Dynamics of shoreline changes in the coastal region of Sayung,
Indonesia
Ratna Sari Dewi
a,⇑, Wietske Bijker
b aGeospatial Information Agency (Badan Informasi Geospasial), Jl. Raya Jakarta-Bogor Km. 46, Cibinong, Bogor 16911, Indonesia
b
Faculty of Geo-Information Science and Earth Observation (ITC), University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands
a r t i c l e i n f o
Article history: Received 29 July 2018 Revised 10 August 2019 Accepted 15 September 2019 Available online xxxx Keywords: Shoreline Shoreline change Change detection Monitoring Multi-temporal imagesa b s t r a c t
Monitoring shoreline is important for planning and development in the coastal region. This study utilizes remote sensing and GIS (Geographic Information System) techniques to observe the dynamics of the shoreline in the coastal region of Sayung, Indonesia, from 1988 up to 2017. Multi-temporal remote sens-ing images from Thematic Mapper (TM), Advance Spaceborne Thermal Emission and Reflection Radiometer (ASTER) and Operational Land Imager and Thermal Infrared Sensor (OLI/TIRS) were used in fuzzy classification followed by thresholding to generate shorelines. Post classification change detection was implemented to analyse changes in shoreline position in the study area. From these changes, erosion and accretion areas along the coast were inferred. The results show a general trend of continuous shore-line changes as a result of increasing coastal inundation. Between 1988 and 2017, 25% of the area has changed from non-water to water and 5% of the area has changed from non-water to shoreline margin. Comparison with the land use/cover (LUC) maps showed that these changes were related to the changes from other crops and paddy field into fishpond and from fishpond into water body.
Ó 2019 National Authority for Remote Sensing and Space Sciences. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license ( http://creativecommons.org/licenses/by-nc-nd/4.0/).
1. Introduction
A coastal region is a transition area between the land and the sea. Coastal regions cover less than 20% of the global land surface but host more than 45% of the world’s population (Crossland et al., 2005; Mentaschi et al., 2018). They are used for human set-tlement, agriculture, industry, fishing, transportation, and recre-ation and are potentially a source of energy from tidal and wave power (Clark, 1927; Davidson-Arnott, 2010). Many of those activi-ties pose threats to the coastal region, accelerating processes such as coastal erosion, inundation, sedimentation, and habitat and resource degradation. More than 70% of the world’s coastal regions are experiencing coastal erosion (Ghosh et al., 2015) and climate change is likely to impact the coastal regions’ exposure to coastal flooding (Addo et al., 2011). In addition, more than 200 million people worldwide are vulnerable in case of flooding by extreme sea levels (Nicholls, 2010).
The need for assessing, monitoring and mitigating coastal pres-sures is increasing because of the increase of coastal population and infrastructure (Cenci et al., 2018). In addition, effective coastal planning and management are a prerequisite for sustainable coastal development (Kay and Alder, 2005; Kumar and Chauhan, 2010). Monitoring spatio-temporal changes of the coastal environ-ment is critical. It can help to understand the spatial distribution of coastal erosion, to predict the trends and to support research on coastal erosion and its countermeasures (Zhang, 2011).
The shoreline is the boundary between the land and the sea. It is one of the primary indicators of environmental change showing changes in coastal conditions such as fluvial process, sediment sup-ply, and relative sea level (Morton, 2002). The position of the shoreline keeps changing due to natural and anthropogenic factors (Aouiche et al., 2016; French, 2001; Pardo-Pascual et al., 2012). Increasing human activities in the coastal region such as the devel-opment of residential areas, industries, tourist destinations and the construction of harbours and jetties cause massive changes of shorelines (Alharbi et al., 2017; Stefano et al., 2013). Waves, tides, and storms intensively erode and reshape the shorelines. In addi-tion, the disturbance in sediment transport processes which supply sediment to the coastal system, may cause erosion and coastal inundation that further contribute to the change in shoreline
https://doi.org/10.1016/j.ejrs.2019.09.001
1110-9823/Ó 2019 National Authority for Remote Sensing and Space Sciences. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer review under responsibility of National Authority for Remote Sensing and
Space Sciences.
⇑ Corresponding author.
E-mail addresses: ratna.sari@big.go.id (R.S. Dewi), w.bijker@utwente.nl
(W. Bijker).
Contents lists available atScienceDirect
The Egyptian Journal of Remote Sensing and Space Sciences
j o u r n a l h o m e p a g e : w w w . s c i e n c e d i r e c t . c o mpositions (Ahn et al., 2017; Bouchahma and Yan, 2014; Brown et al., 2011). Due to the importance of the processes that occur along the shoreline, rapid and reliable techniques are required to monitor the changes. Shoreline monitoring is an important task in providing information for coastal management, coastal environ-ment protection and coastal hazard assessenviron-ment. Frequent observa-tions of shoreline position can provide answers to quesobserva-tions such as ‘‘where” and ‘‘how fast” the coast has changed, which can help coastal planners in the future.
