Abstract
Remote sensing data provide unprecedented opportunities for detecting and monitoring forest disturbance and loss. Disturbance and loss have been successfully mapped where cleared land is of sucient extent to provide discrimination within an image. However, methodologies for mapping and monitoring forest degradation are still lacking, primarily because these features are small in size. In Tanzania, the size and extent of degradation is poorly understood, yet there has been an increase in its drivers, such as shifting cultivation, settlements, logging and charcoal burning of indigenous trees.This study aimed to establish the extent remote sensing can map forest cover and change for REDD+ monitoring in Tanzanian savanna woodlands. A forest baseline for Liwale in south-eastern Tanzania was derived using the Land Cover Classication System (LCCS) from a 2012 RapidEye image, and validated using LiDAR data from 2012. The baseline and a 2014 RapidEye image were then used to perform an object-based change detection, by rst identifying potential change
features by automatically thresholding the data using a method based on optimising the skewness and kurtosis for distribution of the class of interest, and then classifying, using the random forests algorithm, the true change features. The change results were validated using 2014 LiDAR data.
These methods were then scaled out to courser Landsat imagery. The study concluded that both high and low resolution optical RS data have great potential for forest monitoring of the Tanzanian savanna woodlands, even though Landsat does not provide the level of detail to accurately depict small-scale change. Components of a remote sensing-based monitoring system are proposed. However, it was noted that for national monitoring, an integration of both high and low resolution data was best. Being a deciduous environment, it was found that seasonality and persistent cloud cover in the region greatly limit the window for appropriate monitoring data. The developed methodology is robust, and thus can be scaled up to a broader national scale and across similar environments in Africa.
| Date of Award | 2016 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Pete Bunting (Supervisor), Andy Hardy (Supervisor), Sandra Brown (Supervisor) & Richard Lucas (Supervisor) |