Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Improving biome classification of tropical grasslands using fire frequency data

Nilsson, Ylva LU (2025) In Student thesis series INES NGEK01 20251
Dept of Physical Geography and Ecosystem Science
Abstract (Swedish)
Tropical grasslands are ecosystems of immense value, both ecologically and socially. Despite this, they are commonly misclassified as degraded forests, partly due to the large role of fire, rather than climate, in limiting the establishment of woody vegetation in those ecosystems. Misclassification can cause misguided management practices to be implemented, which explains the need for improved classification schemes for tropical grasslands. This study therefore investigates the possibility of improving biome classification of tropical grasslands by using both climate and fire frequency as input variables. To this end, two different classification maps for northern Africa were produced, using clustering with (1) only climate variables and... (More)
Tropical grasslands are ecosystems of immense value, both ecologically and socially. Despite this, they are commonly misclassified as degraded forests, partly due to the large role of fire, rather than climate, in limiting the establishment of woody vegetation in those ecosystems. Misclassification can cause misguided management practices to be implemented, which explains the need for improved classification schemes for tropical grasslands. This study therefore investigates the possibility of improving biome classification of tropical grasslands by using both climate and fire frequency as input variables. To this end, two different classification maps for northern Africa were produced, using clustering with (1) only climate variables and (2) climate and fire frequency variables as input. The classifications were then compared against MODIS land cover data to get an estimate of the accuracy for the different approaches. There was a clearly distinguishable difference between the two resulting maps, indicating that the inclusion of fire does have an impact on biome classification schemes. While both classifications had classes for which they performed better than the other, the one including the fire variable was significantly more accurate for the classes located at the boundary between tropical forests and grasslands. This is a promising result, indicating potential to further develop the method to achieve both better accuracy and global coverage. This could allow for better identification of tropical grasslands, which is an important step in order to properly manage those highly valuable ecosystems. (Less)
Please use this url to cite or link to this publication:
author
Nilsson, Ylva LU
supervisor
organization
course
NGEK01 20251
year
type
M2 - Bachelor Degree
subject
keywords
Tropical grasslands, Fire, Biomes, Alternative stable states, Forest, Savanna, Clustering
publication/series
Student thesis series INES
report number
701
language
English
id
9199216
date added to LUP
2025-06-16 09:31:28
date last changed
2025-06-16 09:31:28
@misc{9199216,
  abstract     = {{Tropical grasslands are ecosystems of immense value, both ecologically and socially. Despite this, they are commonly misclassified as degraded forests, partly due to the large role of fire, rather than climate, in limiting the establishment of woody vegetation in those ecosystems. Misclassification can cause misguided management practices to be implemented, which explains the need for improved classification schemes for tropical grasslands. This study therefore investigates the possibility of improving biome classification of tropical grasslands by using both climate and fire frequency as input variables. To this end, two different classification maps for northern Africa were produced, using clustering with (1) only climate variables and (2) climate and fire frequency variables as input. The classifications were then compared against MODIS land cover data to get an estimate of the accuracy for the different approaches. There was a clearly distinguishable difference between the two resulting maps, indicating that the inclusion of fire does have an impact on biome classification schemes. While both classifications had classes for which they performed better than the other, the one including the fire variable was significantly more accurate for the classes located at the boundary between tropical forests and grasslands. This is a promising result, indicating potential to further develop the method to achieve both better accuracy and global coverage. This could allow for better identification of tropical grasslands, which is an important step in order to properly manage those highly valuable ecosystems.}},
  author       = {{Nilsson, Ylva}},
  language     = {{eng}},
  note         = {{Student Paper}},
  series       = {{Student thesis series INES}},
  title        = {{Improving biome classification of tropical grasslands using fire frequency data}},
  year         = {{2025}},
}