An exploratory GIS-based method to identify and characterise landscapes with an elevated epidemiological risk of Rhodesian human African trypanosomiasis

Table 2 Image features used for level 1 and level 2 classifications

	Classes	Features	Image	Notes
Level 1 classification	Open water	Object mean band 4 (near infrared): threshold of 5.5%	April	Light in band 4 absorbed by water
	Built up & bare ground; wet vegetation types; dry vegetation types	Object mean band 3 (red)	April	Vegetation type discrimination
		Object mean band 2 (green)	November	Identification of healthy vegetation
		Object mean band 3	November	Vegetation type discrimination
		Object mean band 5 (mid infrared)	November	Vegetation and soil moisture content
		Object mean band 7 (mid infrared)	November	Identification of built up areas and bare ground
		Object length/width
		NDVI difference, January to November		Differentiation of vegetation types
Level 2 classification	Woodland & dense savannah; crops & open savannah	Object mean band 2	April	Related to healthy vegetation, biomass, plant type or vegetation moisture content
		Object mean band 3	April
		Object mean band 4	April
		Object mean band 5	April
		NDVI difference, January to April, January to November and April to November		Differentiation of vegetation types
	Seasonally flooding grassland; lake fringe swamps	Object mean band 4	November	Identification of vegetation and soil moisture content, biomass, plant vigour and water
		Object mean band 5	November

All classifications used a nearest neighbour algorithm with the exception of open water, which utilised a threshold [26, 27].
*Obtained using sub-classifications of the “dry vegetation types” class.
**Obtained using sub-classifications of the “wet vegetation types” class.

ISSN: 1471-2334