Band 11 der Reihe Fernerkundung und angewandte Geoinformatik
Buch, Englisch, 134 Seiten, KART, Format (B × H): 176 mm x 250 mm, Gewicht: 300 g
ISBN: 978-3-944101-20-0
Verlag: Rhombos-Verlag
Wafa Mohamed Tahir Nori
Detection of land cover changes in El Rawashda forest, Sudan: A systematic comparison
Band 11 der Reihe „Fernerkundung und angewandte Geoinformatik“
Herausgegeben von Univ. Prof. Dr. habil. Elmar Csaplovics, Lehrstuhl Remote Sensing, FR Geowissenschaften, TU Dresden
134 Seiten, Format DIN B5, Zahlreiche Abbildungen, davon 34 farbig. Sprache: Englisch. Preis: 39,80 Euro. ISBN 978-3-944101-20-0. Rhombos-Verlag, Berlin 2014
About this book
This research evaluates the potential of remote sensing for monitoring forest cover change in El Rawashda forest, Sudan, using Landsat ETM and Terra ASTER imagery. This was accomplished by performing eight change detection algorithms. Firstly a simplified post-classification with only 4 forest classes, namely close forest, open forest, bare land and grass land, was used. A RGB-NDVI change detection strategy to detect major decrease or increase in forest vegetation was developed as well. This method was found to be more effective than NDVI image differencing as it distinguishes different change classes by different colour tones. The Tasseled Cap green layer (GTC) composite was proposed to detect the change in vegetation of the study area. This method performs better than RGB-NDVI. Change vector analysis (CVA) based on Tasseled Cap transformation (TCT) was also applied for detecting and characterizing land cover change. The calculated date to date change vectors contain useful information, both in their magnitude and their direction. A powerful tool for time series analysis is the Principal Components Analysis (PCA). This method was tested for change detection in the study area by two ways: Multitemporal PCA and Selective PCA. A recently proposed approach, the Multivariate Alteration Detection (MAD), in combination with a posterior Maximum Autocorrelation Factor Transformation (MAF) was used to demonstrate visualization of vegetation changes in the study area. As a final step a quantitative accuracy assessment at the level of change/no change pixels was performed. Among the various investigated methods of forest cover change analysis the highest accuracy was obtained using post-classification comparison based on supervised classification.
Autoren/Hrsg.
Weitere Infos & Material
Contents
1 Introduction 1
1.1 Background 1
1.1.1 Forest in Sudan 1
1.1.2 Remote Sensing for Ecosystem Management 3
1.2 Problem definition 4
1.3 Motivation 5
1.4 Current status of research 7
1.5 Objectives 9
1.6 Structure of the thesis 10
2 Review of forest change detection techniques 13
2.1 Introduction 13
2.1.1 Image acquisition 13
2.1.2 Pre-processing 14
2.2 Visual interpretation 14
2.3 Pixel-based methods 15
2.3.1 Single-date indices 15
2.3.1.1 Spectral reflectance 15
2.3.1.2 Vegetation indices 16
2.3.1.3 Principal Components 16
2.3.1.4 Tasseled cap components 17
2.3.1.5 Spectral Mixture Analysis 17
2.3.1.6 Classification 18
2.3.2 Multi-date transformation 18
2.3.2.1 Image differencing 18
2.3.2.2 Change vector analysis 19
2.3.2.3 Multitemporal principal component analysis 20
2.3.2.4 Multitemporal Kauth-Thomas transformation 21
2.3.2.5 The Multivariate Alteration Detection 21
2.3.2.6 Chi square transformation 22
2.3.3 Change detection algorithms 22
2.3.3.1 Post-classification comparison (PCC) 23
2.3.3.2 Thresholding 23
2.3.3.3 Classification 24
2.4 Object-based methods 24
2.4.1 GIS-data 25
2.4.2 Image segmentation 26
2.5 Change mapping 27
2.6 Method comparison 28
3 Study area and research data acquisition 29
3.1 Study area 29
3.1.1 Location of the study area 29
3.1.1.1 Topography 29
3.1.1.2 Geology 29
3.1.1.3 Soil 29
3.1.2 Physical Characteristics of El Rawashda Reserved Forest 31
3.1.2.1 Climate 31
3.1.2.2 Vegetation 32
3.1.3 Social and land-use characteristics 33
3.1.3.1 Population in Gedarif State with special
reference to El Rawashda 33
3.1.3.2 Land-use, land use changes and their influence on the forests 34
3.1.3.3 Socio-economic background of the nomad 35
3.1.4 Management Units 35
3.1.4.1 El Rawashda model I 35
3.1.4.2 El Rawashda Model II 36
