The Effectiveness of Remote Sensing Techniques for Land use Classification in Kota Belud, Sabah
Lindah Roziani Jamru1, Mohamad Nazril Sharil1*, Adi Jafar1, Oliver Valentine Eboy1, Colonius Atang1, Mustapa Abdul Talib1
1Faculty of Social Sciences and Humanities, Universiti Malaysia Sabah, Kota Kinabalu, Sabah
*Corresponding Author: [email protected] Received: 30 March 2023 | Accepted: 15 May 2023 | Published: 1 June 2023
DOI:https://doi.org/10.55057/ajress.2023.5.2.10
__________________________________________________________________________________________
Abstract: The objective of this study was to evaluate the effectiveness of remote sensing techniques in classifying land use types in Kota Belud, Sabah. Two classification techniques, unsupervised and supervised, were employed to ensure high accuracy of the land use data. Six land use types were classified, including paddy fields, bare land, water bodies, forest, urban areas, and oil palm plantations. This study found that the accuracy of land use classification for the years 1990, 2000, 2010, and 2020 was 94%, 86%, 98.30%, and 91.60%, respectively.
These results suggest that the supervised classification method was more effective in accurately classifying land use types compared to the unsupervised classification method. The final maps generated by both classification methods showed significant changes in land use over time.
For instance, the expansion of paddy fields was observed in the study area, particularly from 1990 to 2020. The conversion of forest areas and bare land to paddy cultivation was in line with the National Agro-Food Policy 2021-2030, which aims to make Kota Belud a paddy granary. Overall, the study highlights the importance of remote sensing techniques in monitoring land use changes. The results provide valuable insights for policymakers and land use planners in making informed decisions to manage land resources sustainably.
Keywords: Land use, remote sensing, classification techniques, unsupervised, supervised _________________________________________________________________________
1. Introduction
Land use is a complex and dynamic system that can be managed through proper management (Zhao et al., 2017). Land use changes result from human influence on the landscape, which occurs due to significant modifications in the ecosystem (Bakr et al., 2010). The importance of land use studies lies in understanding the patterns of land use change and classifying land use categories to clearly observe changes and loss of original land (Nur Syabeera & Firuza, 2019).
Various studies have been conducted to understand the factors leading to land use changes, and land use patterns are seen as the result of long-term interaction between humans and the environment (Bičík et al., 2015). The main land use categories has been studied, such as urban area (Basavarajappa et al. (2014), wetland (Jamru et al, 2018; Jamru, Rahaman & Ismail, 2013) agricultural land (Vibhute & Gawali, 2013) forest areas (Razafindrakoto, et. al 2018), and other types of land. Changes in land use have raised various issues in terms of their patterns and processes, including spatial configuration and changes over time, which are influenced by different factors.
Classification analysis is a technique used in remote sensing and image processing to categorize pixels in an image based on their spectral characteristics. There are two types of classification:
Supervised and unsupervised. In supervised classification, the analyst provides training samples, which are areas of the image that are known to belong to specific land cover classes (e.g., agriculture, cities, wasteland, water). The maximum likelihood classifier method is then used to assign all other pixels in the image to one of the predefined classes based on their spectral similarity to the training samples.
Therefore, this study aims to use remote sensing techniques to classify the land use for the year 1990, 2000, 2010 and 2020. Land use data was obtained from Landsat TM 5 and Landsat TM 8 data products downloaded from the United States Geological Survey (USGS) website. Both supervised and unsupervised classification techniques were used to classify the land use type.
To evaluate the accuracy of classification, this study uses the equation by Congalton (1986).
2. Literature Review
Concept LULC
Land use refers to the manner in which humans utilize and allocate land for various purposes, encompassing the activities and practices associated with human settlements and territories.
