D3LTool

Everything in one app

Main window

Three steps

Install & sign in

Double-click D3LToolNASA.exe — a single file, no Python needed; the macOS .app and Linux binaries are on GitHub Releases, and CLI users can simply pip install d3ltool. Sign in under “Settings → Sign in to Earthdata” with your free NASA Earthdata account — search works without login, downloads need it. Credentials stay on this machine only (~/.netrc).

Search

Pick a product group and product (or “custom” for any CMR short_name), optional version (blank = latest) and start/end dates. For the area, click tiles on the sinusoidal map (multi-select, or type h04v03, or use the lat/lon locator) — or enter a bounding box for MERRA-2 style data. The GLASS group (v2.6, v2.7) lists 133 variants (131 verified) from the public glass.hku.hk archive across four archive layouts — static collections need no dates, global CMG products need no tiles — and GLASS searches never ask for a login. Hit the red Search button and results appear in the top-right table.

Download

Tick files in the results table and click “Add to download queue”. The task table shows live per-file progress and speed with pause/resume/retry; finished files are skipped on re-download and interrupted ones resume automatically. Download folder, concurrency and proxy live in “Settings”.

Browser hand-off

When a download keeps failing (some proxies trigger 401/403), select the rows and use “Export download list” (“Open in browser” pops the .html index): your signed-in browser or a DownThemAll-style extension fetches the files — the .txt list also feeds wget -i / aria2c -i. The MCP tools export_download_urls / open_download_page give AI clients the same move, with no password ever in chat or MCP.

# command line
pip install d3ltool
d3ltool search VNP46A1 --start 2020-03-01 --end 2020-03-31 --tile h04v03
d3ltool download VNP46A1 --start 2020-03-01 --end 2020-03-31 --tile h04v03 --dest D:/data
d3ltool search M2T1NXSLV --start 2019-08-01 --end 2019-08-31 --bbox 100 20 120 40
d3ltool search LAI_MODIS_250M --start 2018-08-01 --end 2018-12-31 --tile h28v05   # GLASS, no login

GLASS + R post-processing

The GLASS module (v2.5) is ported from the author's R package glass.sqdaitools v0.1.0 (docs) — a matched pair: D3LTool covers product discovery and resumable, size-validated batch download (no sign-in; the archive is public), while fill-value & scale-factor handling, multi-tile mosaicking, reprojection and temporal compositing run in R with glass.sqdaitools:

# R >= 4.1; heavy deps terra / sf
devtools::install_github("GISerDaiShaoqing/glass.sqdaitools")
glass.sqdaitools::glass_process(files, roi = my_roi,
                                product = "LAI_MODIS_250M",
                                fun = "mean", outfile = "lai_mean.tif")

Upgraded in v2.6: the fully verified 131/133 registry, all four archive layouts (composite folders, static collections, year folders, YYYYDDD day folders), global CMG tile immunity, per-variant data-file extensions from the verification log, and the measured glass.hku.hk politeness policy for directory listings (8/24/72 s backoff + cooldown retry, ≥1.2 s between listings).

Self-adapting in v2.7: the ~50 variants carrying no composited cadence no longer trust the registry's year_doy label — search lists each year folder once and follows the archive's real shape (files flat in the year folder with A-header tokens, yyyyMM month codes, or a one-level version wrapper such as NPP's GLASS_NPP_0.05D_V40/); the stride infers from observed composite gaps, span semantics decide selection, and a single-file year keeps the file whole. Verified live: FVC_AVHRR_0_5D for 2018-08-01..09-15 returns exactly 14 files.

Compatibility: D3L Tool 2.7.0 ↔ glass.sqdaitools ≥ 0.1.0 (API changes are announced in both release notes). GLASS data are produced by the GLASS product team (glass.hku.hk) — please follow their citation policy.

Time series

Open the right-edge time-series drawer, pick a product and layer, enter latitude/longitude and a date range, and hit “Plot series” — the pixel's values are sampled over NASA OPeNDAP (LAADS for MODIS/VIIRS, GES DISC for MERRA-2, reusing your sign-in) and drawn as a line chart, exportable to CSV with one click; “Clear chart” resets the plot. For large-scale analysis, download the raw files instead.

MCP server & AI assistant

d3ltool-mcp exposes 13 tools (search, tile math, time- series sampling, in-app plotting, download control, browser-handoff export) over stdio for Claude Desktop, Claude Code, Codex, Cursor, OpenCode, Qoder, ZCode and any MCP client; the right-edge AI Assistant drawer in the GUI drives the same tools with plain language. Wiring examples live in the integration guide.

Menus

The “Websites” menu links the project site, the author's blog and NASA portals; “RS resources” is a curated remote-sensing list; “Help” holds Donate and About. The UI language switches under “Language”.

From 2018 to today

D3L Tool was created by Dr. Dai in 2018 (v1.0), when the LAADS FTP servers were shut down and the official Python scripts were hard for non-programmers — v1 automated order downloads through browser automation. Eight years later, v2 is a complete rewrite on top of NASA's official earthaccess library and the CMR search service.

Dependencies & acknowledgments

Questions & feedback

• Usage problems → the user guide above
• Bugs and ideas → GitHub Issues
• Anything else → dsq1993qingge@163.com

D3L Tool of NASA Satellite uses an MIT License.