Flood Mapping in the Mara Wetlands for Ecosystem Conservation
Mtamba, Joseph¹; Vekerdy, Zoltán²; Van der Velde, Rogier²; Mtalo, Felix¹; Ndomba, Preksedis^1
1University of Dar es Salaam, TANZANIA, UNITED REPUBLIC OF; ²University of Twente-ITC, NETHERLANDS

Wetland, one of significant components of regional ecosystem, has many functional services such as the regulation of climate, purification of environment, and balance-keeping of regional water (Wang, 2003). As a special ecosystem, it locates in the transition zone between land ecological and aquatic ecosystems, and is a suitable habitat for wild flora and fauna. Since late 1980s, wetland has been under the impact of human disturbance in the world. The human disturbance includes reclamation of wetland for agricultural practise, pollution etc. Now a lot of wetland has been lost in the world (Mitsch and Gosselink, 1993). This endangers the sustainable development of humankind. The Mara wetland at the outlet of the trans boundary Mara river basin catchment in Tanzania and Kenya has been on the risk of loss because of anthropogenic activities and poor agricultural management in upstream. As a prelude to the conservation of wetland, flow dynamics is essential component that requires to be understood and distribution at different flow volumes identified for efficient and sustainable management of this wetland.

Flood dynamics over specified time periods and quantification of affected areas often requires timely and synoptic data collection and analysis. Remote sensing techniques can be helpful in wetland studies, and in this paper an investigation, by using a time series of remotely sensed data, has been proposed to be carried out into wetland flooding pattern mapping on Mara wetland in northern part of Tanzania in Mara region. The study will show the usefulness of time series of satellite images and how they can help to undestand the waters dynamics. They help to define which areas were flooded, partially flooded or dry land during flood, moderate and wet seasons. This information can help to improve stratergies sustainable mamagement of wetlands. Maps of flooded areas at different dates can also be applied as control parameters in flood simulations.

A time series of medium resolution radar images (ENVISAT, RADARSAT-2) shall be analysed and three techniques of flood area detection will be evaluated: pixel-based water delineation, object-based classification and machine learning procedure. This paper shall examine and compares advantages and restrictions of these methods. It also validates water under-detection and over-detection on radar images, and examines specifics for their occurrence.