The application of remote sensing for irrigation and water resources management in the Aral Sea basin, Kazakhstan
The application of remote sensing for irrigation and water resources management in the Aral Sea basin, Kazakhstan
A time series of Landsat-7 images was analysed using ERDAS Imagine 8.4 software. The images were used to estimate the areas of various crop types. Six remote sensing techniques were applied: visual interpretation, unsupervised classification, Principal Component Analysis (PCA), supervised classification, NDVI and irrigation maps and knowledge-based classification. The effect of the following factors on land use classification when using the above techniques was examined:
- Single or time-series images: results differed significantly when single images were used, hence time-series images were essential for crop discrimination.
- Atmospheric interactions significantly affected the classification results and atmospheric correlation was necessary for comparing imaging data with ground truth data. Atmospheric interactions also reduced NDVI values.
- The number of classes used in an unsupervised classification significantly changed the cover areas of the classes.
- Thermal data were successfully used and became necessary for crop discrimination.
- Effects of background water and soil: background water had different effects to background soil, e.g. vegetation growing in water had lower near infrared reflectance NDVI than crops growing in soil.
Land use classification showed high sensitivity to the above factors. Results from more than one techniques provided better results than a single technique. A new knowledge-based technique was developed producing results similar to supervised classification. Ground truth data showed that vegetation could be discriminated based both on its radiometric and temperature characteristics, which depend on irrigation practice.
University of Southampton
Perdikou, Paraskevi Nicou
af68cb35-8061-4aa6-8d89-d40df0c5a115
2003
Perdikou, Paraskevi Nicou
af68cb35-8061-4aa6-8d89-d40df0c5a115
Perdikou, Paraskevi Nicou
(2003)
The application of remote sensing for irrigation and water resources management in the Aral Sea basin, Kazakhstan.
University of Southampton, Doctoral Thesis.
Record type:
Thesis
(Doctoral)
Abstract
A time series of Landsat-7 images was analysed using ERDAS Imagine 8.4 software. The images were used to estimate the areas of various crop types. Six remote sensing techniques were applied: visual interpretation, unsupervised classification, Principal Component Analysis (PCA), supervised classification, NDVI and irrigation maps and knowledge-based classification. The effect of the following factors on land use classification when using the above techniques was examined:
- Single or time-series images: results differed significantly when single images were used, hence time-series images were essential for crop discrimination.
- Atmospheric interactions significantly affected the classification results and atmospheric correlation was necessary for comparing imaging data with ground truth data. Atmospheric interactions also reduced NDVI values.
- The number of classes used in an unsupervised classification significantly changed the cover areas of the classes.
- Thermal data were successfully used and became necessary for crop discrimination.
- Effects of background water and soil: background water had different effects to background soil, e.g. vegetation growing in water had lower near infrared reflectance NDVI than crops growing in soil.
Land use classification showed high sensitivity to the above factors. Results from more than one techniques provided better results than a single technique. A new knowledge-based technique was developed producing results similar to supervised classification. Ground truth data showed that vegetation could be discriminated based both on its radiometric and temperature characteristics, which depend on irrigation practice.
Text
905291.pdf
- Version of Record
More information
Published date: 2003
Identifiers
Local EPrints ID: 465024
URI: http://eprints.soton.ac.uk/id/eprint/465024
PURE UUID: 870ee8a6-c50c-48fb-9909-3eeadfe20935
Catalogue record
Date deposited: 05 Jul 2022 00:17
Last modified: 16 Mar 2024 19:53
Export record
Contributors
Author:
Paraskevi Nicou Perdikou
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics