Scientific questions for NDP

Start with a scientific question, then use follow-up questions to explore the result’s coverage, sensitivity and interpretation. CLIO finds relevant NDP collections, obtains the available data and keeps its analysis alongside the sources. You do not need to specify tools or intermediate steps.

NDP can discover and provide access to a collection whose original publisher hosts the files. Alta’s citations therefore still point to the University of Utah’s Hive, and Tahoe’s to TRPA. Keep both the NDP dataset/resource identities and the publisher’s attribution with your results, so a colleague can trace the measurements and reproduce the analysis.

Download the conversations

Open a conversation on its own page to explore the saved charts, figures and recorded tool activity. No CLIO account is needed. These are interactive archives of the actual sessions; opening one does not run the agent, connect to a private Endpoint or start a stream. The pages are large because the saved results and native viewer travel with the conversation.

You can also download the same standalone HTML, including its CSS and viewer, for offline reading in a desktop browser.

Conversation	Questions	Data	Read	Download
Alta-Collins snowfall	3	NDP-acquired observations	Open	40 MB HTML
Lake Tahoe area and shoreline	3	NDP-acquired TRPA geometry	Open	61 MB HTML
Konza Prairie burn observations	3	Previously acquired NDP data	Open	47 MB HTML
Scientific data through an Endpoint	7	Discovery, storage and reuse	Open	33 MB HTML

The examples use Sol 6.1. Each archive includes the full conversation and recorded tool activity, including corrections made during the investigation.

Follow the investigations

Alta, Tahoe and Konza each develop over three questions: establish a result, examine it more closely, then prepare it for a colleague. The seven-question Endpoint conversation follows a collection through discovery, authoring, publication and reuse. Alta and Tahoe start by finding data through NDP; Konza and the Endpoint example reuse data already acquired through NDP.

Alta: from a seasonal total to monthly explanations

1. “How unusual was the 2023–24 snowfall total at Alta-Collins compared with the historical winters available through NDP? Show the comparison and explain any coverage limits.”
2. “Was that driven by especially snowy months, or was snowfall elevated throughout the winter? Let me explore the monthly comparison.”
3. “Can a colleague reproduce the seasonal and monthly comparisons outside CLIO? Package the source data, results and methods, and check that a clean extracted copy reruns.”

The measured October–April total was 635.5 inches, sixth among 18 winters with complete coverage, including 2023–24. Seven earlier winters lack October observations. The follow-up found elevated January–March snowfall alongside below-baseline totals in the other months. The explorer switches between a matched set of complete winters and all available complete months; those baselines answer different questions.

Explore seasonal totals and monthly comparisons, switching between historical baselines. The source measurements and READMEs were acquired through NDP from Hive. Click to enlarge.

The analysis package includes the original measurements and READMEs, a portable standard-library Python script, derived tables, methods and checksums. An independent clean extraction reproduced all three result CSVs byte-for-byte. One station and one winter do not establish a climate trend.

Tahoe: from mapped area to shoreline sensitivity

1. “Can NDP’s Lake Tahoe boundary tell us the lake’s surface area? Make a map and show how you measured it.”
2. “How much does the estimated shoreline length depend on map detail? Show a comparison at several resolutions.”
3. “Can a colleague reproduce both the area and shoreline comparisons outside CLIO? Package the source boundary, results and methods, and check that the files open and the analysis reruns.”

The TRPA polygon acquired through NDP encloses 497.7 km². Its original mapped perimeter is 123.05 km; simplifying the geometry with a 100-metre tolerance reduces that to 112.67 km. The follow-up compares ten settings and matched shoreline views. That tolerance describes geometric simplification, not survey accuracy or image resolution.

The lake-area figure explains the measurement method and its limits. Open the conversation to explore the NDP-acquired TRPA boundary and shoreline comparisons. Click to enlarge.

The portable package includes the original GeoJSON, provenance, pinned Python dependencies, both analyses, an offline report and automated checks. An independent extracted rerun reproduced the numerical results and both images on the tested Windows host. The source does not establish its survey date, reference water level or island exclusions. This measures the mapped footprint, not today’s water extent or the basin’s wetland/riparian acreage.

Konza: particles, wind and carbon dioxide

1. “What do NDP’s Konza burn observations actually show about particles, wind and CO₂? Give me a colleague-friendly explanation and a figure I can explore, with gaps and sensor limits visible. No dot separators in titles or captions.”
2. “Are the biggest particle episodes also the largest CO₂ episodes, or does the answer depend on the time window? Quantify the comparison without implying that correlation proves a cause.”
3. “Which conclusions are robust, and what additional measurements would most improve the explanation? Leave a concise summary and the key figures for a collaborator.”

This conversation continues from files already downloaded through NDP and checked derived tables. The analysis uses reductions of the large wind and CO₂ archives, accounting for multiple measurement packets within individual archive records. Open the conversation to follow the comparison and interpretation; it does not repeat the initial catalog search or raw-file processing.

The strongest particle and CO₂ episodes overlap only partly. The figure compares 958 paired minute observations and shows how descriptive correlations change with the chosen time window. Missing observations, quality flags and unverified clock synchronization limit interpretation. The particle inlet size is not documented, so these readings do not establish PM2.5, PM10 or AQI. Correlation does not establish a combustion source or a causal wind effect.

Compare particle and carbon-dioxide observations across time windows. The association changes with the window; sensor limits and missing observations remain part of the interpretation. Click to enlarge.

From discovery to a reusable Endpoint collection

The Endpoint conversation connects through a saved SSH profile and compares anonymous discovery with the signed-in account’s view. CLIO reads an existing EarthScope station collection, selects metadata epochs covering its retrieval date, creates an organization, and saves a new private dataset. The CSV, map and methods note are each read back and checked against their original byte counts and SHA-256 hashes.

