SMO (Smoke contribution to Ozone or Smoke O3) represents the estimated increase in the MDA8 O3 concentration caused by wildfire smoke. It is calculated from the GAM residual as:

Residual = MDA8 O3 (observed) - MDA8 O3_pred (predicted)

If it is a smoke day: SMO = Residual

While the residual is calculated for every day, the SMO is only reported on smoke days.

A positive SMO value means wildfire smoke likely increased ozone at that monitoring site on that day. Because any model analysis will have random variations, it is important to examine how the SMO on an individual day compares to the distribution of all non-smoke days. This is expressed as the quantile, which indicates where on the distribution the individual day falls. For example, an SMO value of 12 ppb with a 98th percentile quantile indicates the value is highly significant. See more discussion of statistics in the statistics doc.

Published data refers to peer-reviewed scientific model outputs (GAM-v2, GAM-v1, EPA EMBER, Smoke PM2.5) that estimate smoke-attributed ozone and PM2.5 concentrations. These datasets require login with a free account because they are research-grade data products that we want to track usage for academic attribution purposes. Simply sign in with your Google account to access all Published layers.

Each dataset covers a specific time period. For example, the GAM-v2 model covers April–October of 2019–2024 (ozone season only). If you select a date outside this range, no data will be available. Check the Layer Descriptions panel (the ℹ icon on the map) for the exact date range of each dataset. For HMS Smoke and Fire data, data from the current day may not yet be available due to processing delays — try selecting yesterday's date.

Use the statistical plots available in the [Map] page:

• Table: View regional or site-level summary statistics including area burned, smoke coverage, and fire metrics in a structured list.
• Bar/Line plot: View state-level aggregated statistics and click individual states for site-level drill-down.
• ParCoords plot: Multi-dimensional visualization to compare trends across different variables or sites simultaneously.
• Scatter plot: Select multiple data layers and compare their values site-by-site. The first selected layer goes to the Y-axis, and additional layers become X-axis traces.

The Latest Analysis & Prediction section (Resources > Data > Latest Analysis & Prediction) provides summarized research outputs for specific calendar years, categorized into Preliminary (initial models) and Finalized (validated results).

• UW GAM-v2 (+2025): Predictions for wildfire seasons (2025+) using the finalized model architecture, UW GAM-v2 (Lee and Jaffe, 2025).
• UW Smoke PM2.5 (+2025): Summarized PM2.5, PM2.5-criteria, and Smoke day across PM2.5 AQS sites using UW Smoke PM2.5 (Jaffe et al., 2026) methodology.

Finalized reports are typically released in September of the following year (e.g., the 2025 finalized dataset is expected in Sep 2026). These reports use the final QC data reported to the EPA's Air Quality System (AQS). Like all Published data, you must be logged in to download these files.

Note on Preliminary data: Preliminary data combines the most recent EPA pre-generated data with AirNow observations, which are not considered final data. Users should exercise some caution when using preliminary outputs, although in practice, the preliminary data usually does not change significantly in the final dataset. Due to staggered reporting cycles and our goal to generate rapid analyses, we start our analysis with the preliminary dataset and switch to the final AQS dataset as soon as possible. Finalized data is established based on the EPA's official "as of" timestamp at the time of publication. Please use our analyses with awareness of these limitations.

The AQS query tool (located under Resources > Data > Published > AQS query) allows you to extract processed data at the individual AQS monitoring station level.

By selecting a Dataset (e.g., GAM-v2), State, and a specific AQS site, you can import historical model outputs and observations. Once data is imported, you can toggle between:

• Location: View the site's physical metadata and its position on a map.
• Table: Browse the full dataset in a structured list and download it as a CSV for external analysis.
• Plot: Generate interactive time-series charts to visualize trends over custom date ranges.

The State report tool (Resources > Data > Published > State report) provides aggregated air quality summaries at the state level. It is designed to help you understand wildfire impacts in the state level

Key features include:

• Temporal Resolution: Choose between "By Year" for annual trends or "By Date" to analyze specific high-impact periods.
• Report Types: Select from various statistics, such as average smoke-attributed ozone (SMO) levels, exceedance day counts, or site-specific summaries within the state.
• Exportable Data: Like the AQS query tool, all generated reports can be reviewed in a table format and downloaded as a CSV file.

Signing in is required for two reasons:

• Access control: Published research data is restricted to authenticated users for usage tracking and academic attribution.
• Community features: MapPosts (community map pins) and Announcements require user identity to display authorship and enable moderation.

We use Google Sign-In exclusively — no separate username/password is created. We only access your display name, email, and profile photo; no other Google data is accessed.

Guest mode gives you full access to all Live data layers (HMS Smoke, Hotspots, AirNow, Wildfire News, NIFC Fires) without signing in. However, you will not be able to access Published model data, write MapPosts, or view Announcements. If you only need real-time air quality monitoring, Guest mode is perfectly sufficient.

The UID (User ID) is your unique identifier within the Smokelyze platform. It is used to manage access permissions for collaboration features like MapPost.

For example, if you want to share a MapPost with specific colleagues rather than the entire public, you can set the post's visibility to "Group". Only members you have added to your group via their UID will be able to see those posts.