This utilization of land is commonly recognized as the primary objective of how humans interact with and utilize the land. In the study conducted by Rawat and Kumar (2015), they define land cover as the physical characteristics and attributes of the Earth's surface. Land cover is an integral component influenced by cultural, societal, and economic factors, wherein human beings play a significant role in shaping, influencing, or even creating the surrounding environment. Wood and Porro (2000) further emphasize that land use and land cover are interconnected, as alterations in land cover result from the dynamic interactions between human activities and the natural environment. Quentin et al. (2006) extends the concept by introducing the notion of Land Use and Land Cover Change (LULCC), which refers to any modifications or transformations in the biological or chemical processes occurring on the Earth's surface. These changes encompass a broad range of activities, with agriculture often exhibiting significant variations in intensity over different periods, and the magnitude of these changes has been more pronounced in previous iterations.
Supervised Classifications
Supervised classification is a methodology used to categorize different types of information present in image data, including soil, vegetation types, water flow, and other relevant features.
This technique involves assigning training classes that represent the desired types of categories to be classified. By utilizing the available image data, the supervised classification method enables the division of the studied region or area into distinct classes based on the provided class information. There are several supervised classification techniques available, such as the maximum likelihood classifier, minimum distance classifier, parallelepiped classifier, and Mahalanobis classifier (Vibhute & Gawali, 2013). These methods offer various approaches to accurately classify and interpret the image data, facilitating the analysis of different features within the studied area.
Unsupervised Classifications
Unsupervised classification is a technique used to analyze a large number of pixels in an image without prior knowledge of their classification, and it automatically divides them into distinct groups based on similarities in their image values. Unlike supervised classification, this method does not rely on specific training data provided by an analyst. One common type of
unsupervised classification is the K-means algorithm, which groups pixels into clusters based on their similarity. Another approach is the ISODATA (Iterative Self-Organizing Data Analysis) method, which iteratively refines and adjusts the classification results based on statistical parameters (Vibhute & Gawali, 2013). These unsupervised classification techniques allow for the identification of natural groupings within the image data, aiding in the exploration and interpretation of unknown patterns or features present in the image.
Maximum Likehood Classifier
The maximum likelihood classifier is widely recognized as a prominent classification method used in remote sensing. It operates by assigning a pixel to a specific class based on the highest likelihood or probability of that pixel belonging to that particular class. This statistical decision criterion is particularly useful when dealing with overlapping pixel signatures, where multiple classes may exhibit similar characteristics. The classification process involves comparing the posterior probability of each class, considering all available class signatures, and assigning the pixel to the class with the highest likelihood. By analyzing the similarity between the image data and the signature of each class, the maximum likelihood classifier determines the most probable class for each pixel (Vibhute & Gawali, 2013). This approach facilitates accurate classification by leveraging the statistical likelihood of a pixel belonging to a specific class based on the image data and the class signatures.
Kohen Kappa Accuracy
Accuracy evaluation serves as the final step in the analysis of remote sensing data, allowing for the assessment of the reliability and correctness of the obtained results. This evaluation takes place after the interpretation and classification processes have been completed. Thematic or classification accuracy specifically focuses on evaluating the accuracy of thematic maps or classified photographs. It entails determining the correspondence between the assigned class labels and the actual or "true" classes, which are often observed and verified in the field during surveys. For example, a class labeled as "water" on a categorized image or map should actually represent groundwater, aligning with the true characteristics of the area (Anand, 2017).
The term "accuracy" is used in various contexts, particularly in relation to remote sensing data.
The value and reliability of the information derived from remotely sensed data are assessed based on the interpretation and accuracy of the results. Evaluation can be conducted qualitatively or quantitatively. Qualitative evaluations involve comparing the visual representation on the map or image with what is actually observed to assess if it appears correct or accurate. On the other hand, quantitative evaluations aim to identify and quantify inaccuracies in the map through a comparison between the map data and ground truth information, which is considered to be completely accurate and reliable (Anand, 2017). These evaluations play a crucial role in determining the trustworthiness and usability of the remote sensing data and derived maps.
3. Methodology
The United States Geological Survey (USGS) website was used to get the Landsat TM 5 and Landsat TM 8 data products used in this investigation. Data was collected for the years 1990, 2000, 2010, and 2020, and it was utilised to examine how the study area's land cover changed over time as well as how quickly paddy farming increased and decreased. ArcGIS 10.8 software was used to convert the received GeoTiff-formatted data into raster data. Details of the remote sensing data used in this study are displayed in Table 1.
