Burned Area Detection - Part II

Random Forest

• ML classifier that uses n number of decision trees to classify the observation into certain classes

Main Parameters

• Number of Trees (Ntree)
• Number of variables per split
• Bag Fraction

Varibles

• Landsat bands(Thermal band excluded)
• Elevation
• Slope
• GEMI
• NDVI
• NDMI
• MiRBI
• BAI
• BAIMs
• BAIMl

Hyper Parameterisation

• Number of Trees : 50,1000,50
• Number of variables per split :1,10,1
• Bag Fraction : .1,1,.1

Final Model

ntree - 550, split = .2, split = 4

Study Area

Results visual

Results CF

Parameter Importance

Discussion