3.2 Research data acquisition 36
3.2.1 Field sampling 36
3.2.2 Data collection for study area 37
3.2.2.1 Landsat Imagery 37
3.2.2.2 Aster Imagery 38
3.2.3 Image pre-processing 38
3.2.3.1 Atmospheric correction 38
3.2.3.2 Image to image registration 39
4 Methods of Change Detection 41
4.1 Descriptions of land cover classes 41
4.2 Classification of images using Maximum Likelihood Classifier 41
4.3 Accuracy assessment of classified images 43
4.4 Calculation of Vegetation Indices as independent variables 43
4.4.1 NDVI 44
4.4.2 NDVI-RGB 44
4.4.3 SAVI 44
4.4.4 TDVI 45
4.5 Index Differencing 45
4.6 Tasseled Cap Transformation 45
4.6.1 RGB-TCG 45
4.7 Change Vector Analysis 46
4.8 Mapping with Conventional and Selective Principal
Component Analysis 46
4.8.1 Multitemporal PCA: the conventional approach 46
4.8.2 Selective PCA 46
4.9 Change Detection by Multivariate Alteration Detection 47
4.10 Accuracy assessments for change detection techniques 47
4.11 Object-Oriented Classification of reference data by eCognition 48
5 Results 51
5.1 Classification and Change Detection Accuracy 51
5.2 Post Classification Change Detection 53
5.3 Change detection based on Vegetation Indices 58
5.3.1 Vegetation Indices differences 58
5.3.2 RGB-NDVI 59
5.4 Application of Tasseled Cap for change detection 61
5.4.1 Derivation of Tasseled Cap Transformation 62
5.4.2 Derivation of Greenness component for change detection 63
5.5 Change Vector Analysis: An Approach for Detecting Forest Changes 65
5.5.1 Analysis of the change image 2000/2003 65
5.5.2 Analysis of the change image 2003/2006 66
5.6 Mapping with Conventional and Selective Principal Component Analysis 66
5.6.1 Generation of PCA Components 66
5.6.2 Change detection with Multitemporal PCA: the conventional approach 68
5.6.3 Change detection with Selective PCA 72
5.7 Change Detection based on Multivariate Alteration Detection (MAD) 73
5.7.1 MAD Interpretation 74
5.7.2 Maximum autocorrelation factor (MAF) analysis 77
5.8 Accuracy assessments for change detection techniques 79
6 Discussion 85
7 Conclusions and outlook 93
7.1 Conclusions 93
7.2 Outlook 94
8 References 97
Editorial
From this station [Quaicha], however, we were entertained with a most magnificent sight. The mountains at a distance towards the banks of the Tacazze (Setit, border river between Eritrea and Ethiopia, called Tekezé in the eastern Sudan, where it joins the Atbarah south of Shuwak), all Debra Haria (mountain region in Tigray, northernmost province of Ethiopia), and the mountains towards Kuara (Quarra, former Ethiopian frontier province towards the Sudan), were in a violent bright flame of fire. The Arabs feed all their flocks upon the branches of trees; no beast in this country eats grass. When therefore the water is dried up, and they cannot longer stay, they set fire to the woods, and to the dry grass below it. The flame runs under the trees, scorches the leaves and new wood, without consuming the body of the tree. After the tropical rains begin, the vegetation immediately returns; the springs increase, the rivers run and the pools are filled with water. All sorts of verdure being now in the greatest luxuriancy, the Arabs revisit their former stations. This conflagration is performed at two seasons; the first by the Shangalla (Shanqella, ethnic groups of the westernmost parts of Ethiopia, with a Nilo-Saharan ethnic background) and hunters on the southern part of this woody country, begins in the month of October, on the return of the sun, the circumstances of which I have already mentioned; the latter which happens in March, and lasts all April, besides providing future sustenance for their flocks, is likewise intended to prevent, at least to diminish, the ravages of the fly; a plague of the most extraordinary kind, already described. (p.265)
Our journey [on the 18th of April 1772, towards Sennar] for the first seven hours was through a barren bare and sandy plain without finding a vestige of any living creature, without water and without grass. (p.333)
Bruce J (1813) Travels to discover the source of the Nile, in the years 1768, 1769, 1770, 1771, 1772 & 1773 by James Bruce of Kinnaird Esq., 3rd ed., vol.6. Constable, Edinburgh and Longman, London.
