ISSN: 2339-076X (p); 2502-2458 (e), Volume 6, Number 3 (April 2019):1719-1725 DOI:10.15243/jdmlm.2019.063.1719
www.jdmlm.ub.ac.id 1719 Research Article
Land changes detection on Rote Island using harmonic modelling method
Atriyon Julzarika
1,2*, Nanin Anggraini
1, Kayat
3, Mutiaraning Pertiwi
21 Indonesian National Institute of Aeronautics and Space (LAPAN), Indonesia
2 Geodesy Geomatics Engineering, Universitas Gadjah Mada (UGM), Indonesia
3 Ministry of Environment and Forestry (KLHK), Indonesia
*corresponding author: [email protected] Received 14 January 2019, Accepted 9 February 2019
Abstract: Rote Island is one of the islands in East Nusa Tenggara. In this island, land changes occur significantly. This land changes can be detected by Landsat images. These images are obtained from the big data engine. The big data engine used is the Google Earth Engine. This study aimed to detect land changes with harmonic modelling using multitemporal Landsat images from the big data engine. Harmonic modelling is used in monitoring changes in Normalized Difference Vegetation Index values in a multitemporal manner from Landsat images. Processing is done using the Geomatics approach. Land changes on Rote Island generally occur on coastal and savanna. Land changes on land generally have vertical deformation on its movement and horizontal on the savanna. The land changes accuracy result is 95% in 1,96σ. This method can be used for rapid mapping of land changes monitoring.
Keywords: big data engine, harmonic modelling, land changes, Rote Island
To cite this article: Julzarika, A., Anggraini, N., Kayat, and Pertiwi, M. 2019. Land changes detection on Rote Island using harmonic modelling method. J. Degrade. Min. Land Manage. 6(3): 1719-1725, DOI:
10.15243/jdmlm. 2019.063.1719.
Introduction
Remote sensing data has evolved from conventional to dynamic (Kumay, 2015). One example of this dynamic data is the big data engine.
Big data includes remote sensing optical and radar sensor data. Some examples of images contained in big data engines are Landsat, Sentinel, ALOS PALSAR, WorldView, Ikonos, GeoEye, QuickBird, MODIS, and others. In addition to the big data engine, there are also various thematic information and algorithm making according to user application needs. The big data engine is also facilitated in the initial processing, extraction of geobiophysical information, to the presentation of spatial information. This big data engine is dynamic and fast in its processing that depends on the internet and cloud data. One of the uses of this big data engine is in the field of inland water resources, especially lakes. Rote Island is one of the outermost regions in Indonesia. This island has
a lot of uniqueness when viewed from various aspects including geology, culture, ecology, and water resources. However, this uniqueness is not yet comprehensively utilized and explored because there are not many studies and researches conducted in this area. The Integration System Development Activity Through a Multidisciplinary Approach to Natural Resource Mapping and the Environment is expected to open up insight into the uniqueness and provide preliminary data for further management (Julzarika et al., 2018).
Natural Resources and the environment on Rote Island experienced insignificant changes.
Generally, changes are caused by steppe and savanna environments. In addition, it is also caused by vertical deformation or the appointment of Rote Island due to geological activity with the Australian plate degraded, both in quantity and quality. Geographical position and topographic shape cause Rote Island have very diverse
Journal of Degraded and Mining Lands Management 1720 biodiversity. Likewise, the geological structure of
Rote Island causes a variety of natural formations that are very interesting and have quite high tourism potential. Among the uniqueness found on Rote Island are a number of lakes scattered throughout the Rote Island region, from the west to the east. Some lakes on Rote Island have a unique, very high salinity content making some of these lakes as dead sea lakes (Nikodemus et al., 2004).
The condition of the saltwater lake also allows the development of various unique animal and plant species, thus increasing the overall biological richness of Rote Island (Zitierung et al., 2012).
