Image mining for drought monitoring in eastern Africa using Meteosat SEVIRI data Coco M. Rulinda *, Wietske Bijker, Alfred Stein International Institute for Geo-Information Science and Earth Observation (ITC), Department of Earth Observation Science, Hengelosestraat 99, P.O. Box 6, 7500 AA Enschede, The Netherlands 1. Introduction and background Drought is defined as an extended period of abnormally dry weather that causes water shortage and damage to vegetation. It is a creeping and recurrent natural phenomenon and its impacts, covering large areas, can last for weeks or months (Wilhite, 2005). The onset, duration and severity of droughts are often difficult to determine and their characteristics may vary significantly from one region to another. In systems reliant on rainfall as the sole source of moisture for crop or pasture growth, seasonal rainfall variability is inevitably mirrored in both highly variable produc- tion levels as well as in the risk-averse livelihoods (Cooper et al., 2008). Africa has a long history of rainfall fluctuations of varying lengths and intensities (Nicholson, 1994, 2000). At different spatial and temporal scales, studies showed different behavior of rainfall trends in Africa; while studies by Olsson et al. (2005) and Herman et al. (2005) showed an increase of rainfall and greenness in parts of the Sahel region, Swenson and Wahr (2009) showed a decrease of water shortage in eastern Africa between 2003 and 2008 where drought and famine situations were periodically reported (FEWSN, 2005c, 2006b). Drought has particularly negative impacts on agricultural production in the eastern African region, as most of agriculture is dependent on rainfall (Barron et al., 2003; Slegers, 2008; Thorton et al., 2009). In this study we focus on monitoring the impacts of drought on vegetation, referred to as vegetative drought, using a membership function applied to Meteosat Spinning Enhanced Visible and Infrared Imager (SEVIRI) data. 1.1. Vegetation index Satellite vegetation monitoring involves the exploitation of information from the red and near-infrared wavelengths combined into the Normalized Difference Vegetation Index (NDVI) (Tucker, 1979). NDVI is calculated as in Eq. (1): NDVI ¼ l NIR  l RED l NIR þ l RED (1) where l NIR and l RED are the spectral reflectance in the near infrared (0.75–1.1 mm) and red (0.4–0.7 mm) respectively. NDVI is the most commonly used vegetation index and has been shown to be related to vegetation vigor, percentage green cover and biomass (Myneni and Asrar, 1994; Anyamba and Tucker, 2003; Tucker and Stenseth, 2005). It is a non-linear function that varies between 1 and +1, and is undefined when both l NIR and l RED are zero. NDVI values for vegetated land areas generally range from approxi- mately 0.1 to 0.7, with values greater than 0.5 indicating dense vegetation. Values less than 0.1 indicate no vegetation but barren area, rock, sand or snow (Tucker, 1979). 1.2. Monitoring vegetative drought Monitoring vegetative drought usually requires a large amount of temporal data, and Remote Sensing (RS) technologies provide necessary means to collect these at regular intervals. NDVI is International Journal of Applied Earth Observation and Geoinformation 12S (2010) S63–S68 ARTICLE INFO Article history: Received 20 November 2008 Accepted 16 October 2009 Keywords: Image mining NDVI Vegetation monitoring Drought Meteosat SEVIRI Eastern Africa ABSTRACT We propose an image mining approach to monitor drought using Meteosat Spinning Enhanced Visible and InfraRed Imager (SEVIRI) image data. SEVIRI image data provide frequent Normalized Difference Vegetation Index (NDVI) time series which are important to assess the evolution of drought conditions. Vegetation condition is characterized in space by the deviation of the current NDVI observations at locations from their temporal mean values. In this paper we assume a gradual evolution of vegetation stress caused by drought and hence address this aspect with the use of a membership function applied to vegetation stress values to model drought. Our approach is implemented on subset image data of eastern Africa. Vegetated sites in a drought prone area of the region serve as an illustration using the drought spell at the end of 2005. This study shows that the use of a membership function allows capturing the gradual evolution of drought and can be used to model drought from observed vegetation conditions. ß 2009 Elsevier B.V. All rights reserved. * Corresponding author. Tel.: +31 53 4874559; fax: +31 53 4874400. E-mail address: rulinda14348@itc.nl (C.M. Rulinda). Contents lists available at ScienceDirect International Journal of Applied Earth Observation and Geoinformation journal homepage: www.elsevier.com/locate/jag 0303-2434/$ – see front matter ß 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.jag.2009.10.008