El Rawasha Forest is one of the most important relics of the former pasture woodland of the East African Sub-Saharan regions. Transhumants between the Butana region and the northwestern Ethiopian borderland followed livestock corridors three of which intersect at El Rawasha and Wad Kabo Forests. As rainfed agriculture supported by mechanisation increased, the forests became more and more the last resting places for pastoralists moving from north to south and vice versa. Traditionally during the months of June and July as well as of October and November the herds enter the forests for grazing. Forest products and fodder are demanded, but the quantity and quality of range species growing on the forest floor immediately after the rains cannot satisfy the needs of increasing sizes of herds and flocks, and thus browsing more and more affects the forest per se (El Dool 1994).
Nevertheless the increasing pressure of villagers on the forest resources has not to be underestimated. Rainfed agricultural schemes occupy all the land around the forests and the needs for charcoal have led to illegal felling especially of Talh (Acacia seyal) and subsequently to the expansion of bare land inside the forest. Unprotected soils are affected by insolation, are thus deteriorated and exposed to erosion (Gibreab 1997)
El Dool (1994) stated that the forests (in the early 1990s) „still represent the original natural cover characteristic of the area and comprise the last existing forest area in the dry savannah left above the 14th degree north latitude in the whole African region. The (two) forests are therefore od prime importance from the environmental and genetic biodiversity point of view and should be managed and protected on a sustained yield basis.“ (p.134)
Since the late eighties several attempts were undertaken by the Sudanese Forest National Coorporation (FNC) to cope with the multiple problems affecting forest resources by integrated management plans involving the participation of local people. The El Rawashda model 1 management system was set up by the FAO Fuel Wood Development for Energy Project from 1983 until 1989. A forest committee consisting of representatives of the local communities participated in the rehabilitation and the protection (against illegal felling) of the respective forest areas. The El Rawashda model 2 management system was developed by the forestry branch of the Agricultural Development Project for the Eastern Sudan (ADES) and resembles the FAO approach except the fact that the local people are not involved in the actions of controlled felling.
Anyhow, Salah El Shazali (1995) commented that „The current practice of so-called integrated land use planning, exemplified by its „model“ of Rawashda Wad Kabo forests in Gedarif State, seems to be little more than a mockery of the concept.“ (p.133)
Therefore attempts to combat against the progressive degradation and deterioriation of the forest were not successful. The loss of high-quality forest areas as well as the continuous diminution of the diversity of the forest floor with special reagrd to palatable pasture species by overgrazing as well as by the subsequent impact on trees by browsing – which was estimated appr. 70% of the green biomass consumption by livestock in the early 1990s (Gaiballa 1992) – continued to decrease the ecological (biodiversity) and economical (biomass) quality of the forest reserve. Recent studies on the dynamics of change inside the forest need both a sound methodology of spatio-temporal as well as of balanced socio-ecological analysis. The assessment of spatial dimensions of change depends on full-coverage monitoring of patterns of change and thus on the integration of remote sensing land use and land cover change (LUCC) detection. Few projects have up to now focused on the assessment of gradients of change by the problem-oriented application of earth observation data collected by remote sensing satellite systems.
The PhD research of Wafa Nori contributes to urgently needed scientific action to close the gap between local to sub-regional studies mainly based on socio-economic inventories - with limited significance concerning the regional spatial dimension - and medium-scale to small-scale mapping of land use and land cover (change) based on operational earth observation data with low information value concerning the local to regional impact patterns which are caused by heterogeneous interrelations between different demands of different ethnic groups.
