Browsing by Author "Ismail, Riyad Abdool Hak."

Now showing 1 - 5 of 5

An assessment of land cover change patterns using remote sensing : a case study of Dube and Esikhawini, KwaZulu-Natal, South Africa.
(2012) Bassa, Zaakirah.; Bob, Urmilla.; Ismail, Riyad Abdool Hak.
During the past two centuries, land cover has been changing at an alarming rate in space and time and it is humans who have emerged as the dominant driver of change in the environment, resulting in changes of extraordinary magnitudes. Most of these changes occur due to demands placed on the land by the ever-increasing human population and their need for more land for both settlement and food production. Many researchers underscore the importance of recognizing and studying past land-use and land cover changes as the legacies of these changes continue to play a major role in ecosystem structure and function. The objectives of this study were to determine the extent of land cover changes between 1992 and 2008 in the study areas, Esikhawini and Dube located in the uMhlathuze municipality, KwaZulu-Natal, and to both predict and address the implications of the extent of future changes likely to occur in the area by 2016. Three Landsat satellite images of the study area were acquired for the years, 1992, 2000 and 2008. These images were classified into nine classes representing the dominant land covers in the area. An image differencing change detection method was used to determine the extent of the changes which took place during the specified period. Thereafter, a Markov chain model was used to determine the likely distribution of the land cover classes by 2016. The results revealed that aside from Waterbodies and Settlements, the rest of the classes exhibited a great degree of change between 1992 and 2008, having class change values greater than 50%. With regards to the predicted change in the land cover classes, the future land cover change pattern appears to be similar to that observed between 1992 and 2008. The Settlements class will most likely emerge as the dominant land cover in the study area as many of the other classes are increasingly being replaced by this particular class. The overall accuracy of the classification method employed for this study was 79.58% and the results have provided a good overview of the location and extent of land cover changes in the area. It is therefore plausible to conclude that these techniques could be used at both local and regional scales to better inform land management practices and policies.
Examining the utility of the random ensemble and remotely sensed image data to predict Pinus patula forest age in KwaZulu-Natal, South Africa.
(2010) Dye, Michelle.; Mutanga, Onisimo.; Ismail, Riyad Abdool Hak.
The mapping of forest age is important for effective forest inventory as age is indicative of a number of plant physiological processes. Field survey techniques have traditionally been used to collect forest inventory data, but these methods are costly and time-consuming. Remote sensing offers an alternative which is time-effective and cost-effective and can cover large areas. The aim of this research was to assess the capabilities of multispectral and hyperspectral remotely sensed image data and the statistical method, random forest, for Pinus patula age prediction. The first section of this study used spatial and spectral data derived from multispectral QuickBird imagery to predict forest age. Five co-occurrence texture measures (variance, contrast, correlation, homogeneity, and dissimilarity) were calculated on QuickBird panchromatic imagery (0.6 m spatial resolution) using 12 moving window sizes. The spectral data was extracted from visible and near infrared (NIR) QuickBird imagery (2.4 m spatial resolution). Using the random forest ensemble, various methods of combining the spectral and texture variables were evaluated. The best model was achieved using backward variable selection which aims to find the fewest number of input bands while maintaining the highest predictive accuracy. Only five of the original 64 variables were used in the final model (R2 = 0.68). The second part of this study examined the utility of the random forest ensemble and AISA Eagle hyperspectral image data to predict P. patula age. Random forest was used to determine the optimal subset of hyperspectral bands that could predict P. patula age. Two sequential variable selection methods were tested: forward and backward variable selection. Although both methods resulted in the same root mean square error (3.097), the backward variable selection method was unable to significantly reduce the large hyperspectral dataset and selected 206 variables for the model. The forward variable selection method successfully reduced the large dataset to only nine optimal bands while maintaining the highest predictive accuracy from the hyperspectral dataset (R2 = 0.6). Overall, we concluded that (i) remotely sensed data can produce accurate models for P. patula age prediction, (ii) random forest is an effective tool for the combination of spectral and spatial multispectral data, (iii) random forest is an effective tool for variable selection of a high dimensional hyperspectral dataset, and (iv), although random forest has mainly been used as a classifier, it is also a very effective tool for prediction.
Modelling terrain roughness using LiDAR derived digital terrain model in eucalyptus plantation forests, in KwaZulu-Natal, South Africa.
(2017) Munsamy, Roxanne.; Gebreslasie, Michael Teweldemedhin.; Ismail, Riyad Abdool Hak.
South African commercial plantation forests are established primarily to meet both the local and global demands of industries that require direct raw materials such as pulpwood or timber. Consequently, the commercial forest industry in South Africa is held in high esteem as it makes up one of the largest economic forces within the country. For this reason, individuals responsible for implementing strategies pertaining to silvicultural and harvesting operations within commercial plantations require up to date and detailed multi-forest inventory datasets to ensure that optimal yields are guaranteed and that sites are well maintained. Despite this, various drawbacks within commercial plantations exist: steep slopes, high elevations, and other forms of topographic irregularities, can affect the productivity of the site and impact mechanical silvicultural and harvesting operations. In lieu of making more informed and efficient decision-making protocols, forest researchers are often tasked with implementing and utilising alternative technologies such as remote sensing to determine if specific methodologies can be used for gathering multi-forest inventory data that also incorporate terrain information. Light Detection and Ranging (LiDAR), a recent remote sensing technology, has demonstrated that it is highly robust and can lend itself towards providing highly accurate vertical forest structural attributes and horizontal topographic derivatives. This study employs the use of a LiDAR derived Digital Terrain Model (DTM) (1 m x 1 m spatial resolution) to create terrain indices that are representative of the horizontal features within the commercial forest sites of interest. In addition, a machine learning approach using a random forest (RF) ensemble classifier was adopted to determine how much of the variation in forest structural attributes: mean dominant height, mean height, pulpwood volumes and diameter at breast height can be attributed to terrain when using the LiDAR derived DTM terrain variables. The overall findings presented in this study are encouraging and show that a LiDAR derived DTM can be successfully used for creating highly accurate terrain indices and can be used for predicting variability within even-aged Eucalyptus forest structural attributes within commercial plantation forests in KwaZulu-Natal, South Africa, with an acceptable level of accuracy.
Remote sensing of forest health : the detection and mapping of Pinus patula trees infested by Sirex noctilio.
(2008) Ismail, Riyad Abdool Hak.; Mutanga, Onisimo.; Bob, Urmilla.
Sirex noctilio is causing considerable mortality in commercial pine forests in KwaZulu- Natal, South Africa. The ability to remotely detect S. noctilio infestations remains crucial for monitoring the spread of the wasp and for the effective deployment of suppression activities. This thesis advocates the development of techniques based on remote sensing technology to accurately detect and map S. noctilio infestations. To date, no research has examined the potential of remote sensing technologies for the detection and mapping of Pinus patula trees infested by S. noctilio. In the first part of this thesis, the focus was on whether high spatial resolution imagery could characterize S. noctilio induced stress in P. patula forests. Results showed that, the normalized difference vegetation index derived from high spatial resolution imagery has the potential to accurately detect and map the later stages of S. noctilio infestations. Additionally, operational guidelines for the optimal spatial resolutions that are suitable for detecting and mapping varying levels of sustained S. noctilio mortality were defined. Results showed that a pixel size of 2.3 m is recommended to detect high (11-15%) infestation levels, and a pixel size of 1.75 m is recommended for detecting low to medium infestation levels (1-10%). In the second part of this thesis, the focus was on the ability of high spectral resolution (hyperspectral) data to discriminate between healthy trees and the early stages of S. noctilio infestation. Results showed that specific wavelengths located in the visible and near infrared region have the greatest potential for discriminating between healthy trees and the early stages of S. noctilio infestation. The researcher also evaluated the robustness and accuracy of various machine learning algorithms in identifying spectral parameters that allowed for the successful detection of S. noctilio infestations. Results showed that the random forest algorithm simplified the process by identifying the minimum number of spectral parameters that provided the best overall accuracies. In the final part of this thesis spatial modelling techniques were used to proactively identify pine forests that are highly susceptible to S. noctilio infestations. For the first time the random forest algorithm was used in conjunction with geographic information systems for mapping pine forests that are susceptible to S. noctilio infestations. Overall, there is a high probability of S. noctilio infestation for the majority (63%) of pine forest plantations located in Mpumalanga, South Africa. Compared to previous studies, the random forest model identified highly susceptible pine forests at a more regional scale and provided an understanding of localized variations of environmental conditions in relation to the distribution of the wasps.
A statistical approach for modelling forest structural attributes using multispectral remote sensing data within a commercial forest plantation.
(2017) Reddy, Nicole.; Gebreslasie, Michael Teweldemedhin.; Ismail, Riyad Abdool Hak.
Abstract available in PDF file.