PhD Defence Ms Coco Musaningabe Rulinda

Department of Earth Observation Science

Title of defence

Assessing vegetative drought from multi-temporal NDVI images

Summary

The impact of drought on vegetation, referred to as vegetative drought, is commonly monitored at a regional scale using satellite based vegetation indices such as the Normalized Difference Vegetation Index (NDVI). Current monitoring methods of this phenomenon, however, do not make use of the recent advancements in data types and operators that account for vagueness. This research aimed at improving methods to analyze vegetative drought at a regional scale by handling its conceptual vagueness. For that purpose, five studies were conducted at various scales in East Africa.

The first study aimed at modeling vegetative drought as a vague phenomenon. Four-day maximum composite NDVI images from the Meteosat Spinning Enhanced Visible and InfraRed Imager (SEVIRI) were used as input. A membership function was used to model vegetative drought. The study demonstrated that the use of the membership function, rather than the crisp one, quantified the gradual onset of a vegetative drought.

Figure 1:  Vegetative drought images of East Africa from the first dekad of October 1999 to the third dekad of January 2000. Vegetative drought membership values vary from 0 (no drought) to 100 (severe drought).

The second study aimed at validating the NDVI values obtained in the first study. Measures of chlorophyll content and percentage of vegetation cover were taken in four pixel-size sites of Bugesera, Rwanda, in May 2010. Results showed that a NDVI variation between pixels was influenced by the variation of both parameters combined.

In the third study, the aspect of assessing the vagueness and spatial extent of vegetative drought was addressed. A time series of 10-day NDVI images were used for the study area.  The drought period from September 2005 to April 2006 was considered. Measures were implemented and results showed that vegetative drought had different degrees of vagueness through time, independently of the width of the transition range. The proposed measures present advantages over the measures provided by a crisp approach as they can express both the degree of vagueness in the characterization of drought, and the spatial extent for different levels of certainty.

In the fourth study, the changes of the spatial extent previously quantified were modeled using Markov chains. The method was implemented using data from 1998 to 2008, for the two main rainy seasons. Three classes were defined and probability transition matrices were calculated. Results showed that Markov chains offer potentials to model the dynamics of the spatial extent of vegetative drought at a regional scale.

The fifth study proposed a method to improve the spatiotemporal analysis of vegetative drought by assessing its movement in space and time using an object-oriented approach. Vegetative drought (vd) and rainfall deficit (rd) objects were extracted and tracked. The objects were plotted in the space-time cube. Their speed, direction and relationships were assessed. This study further quantified the spatiotemporal relationship that existed between the two types of events.

Figure 2: Centroids of combined vd and rd objects during both growing seasons of 1999 (above and below respectively) in East Africa, represented in 3D and 2D using the space-time cube.

This PhD research offered a quantitative approach to analyze the dynamics of vegetative drought as a vague phenomenon on a regional scale. Results showed that this approach provided new qualitative and quantitative information about vegetative drought that improved our understanding of the phenomenon.

Biography

Coco G. Musaningabe Rulinda was born on February 9, 1979 in Kinshasa, Democratic Republic of Congo.  From 1998, she studied in the Faculty of Applied Sciences, University of Rwanda (Rwanda), where she received her BSc. degree with distinction in Computer Science.  After her studies she joined the Center for GIS and Remote Sensing (CGIS) of the National University of Rwanda, where she worked as the IT officer.  In 2005 she joined the department of Computer Science, Faculty of Applied Sciences, as an assistant lecturer and was also appointed as a trainer and research assistant at the CGIS.  In 2005 she was granted a NPT/RWA/071 scholarship to pursue her MSc degree in Geo-Information Science and Earth Observation at ITC, with a specialization in geoinformatics.  Her MSc thesis,  completed with distinction, was on modeling drought uncertainty from NDVI images using fuzzy set theory.   In September 2007, she  started a PhD research within the same context. Her research interests in remote sensing include uncertainty modeling and spatiotemporal change analysis.

Musaningabe Rulinda, C.G. (2012) Assessing vegetative drought from multi - temporal NDVI images. PhD thesis University of Twente; Summaries in Dutch and English. ITC Dissertation 204, ISBN: 978-90-6164-329-6.

   Full text

previous page
more events