How to set up a group:

1. Go to your Profile (User icon) > Settings.
2. Enter your colleague's UID in the input field.
3. Click "Add" to register them as a group member.

Once registered, they will be granted permission to view and interact with your "Group" visibility MapPosts.

Sometimes browsers store an old version of the site in their Cache to save loading time. If you encounter display issues or don't see the latest updates, you should perform a "Hard Refresh":

• Windows/Linux: Press Ctrl + Shift + R (or Ctrl + F5)
• Mac: Press Cmd + Shift + R

This forces the browser to discard the cached files and download the most recent version of the platform. If issues persist, you may need to clear your browser's Cache and Cookies (Shortcut: Ctrl + Shift + Del on Windows/Linux, Cmd + Shift + Del on Mac).

Smokelyze works best on modern browsers:

• Edge, Chrome, Firefox, Safari, etc.
• Mobile browsers (e.g., Chrome for Android, Safari for iOS)
• Internet Explorer is not supported

For the best experience, use a desktop browser with a screen width of 1024px or wider. The map and statistical plots are fully responsive but designed for larger screens.

On some computers with integrated graphics (on Windows), we have found that [Chrome] has a problem in rendering the high resolution graphics from our app. If you see a problem while using [Chrome], we recommend switching to one of the other browsers. If issues persist, please contact us via the Contact page.

These display issues are almost universally caused by Graphics Memory (GPU VRAM) Exhaustion. Our application renders massive amounts of geometric data directly in your browser. Here is how to prevent and fix memory crashes:

• Browser Limits: On certain Windows systems (especially with integrated graphics), Chrome may throttle GPU memory usage around 1GB to prevent system instability. If you encounter black screens, ensure Chrome is updated, or try Edge, which often handles hardware-specific rendering more efficiently on Windows.
• Change Map Type (Switch to [Default] or [Light] Map): Dense background maps (like Vector or Topo) consume the majority of your graphics memory. If you experience crashes, immediately switch your Map Type (bottom left menu) to the Light map. The Light map is highly optimized and practically eliminates VRAM overflows.
• Avoid Rapid Clicking: Clicking through multiple dates or layers extremely fast without waiting for the map to finish drawing can forcefully overload the Javascript Garbage Collector, triggering a Black Screen. Wait for the loading spinner to resolve before clicking the next day.
• Hard Refresh: If your browser tab has already crashed into a Black or Grey screen, simply refreshing normally might not clear the GPU buffer. Perform a Hard Refresh (Ctrl + Shift + R) to completely reset the application memory.

Each dataset has its own specific research paper for citation. Please cite the relevant paper based on the data you used:

• GAM-v2 (Smoke O3): Lee and Jaffe, 2025 (JGR-Atmos)
• GAM-v1 (Smoke O3): Lee and Jaffe, 2024 (ES&T)
• Smoke PM2.5: Jaffe et al., 2026 (GeoHealth) (Forthcoming; currently in minor revisions)

If you wish to cite the Smokelyze app itself, please cite the Lee and Jaffe (2025) (GAM-v2) paper for results relating to Smoke O3, or the Jaffe et al. (2026) paper for results relating to Smoke PM2.5, as the current platform's development is fundamentally linked to this research.

For full citation strings and DOI links, please visit the Data Archive page.

The Announcements page is where we post project updates, new layer releases, system maintenance notices, and key research milestones. Check this page regularly to stay informed about the latest developments in the Smokelyze platform.

The Links page is a curated collection of useful external resources, including official air quality monitoring sites, wildfire tracking tools, and academic partner organizations.

The AI assistant requires a Google Gemini API Key (free tier available):

• 1. Visit Google AI Studio and generate your API key.
• 2. In Smokelyze, click on your Profile icon → Settings → paste your key into the Google Gemini API field.
• 3. Open the Ai Chat panel from the map page and start asking questions!
• Important Notes on Tiers & Pricing:
  • Free Tier: You can use the Gemini API for free. However, please refer to the Gemini API Rate Limits as the free tier has specific quotas for requests per minute and per day.
  • Paid Tier (Optional): If you require higher usage limits or unrestricted access, you can manually switch to a Paid Tier by setting up a billing account directly within your Google AI Studio/Cloud project (Check detailed pricing here).
  • Model Version: The assistant currently utilizes Gemini 3.1-flash-lite. We prioritize models that offer the highest free-tier quotas and most cost-effective performance; therefore, the model version may change at any time to ensure the best user experience.

Your API key is stored only in your browser's local storage and is never saved on our servers. Any costs associated with a paid tier are billed directly to your Google Cloud account.

The AI assistant is a full-stack map controller — not just a chatbot. It can:

• Navigate: Change dates, toggle layers, switch datasets, and fly to specific locations on the map (e.g., "Set the date to 2023-06-10 and show the HMS smoke layer").
• Analyze: Run SQL queries against the entire historical database to find rankings, trends, and comparisons (e.g., "Which site in California had the highest SMO on 2023-07-15?").

Yes. Your API key follows a BYOK (Bring Your Own Key) model:

• The key is stored in your browser's localStorage — it never leaves your device except when making API calls.
• When communicating with our backend, the key is sent via a secure HTTP header (not in the request body) to minimize logging exposure.
• Our server uses the key only to forward your request to Google's API, then discards it. We do not store, log, or cache your key.