In general a sound (spatial) SWOT analysis of El Rawashda Forest Reserve depends on the application of the concept of regionalisation and thus on the integration of LUCC investigations of not only the focus area (El Rawashda) but also of the surrounding subregions, explicitely of the rainfed agricultural schemes and their development, of the historical distribution and routes of livestock corridors (analysed by aerial photography of the 1950s and later) as well as of the interrelations with the Butana region and the degradation processes which occured there during the last 50 years, thus successively changing the oscillations of transhumance towards the south.
The PhD thesis of Wafa Nori deals with an in-depth analysis of the reliability of spaceborne remote sensing earth observation for multi-temporal assessment of dynamics of land use and land cover change (LUCC) of the El Rawashda Forest Reserve with specific regard to a comparative research into the accuracy, thus effectiveness and apporpriateness of different methods of change detection based on time series of satellite imagery. The selected time frame covers the period 2000-2003-2006, which is a crucial period for monitoring and assessing the degree of impact of several management projects coordinated by the Sudanese FNC. The aim of the research is a methodological analysis of remote sensing image classification approaches on the one hand and the application-oriented analysis of change towards the evaluation of influences of management regulations on the status of conservation and the sustainable use of the forest on the other hand. In-depth research into the reliability of pixel-based comparative multi-temporal image analysis of forest cover change is provided. The application of Tasselled Cap Transformation (TCT), especially of RGB synthesis of the greenness component, and of TCT-based Change Vector Analysis represent a valuable baseline – as does the Multivariate Alteration Detection together with Maximum Autocorrelation Factor Analysis (MAD/MAF components) - for further research and application-oriented adaptation towards operational regional monitoring. Remarkably both visual on-screen interpretation of specific RGB- visualisation together with subsequent visual segmentation as well as traditional post classification comparison show an undiminished potential for in-depth change detection assessment with specific regard to class-to-class change analysis.
Evidently spatial monitoring and assessment of forest cover change in El Rawashda Forest remain fragmentary as long as the interdependencies between seasonal and annual land use and land cover changes in the forest both in space and time and changes in land use around the forest which are driven by the implementation of large-scale rainfed agricultural schemes and subsequently by an increase of local population and thus by increasing demand of (forest) resources are not thoroughly investigated. North-south migration routes of transumhumants and related livestock corridors are deteriorated due to the closure of former migration routes which is caused by a tremendous increase of agricultural land at the cost of grazing land during the last five decades. Changes in cycles of migration are driven by increasing impacts of local to regional degradation and desertification as a result of interrelated effects of drougth and overexploitation (Crummey 2005, UNDP 2006).
The PhD thesis of Wafa Nori represents an important contribution to the improvement and urgently needed further development of regional monitoring and assessment of anthropozooic impact patterns in El Rawashda Forest Reserve. Low cost and easy-to-handle appropriate technologies of multi-temporal remote sensing image analysis implemented in regional officies responsible for land and forest management should pave the way for a significant improvement of measures of conservation and sustainable development as well as of management of forest resources with participation of the respective spectrum of stakeholders.
Prof. Dr. techn. habil. Elmar Csaplovics
References
Crummey D (2005) Land, Ethnicity, and Political Legitimacy in Eastern Sudan. Red Sea Press, Trenton NJ
El Dool YMA (1994) Integrated management plan for Rawashda and Wad Kabo forest reserves, Gedaref State, 1995-1999, Part 1, Description. FAO, Rome
Gaiballa AK (1992) Nomads and Forest Management. Sudan Fuel Wood Development for Energy Project, GCP/SUD/047/NET. FAO, Khartoum
Gibreab G (1997) People on the Edge in the Horn: Displacement, Land Use and the Environment in Gedaref Region, Sudan. Red Sea Press, Trenton NJ
Salah El Shazali I (1995) The Importance of Forest Resource Managment in Eastern Sudan: The Case of El Rawashda and Wad Kabo Forest reserves, in: OSSREA (ed) Managing Scarcity: Human Adaptation in East African Drylands, OSSREA, Addis Ababa, pp. 131-138
UNDP (ed) (2006) Share the Land or Part the Nation: The Patoral Land Tenure System in Sudan. UNDP, Khartoum
