Derivation of land-cover change data and their assimilation in ecosystem models IB

You are here

Context and objectives

The general objective of the research was to develop new methodologies, and to advance and refine existing methodologies to allow for:
(i) a more realistic description of long-term processes of land-cover changes, based on a variety of data sources, and
(ii) a better integration of remote sensing data into ecosystem models in order to better address key issues on land-cover changes and their impacts.

Project outcome

Expected scientific results

Digital change detection routines were optimized. Subtle changes could be detected, primarily in a qualitative, but still detailed way, by means of a newly developed change detection routine. The quantitative change detection study demonstrated that the Maximum Likelihood and Mahalanobis classification procedures perform better in digital change detection. The need for detailed and accurate ground reference data was also proven.
We developed a method to increase the comparability between land cover maps coming from panchromatic aerial photographs and SPOT XS (multispectral) data by equalising their level of thematic content and spatial details.
The methodology was based on the hypotheses that: (1) map generalisation can improve the integration of data for change detection purpose, and (2) the spatial structure of a land cover map, as measured by a set of landscape metrics, is an indicator of the level of generalisation of this map. There are large interannual variations both in the strength and in the parameters of the relationship between integrated NDVI and biomass. Thus, in semi-arid regions, NDVI values integrated over the growing season is not a very robust proxy variable for biomass. From all the possible indicators of ecological conditions that were tested in this study (i.e. biomass, integrated NDVI, ecosystem resilience, rain use efficiency and floristic modification), the indicator that is most likely to reveal a continuous trend in land-cover modification at the scale of a decade, independent of interannual fluctuations in rainfall conditions, seems to be rain-use efficiency.
Measurement techniques using the Licor-2000 PCA were optimized, operationalized and standardized. Operational settings take the influences of (in)direct sunlight and of areas of interest into account. Optimal measurement setups to guarantee sufficient accuracy were set out using Monte Carlo simulation analysis. As such, an experimental sampling design was developed. Moreover, the errors generated in estimating LAI in a certain forest stand were quantified.
Combustion efficiency is lower for fragmented burnt areas compared to continuous burnt areas. It would thus be possible to decrease uncertainties on estimates of trace gaze emissions from fires by replacing fixed burning efficiency values in emission models by values that would vary in space and time, based on measures of the spatial pattern of burnt areas as detected by remote sensing. In dense forests, burning is strongly associated with land-cover changes while in savannahs, the occurrence of (mostly) early fires does not lead to land-cover change. Our results confirm recent findings concerning the human control on the timing of burning in savannahs. Early fires fragment the landscape and prevent the spatial diffusion of later damaging fires. Where no human settlements are present, late fires become more prevalent. Using burnt area rather than active fire data allowed to better analyse spatial associations between landscape attributes and burning events.
Project leader(s): UCL - Georges Lemaître Centre for Earth and Climate Research
Location: Region:
  • Central Africa,Zambia

Related publications: