dc.contributor.author | Mwesigwa, Isaac | |
dc.date.accessioned | 2021-05-03T13:18:37Z | |
dc.date.available | 2021-05-03T13:18:37Z | |
dc.date.issued | 2021-05-03 | |
dc.identifier.citation | Mwesigwa, I (2020). Soil salinity mapping of Mubuku irrigation scheme using sentinel 1 imagery. Unpublished undergraduate dissertation. Makerere University, Kampala, Uganda. | en_US |
dc.identifier.uri | http://hdl.handle.net/20.500.12281/10518 | |
dc.description | A Project report submitted to the Department of Geomatics and Land Management in
partial fulfilment of the Requirements for the Award of the Degree of Bachelors of Science
in Land Surveying and Geomatics of Makerere University. | en_US |
dc.description.abstract | Soil degradation has adverse effects on the economic growth of countries and its divided into
salinization, desertification, pollution and many others. Soil salinization is the process by which
soluble salts accumulate in the root zones of soil negatively affecting agriculture. Due to its direct
impact on agriculture, it needs to be managed. The initial step in managing soil salinization is
providing accurate data on the areas affected by the salts and by what magnitude.
The project was aimed at modelling soil salinity of Mubuku Irrigation scheme using Sentinel 1
GLCM texture features in conjunction with the electric conductivity values of twenty soil samples
collected from Mubuku Irrigation scheme. Specifically, features that have the greatest significance
to soil salinity modelling were determined and then relationship between the Sentinel 1 derived
features and electric conductivity values also determined.
Soil samples were collected from Mubuku Irrigation Scheme together with their coordinates and
their electric conductivity values measured in the laboratory. Sentinel 1 imagery corresponding to
the same date when the samples were collected was also obtained. Image was pre-processed
through the following steps: application of the orbit file, speckle filtering, geometric and terrain
correction and many other steps preparing image for processing. GLCM texture features were then
generated from preprocessed image and then values corresponding to coordinates were extracted
and exported into a workbook.
Feature selection was carried out on the exported data to select the most significant features for
model building in R. Regression was carried out on the selected features to create models for the
selected features. Four models were generated with model1 having R2 0.701268, model2 having
R2 0.75100, model3 having R2 of 0.80939 and model4 having R2 of 0.742446. The results of the
models were used in band math to generate the final predictive maps for salinity of Mubuku
Irrigation Scheme. Recommendations for future studies include using other selection algorithms
like the genetic algorithm, use of more features for example Histogram textures and also use of
other regression techniques that map input data onto a higher dimension feature space especially
for nonlinear cases. | en_US |
dc.language.iso | en | en_US |
dc.subject | Soil salinity mapping | en_US |
dc.subject | Mubuku Irrigation Scheme | en_US |
dc.subject | Sentinel1 imagery | en_US |
dc.subject | Soil salinity | en_US |
dc.subject | Modelling | en_US |
dc.subject | Radar | en_US |
dc.title | Soil salinity mapping of Mubuku Irrigation Scheme using Sentinel 1 imagery | en_US |
dc.type | Thesis | en_US |