Interactive Visualizations

Category Dominance

This ranked bar chart highlights the extreme category imbalance in the EONET record. Wildfires are visually emphasized and annotated because they dominate the observed event volume.

Temporal Trends

The annual view combines raw counts with a smooth trend line to separate short-term fluctuations from broad movement in observation volume.

Use legend click and the chart controls to isolate Wildfires or non-wildfire categories and compare how the recent surge differs by hazard type.

Seasonality

Absolute Counts

This panel shows where absolute volume is concentrated by month and category.

Within-Category Normalized

This normalized view rescales each category to reveal seasonal shape even for smaller categories.

Monthly Curves by Category

Line-based monthly profiles complement the heatmap and make category-specific seasonality easier to trace.

Spatial Distribution

All Events

Clustered markers improve performance and make high-density regions easier to inspect.

Wildfires Only

Sea and Lake Ice Only

Volcanoes Only

Severe Storms Only

Predictive Benchmarking

The full multi-category target and the wildfire-only target are treated as parallel predictive tasks. This is intentional: model ranking can change when the target scope changes.

All four models are available in the current environment.

Category-level RMSE comparison shows that gains are uneven and depend strongly on hazard type.

Prediction Scope Comparison

scope	model	MAE	RMSE	R2
Full multi-category	Seasonal baseline	331.200	456.327	-0.298
Full multi-category	OLS	184.227	324.520	0.344
Full multi-category	GAM	195.166	324.258	0.345
Full multi-category	XGBoost	464.470	548.158	-0.872
Wildfire-only	Seasonal baseline	694.222	724.923	-0.958
Wildfire-only	OLS	494.885	571.294	-0.216
Wildfire-only	GAM	556.000	759.996	-1.152
Wildfire-only	XGBoost	787.509	831.735	-1.577

This module aligns with the written report interpretation: in the full-scope task, the best RMSE comes from GAM (324.258), while in the wildfire-only task, the best RMSE comes from OLS (571.294). Even with partial RMSE gains, predictive fit remains weak in absolute terms because best R2 stays negative in both scopes (0.345 full scope; -0.216 wildfire-only scope).

Wildfire-only Analysis

This section gives a direct wildfire-only entry point. It complements the wildfire-only prediction benchmark and makes the dominant-category dynamics easier to review without cross-category mixing.

Report

For full methodology, statistical testing, predictive benchmarking, and limitations, see the full report.