Table 1: Details of Remote Sensing data
Data availability Date Source Format (band)
Landsat 5 Satellite Image
1990 2000 2010
USGS 7 band
Landsat 8 Satellite Image 2020 USGS 11 band
Figure 1 presents an overview of the data processing steps utilized in this study. The analysis techniques employed include Radiometric Correction, Mosaic, Cloud Masking, Classification, and Vector Processing. These techniques are sequentially applied to the data, encompassing the entire process from the initial data acquisition to the generation of the results.
Figure 1: Research methodology framework
4. Result & Discussions
Figures 2, 3, 4, and 5 depict the results of land use classification using unsupervised and supervised techniques for the years 1990, 2000, 2010, and 2020, respectively. The classification identifies six types of land use, namely paddy fields, bare land, water bodies, forests, urban areas, and oil palm plantations, as shown in Table 3.
The accuracy of the land use classification for the respective years was found to be 94% (1990), 86% (2000), 98.30% (2010), and 91.60% (2020) as presented in Table 2. These findings indicate that the supervised classification method outperforms the unsupervised classification method in accurately categorizing land use types.
The resulting maps generated by both classification methods reveal notable changes in land use patterns over time. Particularly, there is a noticeable expansion of paddy fields in the study area, especially between 1990 and 2020. This conversion of forested areas and vacant land to rice cultivation aligns with the objectives outlined in the National Agro Food Policy 2021- 2030, aiming to transform Kota Belud into a rice granary.
These findings emphasize the effectiveness of the supervised classification method in accurately assessing and monitoring land use changes over time, providing valuable insights for land management and policy-making in the region.
Figure 2: Landuse classification 1990 using (A) unsupervised (B) supervised techniques
Figure 3: Landuse classification 2000 using (A) unsupervised (B) supervised techniques
Figure 4: Landuse classification 2010 using (A) unsupervised (B) supervised techniques
Figure 5: Landuse classification 2020 using (A) unsupervised (B) supervised techniques
Accuracy assessment
The accuracy of the classification of land use in the research region is evaluated using the accuracy assessment method. This will guarantee that the classification's findings are accurate and trustworthy.
This study employs Congalton's (1986) equation to assess the precision of the land use classification.
The accuracy evaluation findings for land use classification are displayed in Table 2.
Overall Accuracy = Total Number Of Correctly Classified Pixels (Diagonal) x 100 Total Number Of Refference Pixels
Table 2: Accuracy Assessment of land use classification
Satellite image Percentage of accuracy
1990 94%
2000 86%
2010 98.30%
2020 91.60%
Table 3: Land use classification of the study area 1990-2020 Area ha
Land Use 1990 2000 2010 2020
Paddy 4,329 10,917 7,935 12,564
Bare Land 17,887 13,619 15,800 10,151
Forest 91,393 85,426 80,788 77,848
Urban 7,008 8,208 8,636 9,535
Oil palm 1,092 1,660 2,088 5,589
5. Conclusion
In conclusion, supervised and unsupervised classifications serve the purpose of categorizing pixels in images, but they diverge in their approach to training samples and the generation and interpretation of classes. The output map derived from supervised classification tends to be more accurate and informative. This is attributed to the fact that supervised classification relies on pre-existing knowledge of the study area, enabling a more guided and precise classification process.
Acknowledgements
This paper was a part of the research grant SLB2268 “Application of GIS and remote sensing in the agricultural sector in developing abandoned land for paddy cultivation in Kota Belud, Sabah”. The author would like to take this opportunity to express a thousand thanks to the Research Management Centre (RMC), Universiti Malaysia Sabah (UMS) for covering the research and publication costs of this academic article.
References
Anand, A. (2017, January). (PDF) UNIT 14 ACCURACY ASSESSMENT. ResearchGate.
https://www.researchgate.net/publication/324943246_UNIT_14_ACCURACY_ASS ESSMENT
Bakr, N., Weindorf, D.C., Bahnassy, M.H., Marei, S.M., & El-Badawi, M.M. (2010).
