Practical Part 1

Phenology assessment using original and synthetic optical RS data and RF-based fusion model

Objectives:

Develop skills to:

• Apply RF model on PlanetScope data to create synthetic S2 data and fill time-series gaps [Section 1_data_fusion]
• Build time-series of Sentinel-2 data and retrieve phenometrics (SoS, EoS, PoS) [Section 2_phenometrics]
• Compare retrieved LSP metrics to references [Section 3_visualisation_validation]

Practical part flow (content):

1. Planet-Scope_Sentinel-2 Data Fusion with Random Forest [1_Planet_Scope_to_Sentinel_2_rf_synth.ipynb]

   • Loading S2 and PS original NDVI data
   • Data fusion based on the Random-Forest model (RF)
   • RF-model validation (MSE, MAE, SSMI Structural Similarity Index)
   • Saving synthetic data

2. The LSP Retrieval Based on S2 and Fused Time Series [2_Phenometrics_pixel_out_13.ipynb]

   • Reading S2-based and Fused-based time series
   • Fitting the B-spline function and proving the results
   • Phenometric retrieval (SoS, PoS, EoS) with Median and Relative Amplitude methods

3. Visualisation and validation of the LSP metrics derived from S2 and Fused Time Series [3_Val_Vis.ipynb]

   • Comparison of LSP metrics with the reference DWD and HR-VPP on violin plots
   • Box-plots with the data density impact on phenometrics accuracy