Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Master → Pays Bas → 2002 → Mapping of woody vegetation in arid zones : a multi - sensor analysis

International Institute for Geo-information Science & Earth Observation (ITC) 2002

Mapping of woody vegetation in arid zones : a multi - sensor analysis

Hernandez Riquene, A.

Titre : Mapping of woody vegetation in arid zones : a multi - sensor analysis

Auteur : Hernandez Riquene, A.

Etablissement de soutenance : International Institute for Geo-information Science & Earth Observation (ITC) Netherlands

Grade : Master of Science (MSc) Forestry for Sustainable Development 2002

Résumé
Botswana woody vegetation presents large spatial and temporal variability. Elevation, soil type and texture, and groundwater depth gradients are some of factors influencing this variation. The vegetation distribution analysis in the area is a challenge because changes in vegetation cover and structure can have impact on the groundwater recharge. The main objective of this study was to identify and map the different woody vegetation cover types in the Serowe area using LANDSAT TM, ASTER and IKONOS satellite data, based on floristic and structure approaches. Besides, their relationships with a selected number of environmental factors were analyzed. Vegetation mapping was mainly done using Landsat TM, ASTER and IKONOS images from year 2001. First of all, based on Landsat TM 2000 images and supervised classification a preliminary vegetation map was obtained because images from year 2001 were not available. Three sub-areas (10x10 km) representative of major landscape types (sandveld, escarpment and hardveld) were evaluated where vegetation cover characteristics, environmental data and other plot information were collected using stratified random sampling design. Relevés were clustered following two approaches : floristic (using TWINSPAN software) and Structure (using a system proposed by the Kenyan Soil Survey), and then Landsat TM, ASTER and IKONOS images were classified based on these two approaches. Comparison between approaches in each sensor and between sensors was done using confusion matrix, kappa statistic and ‘Z’ test. Besides, CANOCO software was used to find relationships between sensor bands and crown cover by means of Detrended Canonical Correspondence Analysis (DCCA) and between environmental variables and crown cover by means of Canonical Correspondence Analysis (CCA). Results suggest the presence of 4-plant communities in the place, grouped using floristic approach and 10 classes following and structure approach. Moreover, Landsat TM and ASTER sensors were feasible for vegetation mapping using this two approaches being floristic approach the best in ASTER and there was not significant difference when comparing Landsat approaches. However, ASTER sensor was better than Landsat TM using floristic and structure approaches at 95% confidence interval when comparing them. Landsat TM5, TM7, TM3 bands and ASTER AB4, AB3, AB5 were found adequate to vegetation classification in arid regions. On the other hand, soils types and soil texture gradients are strongly related to vegetation distribution and water table depth and elevation gradients had less effect on it. Landsat result suggested that this sensor is suitable for bushy areas discrimination and for ASTER the result indicated that it was suitable when mapping wooded areas. The understory brightness due to dry vegetation (no chlorophyll), strong soil signal of brighter soils, absorption power of black soils and also to the darkening effect produced by vegetation cover introduce noise in vegetation classification. The two sensors were significantly different at 95% confidence interval using floristic and structure approach. These results showed that use of ASTER satellite data gives more reliable vegetation maps than Landsat TM satellite data. On the other hand, soils types and soil texture gradients were found governing the vegetation distribution in the zone.

Version intégrale

Page publiée le 28 décembre 2016, mise à jour le 14 octobre 2018