Characterising and quantifying vegetative drought in East Africa using fuzzy modelling and NDVI data C.M. Rulinda a, * , A. Dilo b , W. Bijker a , A. Stein a a Faculty of Geo-information Science and Earth Observation (ITC), University of Twente, The Netherlands b Faculty of Electrical Engineering, Mathematics and Computer Science, University of Twente, The Netherlands article info Article history: Received 4 April 2010 Received in revised form 8 September 2011 Accepted 15 November 2011 Available online xxx Keywords: Remote sensing data Drought Fuzzy set theory Measures Vague spatial objects abstract This study aims at improving the characterisation and quantification of vegetative drought as a vague spatial phenomenon. 10-day NOAA-AVHRR NDVI images of East Africa from September 2005 to April 2006 are used. Vegetative drought is characterised using a membership function to model the gradual transition between drought and non-drought classes. Measures are implemented to quantify the areas and vagueness of vegetative drought, and to visualise its evolution in space and time. Results show a severe drought, affecting more than 60% of the vegetated area in the region. Different degrees of vagueness are observed in time, independently of the change of the transition range; the vagueness remains higher at the onset than at the termination of drought, reflecting a more gradual movement to drought and a crisper return to normal conditions. The vagueness was the lowest at the drought peak. The mean-area is less vulnerable to the change of the transition range, compared to the core-area. A Crisp approach, using the median of the transition range as the threshold value, does not quantify the vagueness of vegetative drought. This method can also be used in other regions, or adapted to characterise and quantify other vague spatial phenomena. Ó 2011 Elsevier Ltd. All rights reserved. 1. Introduction In East Africa, the dynamics of vegetation strongly depend on water availability (Unganai and Kogan, 1998; Sannier et al., 1998; Nicholson et al., 2005). Drought episodes reported in the past years have periodically affected negatively the vegetation health in the region. For instance, during the end of the year 2005 and the beginning of the year 2006, the US Famine Early Warning System Network (FEWS NET), the Food Aid Organization (FAO), and local institutions have all reported a severe drought episode that resul- ted in crop failure, pasture degradation, and water shortage, leading to a severe food insecurity in most parts of East Africa. 1 In subsequent years, drought episodes were also reported mostly in parts of Kenya, Somalia and Ethiopia, resulting in food crises. It is in this context that timely and affordable ways to assess the impact of drought on vegetation, especially in regions with limited ground collected data, has been identified as an important factor to improve strategies of drought mitigation (Ambenje, 2000) and reduce the negative impact. Droughts are complex to quantify due to their creeping nature; they accumulate slowly and extend over long periods of time (Wilhite, 2005). The areas affected by drought evolve gradually as the symptoms of moisture stress in plants often develop slowly. In the process of modelling the impact of drought on vegetation, hereafter referred to as vegetative drought or drought, it is standard practice to use a crisp classification vegetation stress values. This approach does not reflect the vague nature of drought, and hence could hinder its detection. Frank (2008) suggested that a realistic model could be a more critical issue in improving decision making processes than the quality of input data. In this context we focus on reflecting the vague nature of vegetative drought using a more realistic model, hence improve its detection. The aim of this study is to characterise and quantify vegetative drought while taking into account its vagueness. We present measures to quantify vegetative drought at a regional scale and we use remote sensing data as input. The proposed measures to quantify vegetative drought are the total-area, the core-area, the mean-area and the vagueness, which are explained in Section 4. We track drought objects in the images and calculate their centroids, which are used to visualise their evolution in space and time. The rest of the paper is organised in the following way: Section 2 briefly presents previous work on vegetative drought monitoring; Section 3 presents the characteristics of the study area and the * Corresponding author. Tel.: þ31 0 53 4874559; fax: þ31 0 53 4874400. E-mail address: rulinda14348@itc.nl (C.M. Rulinda). 1 www.fews.net: the USAID FEWS NET Weather Hazards Impacts Assessment for Africa. Contents lists available at SciVerse ScienceDirect Journal of Arid Environments journal homepage: www.elsevier.com/locate/jaridenv 0140-1963/$ e see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.jaridenv.2011.11.016 Journal of Arid Environments xxx (2011) 1e10 Please cite this article in press as: Rulinda, C.M., et al., Characterising and quantifying vegetative drought in East Africa using fuzzy modelling and NDVI data, Journal of Arid Environments (2011), doi:10.1016/j.jaridenv.2011.11.016