wetPf2 has no penetration depth above the ground
• Header wraps instead of making the app scroll sideways below about 1460 px
• Compare averages carry the standard error and 95% confidence interval of the mean
• Loading the bundled examples refreshes the Explore product and date filters
• Links use the theme colours, readable on the dark theme
• After an update the settings panel lists options the new version dropped, and closing it leaves settings.json untouched
• Automatic update starts the new build as an independent program, so it reopens after the swap
• Compare tab says why an ivmL2m selection has no average over height
• Settings keep their stored values across versions; after an update a panel shows what is new and which settings changed
• Weekly automatic update from the GitHub releases, with Check now and a setting to turn it off
• Release archives hold one install folder and are stamped with their tag, one per platform
• Tutorial: describe the Statistics tab
• Browser test of every tab and main action
• Compare and Export follow the searched product
• Explore map: remove the old selection handler before adding a new one
• Validate profiles against NOAA IGRA v2 radiosondes
• Maps and zonal means of per-profile diagnostics, gravity-wave energy by layer and filter
• Gravity-wave energy: high-pass background option, checked against a known wave
• Help: repair the section-departure formula
• Latitude-height sections of any quantity, with departure from the mean
• Statistics tab: sampling, quality flags and penetration depth
• README: remove the planetary-profile claim and match features to the app
• Help, About, tutorial and data policy: describe what the app actually does
• Correct references and DOIs; data terms without a public-domain claim
• Start the CPU worker pool in the background at launch
• Scatter plots: diagnostic labels with units on axes and colour bar
• Number formatter keeps integer zeros with 0 decimals
• Export tab: Write file no longer stops on an undefined variable
• Exported figures: product-aware axis titles
• Plot titles: name atmPrf Temp as dry temperature
• Profiles tab: pick a quantity and height the selected product can draw
• Rebuild index: keep the archive day of each granule's cache folder
• Explore: default date range includes observations from just before the archive day
• Composite figure export: use the Compare tab's quantity, height and grid
• Compare tab: choose the height coordinate; scintillation composites against tangent height
• API: convert float arrays to JSON in one pass
• Profiles tab: draw large selections on the CPU workers
• Composites and profile plots: derive fields only when the quantity needs them
• CSV exports: readable software licence line in the header
• Coverage summary: bin in SQL without the optional FLOOR function
• Profiles: diagnostics panel follows the selection when nothing can be drawn
• Profiles: say when an ionPrf profile has no F2 peak
• Tropopause: skip the moist-layer artefact in atmPrf dry temperature
• Ionosphere: TEC counts negative densities as zero; tec0/tec1 labelled as CDAAC defines them
• Downloads: spread sampling keeps every whole-day ivmL2m file
• Downloads: the concurrent-downloads setting applies without a restart
• README and CHANGELOG: document this round of fixes, CPU workers, science notes
• Pipeline provenance takes the software version from the package
• Pipeline: do not record a Savitzky-Golay step that was not applied
• Index longitudes in [-180, 180)
• Composites: a one-profile group has undefined spread, not zero
• netCDF export: record each profile's product and what Temp/Pres mean in it
• atmPrf Temp/Pres are dry quantities: label them so, never average with wetPf2
• Figures: keep a linear axis when the log switch meets non-positive data
• Settings drawer: Save works again
• Profiles tab: fix crash at the end of the diagnostics panel
• Profile figure export: draw every level, not the display subsample
• Figures: keep captions, embed TrueType fonts, and isolate concurrent exports
• netCDF export: write UDUNITS unit strings
• Settings: 'require good QC', temperature units and export grid now take effect
• Exported figures follow the Plot Theme setting
• Profile error bars: measure scatter about the running mean, not the gradient
• Run per-granule work on a user-set number of worker processes
• Composites: grouping by date works
• Settings: the default stream choice is now saved
• Index: refresh quality code and identifiers when a granule is re-indexed
• Downloads: index granules received before a cancel or network failure
• Day manifest: list collections and fetch file sizes concurrently
• Archive listing: re-check missing directories after 5 minutes
• API: validate grid, DPI and settings values; panels figure without a position
• API: restrict CORS to loopback origins
• Tests: stop the API server and delete the throwaway data folder at the end
• Export provenance: say that plot processing was not applied to the data
• Exports: write numbers at full precision
• Analysis: fix TEC, E-layer, IWV, density, tropopause and interpolation details
• Readers: honour declared fill values, faster netCDF-3 reads, correct local times
• Link to the Releases page (0.x builds are pre-releases, not shown as latest)

All changes, grouped by topic: CHANGELOG.md.