Rote Island have dead sea lake areas. This area has more than 15 saltwater lakes. The saltwater lakes are like Lake Oemasapoka, Oeinalaen, and others. This area also has higher vertical deformation which causes the potential to form a new salty lake around the "Mulut Seribu"
area. Vertical deformation factor is one of the causes of land changes in this region. This area also has a variety of endemic flora and fauna and still lacks research exploration in this area (Julzarika et al., 2018).
Land changes as a consequence of increased demand for space have various forms. One way to eliminate land changes due to environmental degradation (Pratisto and Danoedoro, 2008). It is necessary to build and grow awareness and concern from the elements of the government, society and the business world so that they can participate in tackling the problem of environmental degradation in accordance with their respective capacities and capabilities (Buma et al., 2018) ; (Tang, et al., 2016). These factors affect the land changes on Rote Island. The grassland ecosystem on Rote Island is very vulnerable and affects environmental variables both in terms of geophysical-chemical, biological, socio-economic and cultural aspects as well as aspects of public health. Grassland ecosystems that have their own peculiarity from other regions in Indonesia and in the world because this grassland ecosystem is formed in dry climatic conditions with a hilly and wavy topography (Karimi and Bastiaanssen, 2015).
Common causes of land changes are found in various cases in the world. Economic development can be seen simply from the increased purchasing power of the people (Suroso and Susanto, 2006;
François et al., 2016). This increase in purchasing power was responded by increasing facilities, infrastructure, as well as things needed by the community (Priasty, 2014). The land changes discussed in this study are land changes on Rote Island. This study aims to detect land changes with harmonic modelling using multitemporal Landsat images from big data engine.
Material and Methods
This study uses data from the big data engine in the form of Landsat. Big data Engine is a platform for petabyte-scale scientific analysis and visualization of geospatial datasets, both for the public interest and for business and government users. Big Data Engine stores satellite imageries, manages it and makes it available for the first time for global scale data mining. Public data archives cover the images of the historical earth that has existed for more than forty years, and new images are collected every day. Big data Engine also provides Application Program Interface (APIs) in JavaScript and Python, as well as other tools, to enable analysis of large datasets (GEE, 2018).
Landsat is a satellite owned by the United States of Geological Survey (USGS) which has a mission to survey geological and natural resource mapping (USGS, 2018). Currently, Landsat has entered the eighth generation. In the big data engine, there are Landsat 1, Landsat 2, Landsat 3, Landsat 4, Landsat 5, Landsat 7, and Landsat 8 images. Landsat images began in 1972 and still continue today (Trisakti et al., 2016).
Geographically, Rote Island is located at 10°
25'- 11° 00' South Latitude and 121° 49'- 123° 26' East Longitude (Rote Ndao District, 2018). Rote is one of the islands in East Nusa Tenggara which has unique geological and biodiversity characteristics.
This geographical position causes Rote Island to be the southernmost island of the Republic of Indonesia. Rote Island itself consists of at least 94 large and small islands that are included in the area of Rote Ndao Regency, see figure 1 (Rote Ndao Regency, 2018).
Figure 1. Rote Island
The method used in this study is the Harmonic Modeling method. The Geomatics approach is used to land changes detection. Geomatics approach is used to its processing in big data engine. Harmonic modelling method is the main method on its processing.
Journal of Degraded and Mining Lands Management 1721 Data in the form of satellite images is obtained by
the big data engine. Big data engine in this paper is Google Earth Engine (GEE). The data used is Landsat images. Landsat data began from 1972- present. Then all of the Landsat data is focused on Rote Island. All Landsat data is processed to monitor changes on Rote Island. The method used is harmonic modelling. The results obtained from this big data engine are land changes on Rote Island. Field validation needs to be done to determine the accuracy of information on land changes on Rote Island. The flow chart of this research can be seen in Figure 2.