Follow-up questions update the description without replacing the files, submit a separate public description to the configured NDP test catalog, and reload the stored CSV into a linked map and table. The public record contains only the original EarthScope service link; the derived collection remains private.

Of the 312 selected entries, 52 are explicitly labeled synthetic by the source. The other 260 are not thereby confirmed as operating physical stations. Epoch dates describe metadata coverage; this conversation does not test waveform availability or live telemetry. Test-catalog submission is also distinct from production publication or scientific approval.

Share the conversation and the reproducible work

The four downloads on this page are HTML only, with the viewer assets and saved results embedded. They include the conversation, recorded tool activity, loaded skills and captured views. A transcript does not include every original data file or analysis artifact.

To share reproducible work as well as a conversation, ask CLIO to package the original data, analysis scripts, dependencies and methods. The HTML downloads above do not include the separate Alta and Tahoe analysis packages. Opening an interactive result is different from rerunning its calculation.

Snow and water

Has the snow-to-liquid ratio at Alta-Collins changed in the 2023–24 winter compared with its historical record? I think NDP has the observations. Show the evidence, with a chart and the limitations of the comparison.

NDP lists historical Alta-Collins observations and a 2023–24 collection, each with observations and a readme. Compare compatible observation intervals and units; handle missing values, trace precipitation and seasonal coverage. A ratio of seasonal totals differs from an average of individual snowfall-event ratios. One unusual winter does not establish a climate trend.

Other meaningful questions include:

• How sensitive is Alta’s historical snow-to-liquid ratio to excluding very small precipitation events, and does the interpretation of 2023–24 change?
• Do the Alta and Highland disdrometer collections cover the same storms, and what differences in precipitation properties can the overlapping measurements support? Establish sensor variables, overlap and accessible file sizes first.

Fire and atmosphere

Can the NDP observations from the 2022 Konza Prairie burn tell us whether the smoke or CO2 signal changed with the wind? Show what the measurements support.

The Konza burn experiment lists wind, gas and particle instrumentation alongside other resources. Sensor time alignment, units, location and data quality determine which comparisons are valid. A smoke-classification package is not automatically a concentration time series, and an association does not establish causation. Some resources are large or use executable serialization formats; inspect metadata and use suitable readers rather than blindly executing notebooks or deserializing objects.

Further investigations:

• How does the timing of detected smoke agree across the Konza camera nodes, and does wind direction explain any lag? This needs actual timestamped signals from multiple nodes, not catalog descriptions alone.
• How closely do recent VIIRS thermal detections coincide with reported wildfire locations in a chosen California region? Match acquisition/report dates and distinguish satellite pixels from incident boundaries and false detections.
• Which LANDFIRE fuel or vegetation classes surround a selected fire perimeter? Establish compatible dates, classification and spatial resolution before interpreting exposed area as risk or fire behavior.

Habitats and lake basins

What can NDP tell us about how much of the Lake Tahoe basin is wetland or riparian habitat, and how that relates spatially to the lake? Make a small map and quantify what the available scientific data support.

The catalog contains Lake Tahoe and TRPA geospatial collections. A useful result needs actual habitat and boundary features, compatible coordinate systems, a defined basin and overlap handling. Do not add overlapping polygon areas or treat different wetland/riparian definitions as interchangeable. A service landing page is not a geometry file.

Related questions:

• Can NDP’s Lake Tahoe boundary tell us the lake’s surface area? Make a map and show how it was measured. Then investigate how much the mapped shoreline length changes with geometric detail. This is a boundary-measurement question; it does not measure wetland acreage or today’s water-covered surface. Inspect the downloaded coordinate system and distinguish numerical agreement from survey accuracy.
• Which mapped habitat patches lie near the shoreline, and how does their area change with the chosen shoreline buffer? State the buffer definition and units.
• How much of a mapped flood zone overlaps particular land-cover classes? Check whether the flood layer is a probability scenario or an observed event.

A harder hydrology question

Can NDP help compare winter ground displacement around Lake Tahoe over the last five winters, and assess whether it is consistent with changing regional water storage?

This is a research-feasibility task until the required coverage is established. It needs appropriately located GNSS stations with multiple winters of compatible displacement observations, alongside independent hydrological measurements such as snowpack, precipitation, lake level or a water-storage product. A short segment from one station cannot answer a multi-winter regional question.

GNSS displacement can reflect regional hydrological loading, tectonics, snow, reference-frame choices and instrument effects. Establish component units, station epochs, offsets, trend removal and uncertainties. It is not a direct measurement of lake volume. A defensible first result may be a map of suitable stations and a coverage/gap table, followed by an analysis only where the data support it.

For scientific context, Borsa, Agnew and Cayan’s regional GPS study used seasonally adjusted station series and a deformation-to-mass inversion. It supports investigating hydrological loading; it does not establish a direct conversion from one lake-adjacent station’s motion to that lake’s stored volume.

Streams and sharing results

Can we obtain current observations for a station near our study area, measure their freshness and gaps, and compare them with an archived interval?

This requires successful acquisition and consumption, measured timestamps and cadence. Station metadata, a running producer or archived SSE delivery alone cannot establish that result. See sampling and analyzing streams.

Put this investigation’s report, chart and derived table into one private NDP dataset for our research group, and verify that the files can be read back.

The agent must resolve or, when authorized, create the organization, store the selected artifact versions, register their real locations and read back both the dataset record and stored bytes. A catalog record does not itself upload files. Publishing needs write permission and a storage destination. Creating a dataset in an Endpoint’s local catalog does not automatically publish it to the central NDP catalog.

These suggested questions depend on the data and permissions available at your connected service. For your own investigation, ask follow-up questions about coverage, sensitivity and provenance before sharing the conclusions.