Monitoring land cover changes in a newly reclaimed area of Egypt using multi- temporal Landsat data. Applied Geography, 30(4), 592-605.
Basavarajappa, H.T., Dinakar, S., & Manjunatha M.C. (2014). Analysis on land use/land cover classification around Mysuru and Chamarajanagara District, Karnataka, India, using IRS-1D PAN+ LISS-III satellite data. Journal Impact Factor, 5(11), 79-96
Bičík, I., Kupková, L., Jeleček, L., Kabrda, J., Štych, P., Janoušek, Z., & Winklerová, J. (2015).
Land use changes in the Czech Republic 1845-2010: Socio-economic driving forces.
Switzerland, Springer.
Jamru, L. R., & Rahaman, Z. A. (2018). Combination of spatial logistic regression and geographical information systems in modelling wetland changes in Setiu basin, Terengganu. In IOP Conference Series: Earth and Environmental Science (Vol. 169, No. 1, p. 012106). IOP Publishing.
Jamru, L. R. B., Rahaman, Z. A., & Ismail, W. R. (2013) Permodelan Perubahan Tanah Bencah Di Lembangan Sungai Setiu, Terengganu Menggunakan Logistik Regresi Dan Aplikasi Sistem Maklumat Geografi (GIS). International Journal of Environment, Society & Space (Volume 1, No. 1, 2013, Pages 75 to 97)
Mieja Razafindrakoto, Andry Andriamananjara, Tantely Razafimbelo, Jennifer Hewson, Riana H. Andrisoa, Julia P.G. Jones, Ilja van Meerveld, Alison Cameron, Ntsoa Ranaivoson, Nandrianina Ramifehiarivo, Nantenaina Ramboatiana, Ravo N.G. Razafinarivo, Tahiana Ramananantoandro, Andriambolantsoa Rasolohery, Marie P.
Razafimanantsoa, Christophe Jourdan, Laurent Saint-André, Gabrielle Rajoelison, Herintsitohaina Razakamanarivo (2018).Organic Carbon Stocks in all Pools Following Land Cover Change in the Rainforest of Madagascar,Soil Management and Climate Change,https://doi.org/10.1016/B978-0-12-812128-3.00003-3.
Nur Syabeera Begum Nasir Ahmad, Firuza Begham Mustafa. (2019). Analisis perubahan guna tanah Negeri Sembilan melalui aplikasi Sistem Maklumat Geografi (GIS).
GEOGRAFIA OnlineTM Malaysian Journal of Society and Space 15 issue 1 (113- 131). e-ISSN 2680-2491 https://doi.org/10.17576/geo-2019-1501-09.
Quentin FB, Jim C, Julia C, Carole H, Andrew S (2006). Drivers of Land use Change, Final report; matching opportunity to Motivation, ESAI Project 05116, Department of Sustainability and Environment and Primary Industries, Royal Melbouren Institute of Technology, Australia”
Russel G Congalton. (1986) “Remote Sensing Brief Accuracy Assessment: A User’s Perspective M.” Science Direct, vol. 52, no. 3, pp. 397–399.
Rawat SJ, Kumar M (2015). Monitoring Land Use Cover Change using Remote Sensing and GIS techniques: A Case Study of Hawalbagh Block, District Almora, Uttarakhand, India”. The Egyptian Journal of Remote Sensing and Space Sciences 18(1):77-84.
Vibhute, A. D., & Gawali, B. W. (2013). Analysis and Modeling of Agricultural Land use using Remote Sensing and Geographic Information System: a Review. International Journal of Engineering Research and Applications, 3(3), 81–91.
Wood CH, Porro R (Eds.) (2000). Deforestation and Land Use in the Amazon. Korea:
University Press of Florida.
Zhao, L.L., He, D.J., Tan, D.H., You, W.B., Hong, W., & Fan, X.C. (2017). Dynamic changes of land use in coastal zone of eastern Fujian from 2000 to 2014. Chinese Journal of Tropical Crops, 38(5), 803-810.
