Kriging Interpolation

[laflamev]

Tasks

Geostatistical Analysis:

Implement Kriging interpolation to estimate climate values at unobserved locations.

• Select Suitable Library: Choose a geostatistical library (e.g., scipy, PyKrige) for Kriging interpolation.
• Load and Prepare Data: Load the climate dataset containing observed values of precipitation, temperature, latitude, and longitude.
• Spatial Data Exploration: Visualize the spatial distribution of observed data points on a map to understand their geographic spread.
• Interpolation Setup: Define the variogram model (e.g., spherical, exponential) and other parameters necessary for Kriging interpolation.
• Split Data: Divide the dataset into training and testing sets, reserving a portion for model validation.
• Perform Kriging: Apply the chosen library's Kriging function to estimate climate values at unobserved locations based on the observed data.

Model Validation:

Evaluate the accuracy and performance of the Kriging model.

• Cross-Validation: Implement cross-validation by comparing the Kriged values to the observed values in the testing set.
• Calculate Prediction Errors: Compute error metrics (e.g., Root Mean Squared Error, Mean Absolute Error) to quantify the accuracy of the Kriging predictions.
• Generate Validation Plots: Create visualizations (e.g., scatter plots, histograms) comparing predicted and observed values to assess model performance.

Generate Kriged Climate Surfaces:

Create continuous climate surfaces for precipitation and temperature across British Columbia.

• Define Grid for Interpolation: Establish a spatial grid covering the entire British Columbia region to generate the continuous surfaces.
• Apply Kriging to Grid: Utilize the trained Kriging model to estimate climate values at each point within the grid.
• Create Visualizations: Generate maps displaying the Kriged climate surfaces for precipitation and temperature.
• Export Results: Save the Kriged climate surfaces as spatial datasets (e.g., raster files) for further analysis and visualization.

Resources

• Article & tutorial: https://towardsdatascience.com/kriging-the-french-temperatures-f0389ca908dd
• What is Kriging: https://gisgeography.com/kriging-interpolation-prediction/
• Article: https://www.publichealth.columbia.edu/research/population-health-methods/kriging-interpolation
• Methods https://pygis.io/docs/e_interpolation.html#kriging
• Example code: https://geostat-framework.readthedocs.io/projects/pykrige/en/latest/examples/04_krige_geometric.html
• ML in Python: https://pygis.io/docs/f_rs_ml_predict.html