Figure 2. Research flowchart
Land changes can be extracted using harmonic modelling methods. Harmonic modelling is a method in big data engine to detect the land changes. This method is Ordinary least squares (OLS) Regression (Julzarika, 2018). OLS regression is used to adjust separate Fourier series (eCognition, 2016). A Fourier series form based on that presented in (Artis et al., 2007) used in the analysis. Each time series of data is estimated as trigonometric polynomials,
....(1) Where coefficients a0, b1, ..., bn are specified in formula (2). Parameter t is the timestamp
composite value, n is the cycle length, and m is the polynomial sequence and is equal to the number of harmonics in the approximation. Timestamp when Landsat images are stored as milliseconds since the beginning of time (January 1, 1970) and converted to ephemeris day. Value of 365,242,181 days of ephemeris per tropical year is used as the duration of the annual cycle. Regression models with one to four harmonics (i.e. the first sequence to the fourth sequence of the Fourier series) are tested to determine the shape of the model that gives the best match to the data without introducing false oscillations recorded by Hermance (2007) and Bradley et al. (2007).
This harmonic process is only used for evaluation of harmonic regression models, and not as additional data.
… (2) where:
DOY : Day of the Year
GB : the beginning of image acquisition with certain assumptions
GE : the end of image acquisition with certain assumptions
t : the relative time (less dimension) at the time of acquisition
n : the total number of harmonics j : the number of harmonic terms a0, bj,
cj : are harmonic coefficients that are calculated using the standard
• a small f is used rather than F ∗ only to distinguish the DOY function from the function of the normalization time t;
• modM means modulo M operation, which makes sense for areas where GB is larger than GE (for example, in the Southern Hemisphere); M = 365 in our case because 365 is the longest time interval between two different days in a year.
Results and Discussion
Landsat images on Rote Island used are already corrected in geometry and radiometry. The geometry of Landsat images is more stable between images because the sensor is fixed.
Radiometric from big data (Landsat) is in Surface Reflectance (SR).
The next process is making thematic maps. It is created vegetation index maps from multitemporal and multi data from Landsat in big
Journal of Degraded and Mining Lands Management 1722 data engine. This vegetation index map is created
by the Normalized Difference Vegetation Index (NDVI) method. This aims to determine changes in vegetation and non-vegetation in the study area.
This condition will facilitate the harmonic modelling process so that it can know the area that is undergoing change. The harmonic modelling used in this big data (Landsat) is degree 1, see Figure 3. This Harmonic modelling use equation 1 and equation 2. Harmonic modelling is useful in the detection of land changes from multi-temporal and multi data. Degree 1 from harmonic modelling
is used because representing water, land, and dynamic land mathematically.
Checking water and land boundaries need to be done first. Then the land changes checking is carried out in areas that have varying NDVI values.
If the NDVI value does not vary, the objects in the field are relatively fixed, whereas if the NDVI value changes drastically then the region experiences drastic land changes. Every NDVI that experiences a change in value on temporal data is then fitted with a value. The geometry of this stable Landsat images will affect the fitted NDVI.
Figure 3. Harmonic modelling using Landsat multitemporal in land area
Checking is done simply by checking the land near the edge of Rote Island and the water near the edge of Rote Island. The boundary remains in the NDVI value (harmonic modelling) close to 0. In this big data (Landsat), the island fixed boundary is more suitable to use level 1 harmonic modelling. This can be proven by land and water graphs used in harmonic modelling of lake boundary determination.
This land graph aims to find out that the location of the point is on land. This is marked from a graph pattern based on harmonic modelling degrees 1. These restrictions will be separated on the graph for land and water. The fixed boundary of the island lies in black (image). The land is separated with water firmly according to the red colour chart.
The land is separated with water firmly according to the red colour chart. Other checks are also needed on the water graph, namely the location of the point in the water adjacent to the land, see figure 4. In this water graph analysis also uses big data (Landsat) in 1972-2018. The data has
been made geometric and radiometric corrections.
