Monitoring the spatio-temporal dynamics of land cover and primary productivity in Tana river basin in Kenya by means of earth remote sensing (IRO-scholarship)
Description: This doctoral research project fits within a context of examination of possible relationships between the spatial and temporal distribution of land cover types, primary productivity and other river basin characteristics (soil patterns, topography) on the one hand and biogeochemical dynamics of the catchment’s aquatic systems on the other hand. The methodoloy will be developed with application to the Tana River basin in Kenya as a case study of an extensive tropical catchment with large contrasts in elevation, climate and vegetation.
Earth observation is a promising provider of spatio-temporal land cover data. Obviously, higher spatial resolutions of the land cover assessments will result in more reliable and precise correlations with the aquatic biogeochemistry. Whereas optical high resolution imagery outperforms low and medium resolution imagery in terms of the spatial and semantic detail which can be obtained through a multitemporal land cover classification, HR imagery is relatively costly. Moreover its availability at the appropriate time to take advantage of differential vegetative phenological development, is not guaranteed due to low temporal frequency aggravated by frequent cloud cover. Timeliness may be compromised by lack of suitable images and by important processing requirements of HR datasets. Medium and low resolution imagery is cheap. Due to large swath widths of the sensors, revisit time intervals are short allowing for the production of composite imagery as a solution for the cloud problem. Spatial resolution however is too low to distinguish individual land cover units like e.g. agricultural parcels. Pixels are said to be mixed.
Previous research has shown that usable data of the regional extent of various land cover types and their primary productivity, can be obtained by aggregation of spatially semi-explicit area estimates obtained through sub-pixel- or soft-classification of time series of low to medium resolution satellite imagery (MRSC). Artificial neural networks of the feed-forward back-propagation type seem to be the preferred tool in this sub-pixel classification approach. In an agricultural context, convincing results have been obtained for regions like Belgium in north-western Europe where monocultures of crops dominate, crop calendars are quite crisp, agricultural parcels are relatively large, topography is flat to gentle and absolute ground truth data for training of the ANN are abundantly available.
The major objectives of this doctoral research project is therefore to test and improve the MRSC-approach for application over tropical catchments, where multi-cropping is practiced in partially pristine landscapes, annual crops and perennial plantations (e.g., oil palm, Jatropha curcas) may be mixed, where topography is pronounced and ground truth data is scarce. The Tana river basin presents an ideal and challenging case study to test and improve the MRSC approach, as it encompasses a wide range of contrasting settings in terms of e.g., altitude, rainfall, soil and vegetation, from high-altitude evergreen forests to low semi-arid coastal plains.
Research activities will encompass: (i) study of vegetative systems and vegetation phenology and of nature and timing of agricultural practices for the major crops in the area of study, (ii) comparative study of the published approaches to extract areas and primary productivity out of middle resolution remotely sensed imagery from sensors as Terra-Modis, Envisat-Meris and Spot-Vegetation, (iii) study of the spectral separability of the various land cover types and collection of ground truth data, (iv) test and adaptation of the MRSC-approach for area assessments, and (v) uncertainty analysis and quality assessment.
Key words: Food security, land cover, primary productivity, image classification, earth observation, geographic information science
Latest application date: 2009-12-31
Financing: iro-scholarship
Type of Position: scholarship
Duration of the Project : 4 years
Research group: Department of Earth and Environmental Sciences
Remarks: Prof.Dr.ir. Steven Bouillon is co-promoter of this PhD-research project
علاقه مندی ها (Bookmarks)