geoglows-summit-example

A minimal TethysDash plugin that renders a 15-day ensemble streamflow forecast from the public GEOGloWS REST API as a Plotly figure. Designed as a workshop example showing how to build a tethysdash plugin from a Jupyter-notebook hydrology workflow without pulling in the full geoglows / geopandas / statsmodels stack.

What it shows

For a given GEOGloWS 9-digit river_id, the plugin renders:

• High-resolution deterministic forecast (orange line)
• Ensemble median and mean (blue lines)
• 25-75th percentile band (blue shaded area)
• Min-max envelope (light blue shaded area)

Data source: https://geoglows.ecmwf.int/api/v2/forecaststats/<river_id>/?format=json.

Frontend arguments

Argument	Type	Description
river_id	text	The 9-digit GEOGloWS river identifier (e.g. 610217883).

Install

In a fresh venv:

python3 -m venv .venv
source .venv/bin/activate
pip install -e .

This registers the plugin under intake.drivers as geoglows_forecast_viewer.

Quick smoke test

import intake
src = intake.open_geoglows_forecast_viewer(river_id="610217883")
figure = src.read()
print("traces:", len(figure["data"]))
print("title :", figure["layout"]["title"])

Expected output:

traces: 7
title : GEOGloWS Forecast - River 610217883

How tethysdash discovers it

TethysDash uses Intake's intake.source.registry. After pip install, the entry-point declared in pyproject.toml makes the plugin discoverable:

[project.entry-points."intake.drivers"]
geoglows_forecast_viewer = "geoglows_summit_example.forecast_viewer:GeoglowsForecastViewer"

The plugin subclasses intake.source.base.DataSource and exposes the visualization_* class attributes that tethysdash reads to populate its UI (matches the convention used by the bundled RFC streamflow plugin in plugins/RFC_plugins/rfc_plugins/river/river_streamflow.py). The read() method returns the figure dict tethysdash hands to its Plotly renderer:

{"data": [...plotly traces...], "layout": {...plotly layout...}}

Relationship to the GEOGloWS notebook workflow

This plugin is a tightly-scoped slice of the workflow in lib/tethys-agents/examples/initial_conditions_adjustment_q.py - specifically the fetch_geoglows_data → plot_forecast_with_context path, without the bias-correction, AR(1) initialization adjustment, or observed-discharge fetching. Adding those is an exercise: each new statistic becomes another trace on the same figure, or a sibling plugin in the same package.