Add an interactivities plot and add evaluation metrics

[YaoTingYao]

After running Model Assessment, three files are automatically saved locally: a PNG graphic, an HTML file, and a metrics.txt file. The HTML file contains an interactive Plotly-powered scatter plot with hover tooltips displaying the ID, Actual value, Predicted value, Agreement, and Difference for each data point. The plot also supports zooming, making it easy to explore high-density regions near the origin.

New statistics:

• Calculate MAE
  MAE=np.sum(distance_arr)/len(X)
  MAE_percent = (MAE / int(grid_area)) * 100

• agreement=intersection: Minimum of actual and predicted deforestation for each dot and add them all
  agree_arr = [min(X[i],Y[i]) for i in range(len(X))]
  agree = np.sum(agree_arr)

• union
  union_arr = [max(X[i],Y[i]) for i in range(len(X))]
  union = np.sum(union_arr)

• IoU
  iou = agree/union * 100

• Calculate the Difference
  Absolute value of predicted deforestation minus actual deforestation of each point and add them all
  diff_arr=[abs(X[i] - Y[i]) for i in range(len(X))]
  difference=np.sum(diff_arr)

1. Added IoU on the plot graphic file
2. Added IoU, MAE, Agreement and Difference to txt file
3. html file with interactive hover mode with ID, Actual, Predicted, Agreement and Difference for each point.

[Copilot]

Pull request overview

This PR enhances the Model Assessment outputs by adding an interactive Plotly scatter plot (HTML) with hover tooltips and by computing/writing additional evaluation metrics alongside the existing static plot output.

Changes:

• Add Plotly-based interactive scatter plot export to HTML with per-point hover details.
• Compute and persist additional metrics (MAE, IoU, Agreement, Difference) to a .txt file and display IoU on the static plot.
• Update GeoPandas overlay behavior and fix a masking bug in deforestation map creation.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 4 comments.

File	Description
model_evaluation.py	Adds Plotly HTML output and new metrics calculations; updates overlay behavior; fixes fmask array usage in a boolean mask.
UDef-ARP_conda_env.yml	Adds Plotly (and kaleido) to the conda environment dependencies.

Comments suppressed due to low confidence (1)

model_evaluation.py:338

• Setting keep_geom_type=False on the overlay can introduce non-area geometries (e.g., LineString/Point results when polygons only touch at boundaries). This function immediately computes area/max_area and divides by max_area, which can yield max_area==0 or skew percentcell, and later code may carry non-polygon geometries forward. Consider filtering thiessen_gdf to Polygon/MultiPolygon geometries (or extracting polygons from GeometryCollections) before computing area/percentcell, or keep keep_geom_type=True and handle GeometryCollection cases explicitly.

        thiessen_gdf = gpd.overlay(full_voronoi_grid, area_mask, how="intersection", keep_geom_type=False)
        # get area of each polygon
        thiessen_gdf["area"] = thiessen_gdf.area

        max_area = thiessen_gdf["area"].max()
        # using the area, calculate size of cell compared to max
        thiessen_gdf["percentcell"] = thiessen_gdf["area"] / max_area

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.