Harmonic modelling uses degrees 1. The fixed boundary of the lake and sea lies in black (image).
The land is separated with water firmly according to the red colour chart.
If we see a detail of the land changes images, it can be seen that there are some fundamental changes. Such changes such as the addition of new land and there have been low vegetation (light green), changes in vegetation height (dark green), changes in vegetation to settlements (purple), and white-beige on the newly raised land see Figure 5.
Land changes on Rote Island grow significantly; generally occur in the steppe and savanna regions. Settlement growth did not increase significantly. In addition, it is also indicated that there is an addition of new land due to vertical deformation, especially on Coastal Rote, see Figure 5 in a very light green colour. Overall the land changes dominate the coastal and savanna of Rote Island, see Figure 6. Land changes on Rote Island generally consist of vertical deformation on its movement and horizontal on the savanna.
Journal of Degraded and Mining Lands Management 1723 Figure 4. Harmonic modelling using Landsat multitemporal in water area
Figure 1. Vertical deformation on Coastal Rote
Figure 6. The sample area of land changes using harmonic modelling
Accuracy tests are of two types namely horizontal accuracy and vertical accuracy (Julzarika and Dewi, 2018). In this study using horizontal accuracy test. Accuracy testing is done by comparing the results of land changes with conditions in the field. The field survey was
conducted in August 2018. Geostatically, the test sample met tolerance. Locations that are not suitable occur in the savanna area when the survey was very dry. This savanna is generally in the form of syncline and anti-syncline, see Figure 7.
Journal of Degraded and Mining Lands Management 1724 Figure 7. Savanna as the main cause of land
changes on Rote Island
The method of testing accuracy is to compare the results of land changes from the big data engine with field data. This accuracy test uses a confidence level of 95% (1,96σ). The land changes accuracy result is 95%. The results of land changes generally occur in the Savanna area. This accuracy test is processed in the big data engine. This method can be used for rapid mapping of land changes monitoring.
Conclusion
Land changes on Rote Island can be detected using multitemporal Landsat images from big data engine. The land changes use harmonic modelling method in Geomatics approach. In this paper, we use Google Earth Engine for its big data engine.
The land changes dominate the coastal and savanna of Rote Island. Land changes on Rote land generally consist of vertical deformation on its movement and horizontal on the savanna. In addition, there were also changes in savanna and steppe in the dry and rainy seasons. Multitemporal Landsat data can be used for land changes. The land changes accuracy result is 95% in 1,96σ. This method can be used for rapid mapping of land changes monitoring.
Acknowledgement
The authors wish to thank Ministry of Research, Technology, and Higher Education, Universitas Gadjah Mada, LAPAN, LIPI, Ministry for Public Works and Public Housing, Ministry of Environment and Forestry, Ministry for Marine and Fisheries, local government, and local residents in supporting this research project.
Atriyon Julzarika is the main contributor to this paper.
References
Artis, M., Clavel, J.G., Hoffmann, M. and Nachane, D.
2007. Harmonic Regression Models: A Comparative Review with Applications. Institute for Empirical
Research in Economics University of Zurich.
Working Paper Series ISSN 1424-0459.
Bradley, B.A., Jacob, R.W., Hermance, J.F. and Mustard, J.F. 2007. A curve fitting procedure to derive inter-annual phenologies from time series of noisy satellite NDVI data. Remote Sensing of Environment 106(2): 137-145, doi: 10.1016/ j.rse.
2006.08.002.
Buma W.G., Lee, L.I. and Seo, J.Y. 2018. Recent Surface water extent of Lake Chad from multispectral sensors and GRACE. Sensors (Basel) 18(7): 2082, doi:10.3390/s18072082.
eCognition, 2016. Reference Book: Definiens Professional version 5.0.6.2. Definiens AG.
München, Germany.
François, J.P., Andrew, C. and Alan, N.G. 2016. High- resolution mapping of global surface water and its long-term changes. Nature 540: 418–422.
GEE. 2018. Google Earth Engine. Google. The United States of America.
Hermance, J.F. 2007. Stabilizing high-order, non- classical harmonic analysis of NDVI data for average annual models by damping model roughness. International Journal of Remote Sensing 28(12): 2801-2819, doi:10.1080/
01431160600967128.
Julzarika, A., Laksono, D.P., Subehi, L., Dewi, E.K., Kayat, Sofiyuddin, H.A. and Nugraha, M.F.I. 2018.
Comprehensive integration system of saltwater environment on Rote Island using a multidisciplinary approach. Journal of Degraded and Mining Lands Management 6(1): 1553-1567, doi: 10.15243/jdmlm. 2018.061.1553.
Julzarika, A. 2018. Mining land identification in Wetar Island using remote sensing data. Journal of Degraded and Mining Lands Management 6(1):
1513-1518, doi: 10.15243/jdmlm. 2018.061.1513.
Julzarika, A. and Dewi, E.K. 2018. ALOS PALSAR DTM vertical accuracy test for DGNSS-Altimeter combination measurements. Jurnal of Penginderaan Jauh dan Pengolahan Data Citra Digital 15(1): 11-24, doi:10.30536/
j.pjpdcd.2018.v15.a2804
Karimi, P. and Bastiaanssen, W.G. 2015. Spatial evapotranspiration, rainfall and land use data in water accounting—Part 1: Review of the accuracy of the remote sensing data. Hydrology and Earth System Sciences19: 507–532.
Kumay, D.U. 2015. Remote sensing platforms and sensor. NBKRIST Vidyanagar. India.
Nikodemus, P.N., Susanto, S. and Sudira, P. 2004. Water resource prediction on a small island: a case study on Rote Island, East Nusa Tenggara. Jurnal Manusia dan Lingkungan 11(2):55-63 (in Indonesian).
Pratisto, A. and Danoedoro, P. 2008. The impact of land use changes against flood discharge and flood hazard (case study in Gesing Watershed, Purworejo, based on Landsat TM and ASTER VNIR image.
Proceedings of the Annual Scientific Meeting of the Indonesian Remote Sensing Society (PIT-MAPIN), Bandung.
Priasty, E.W. 2014. Analysis of the impact of changes in land use of watersheds in North Bengkulu Regency.
Jurnal of Bengkulu Mandiri 4 January 2014 (in
Journal of Degraded and Mining Lands Management 1725 Indonesian).
Rote Ndao Regency, 2018. Rote Ndao Regency Official Website. Retrieved June 9, 2018, from http://.rotendaokab.go.id/geografis/. Rote Ndao Regency (in Indonesian).
Suroso and Susanto H.A. 2006. Effect of changes in land use against the flood discharge of the Banjaran River basin. Jurnal Teknik Sipil 3(2):75-80 (in Indonesian).
Tang, G., Long, D. and Hong, Y. 2016. Systematic anomalies over inland water bodies of High Mountain Asia in TRMM precipitation estimates:
No longer a problem for the GPM era? IEEE Geoscience and Remote Sensing Letters 13: 1762–
1766.
Trisakti, B., Julzarika, A. and Dewi, E.K. 2016.
Utilization of Standards for the Use of Remote Sensing Data for Monitoring Land Changes. Report activities on remote sensing applications centre, LAPAN, Jakarta (in Indonesian).
United States of Geological Survey (USGS). 2018.
Landsat Mission. USGS. https://landsat.usgs.gov/.
Retrieved on July 2018. Amerika Serikat.
Zitierung, E., Wright, C. and Lewis, B. 2012. On the edge of crisis: contending perspectives on development, tourism, and community participation on Rote Island, Indonesia. ASEAS-Österreichische Zeitschrift Für Südostasienwissenschaften 5(1):
102–127, doi:10.4232/10.ASEAS-5.1-6.