Shoreline position can be monitored using a range of methods. Shoreline survey, aerial photographs, and LIDAR-based methods are well-known as primary techniques for shoreline mapping. However, they are high-cost (both in time and money) and labour intensive methods (IoW-Council, 2006; Morton et al., 1993). Despite the main advantage of shoreline survey and LIDAR in pro-viding accurate results, these data sources are generally limited in their spatial and temporal availability because they are costly. Meanwhile, to interpret aerial photograph requires trained staff (Boak and Turner, 2005; Ghosh et al., 2015; Li et al., 2004).
Nowadays, satellite image-based methods have become increasingly popular to complement the conventional methods for monitoring shoreline change due to their large coverage and low cost (Boak and Turner, 2005). Several methods exist to identify shoreline features including band ratio (Kuleli, 2010; Sarwar and Woodroffe, 2013), water index method (Feyisa et al., 2014; McFeeters, 1996; Sunder et al., 2017), and supervised and unsuper-vised classifications (Dewi et al., 2016; Duru, 2017; García-Rubio et al., 2015; Tamassoki et al., 2014).
For shorelines derived from image classifications, uncertainties can arise for example from the nature of the shoreline (Foody and Atkinson, 2002), and the application of image classification and change analysis methods (Cheng et al., 2001). The position of the land-water boundary cannot be defined precisely due to for instance the degree of wetness of the beaches. Furthermore, a shoreline is vague as its position is changing through time. In this case, the uncertainty of the shoreline will arise when a method that does not accommodate the vagueness of the shoreline is selected. To identify changes in water bodies, it is necessary to process satel-lite images in such a way that water and land can be discriminated. Fuzzy c-means (FCM) classification is applied to estimate water membership in each pixel (Bezdek et al., 1984), then thresholding is applied to differentiate land form water bodies. Fuzzy c-means classification is a clustering method to separate data clusters with fuzzy means and fuzzy boundaries allowing for a partial member-ship. This method allows multiple memberships for a pixel in order to deal with the uncertainty of shoreline positions (Dewi et al., 2016; Shi, 2010).
Sayung sub-district is located in the northern area of Java Island, Indonesia, that has been subjected to severe coastal inunda-tion and erosion. The area has shown massive changes of shoreli-nes during the last three decades due to permanent coastal inundation and severe erosion (Dewi et al., 2018). Remote sensing has proved its effectiveness in providing accurate information, but information regarding shoreline changes for the study area is lim-ited. Shoreline information (such as spatial distribution and the trend of changes) is very important for sustainable coastal devel-opment. Furthermore, the shoreline change is an important indica-tor for coastal hazards and risk assessments as most of the hazards, for instance beach erosion, landslide and tsunami including coastal inundation, are related directly to the stability of the shore (Morton, 2002).
In this study, an effort has been made to map the shoreline in Sayung, Indonesia and to identify shoreline changes caused by coastal inundation and erosion using a series of Landsat images from 1988 to 2017 by means of established remote sensing
techniques. Fuzzy classification and post classification comparison were proposed to identify shorelines and monitor their changes (Dewi, 2018). Thresholding was applied to discriminate the land-water interface. The LUC maps were used to find the connection between changes in land use/cover and changes in shorelines. This research supports disaster risk reduction by addressing specific questions: a) in which land use/cover class(es) have shoreline changes occurred; and b) which parts of the shoreline are more vulnerable to erosion and accretion?
2. Methodology 2.1. Study area
The study was conducted in the northern coast of Sayung sub-district in Central Java Province, Indonesia (Fig. 1), which is approximately 69 km2. The study area is characterized by a low-land low-landscape with an elevation of less than 20 m above mean sea level. The area is a deltaic plain formed by sedimentation from many rivers that run across the area. It is a relatively low wave-energy coast and is adjacent with Java Sea in the north (Ongkosongo, 2010). The area has a mixed semi-diurnal tide with a tidal range of approximately 1.0 m (BIG, 2017).
Sayung sub-district consists of 20 villages with 77,800 inhabi-tants. This area has experienced massive shoreline changes since the last three decades influenced by coastal floods that occur in line with tide cycles. Flooding is becoming worse as a result of land subsidence and sea level rise. (Abidin et al., 2013; Marfai and King, 2007a). Two hamlets in this area have submerged permanently and all the villagers were relocated to other places in 2004 (Damaywanti, 2013). There are many factors that intensify flooding in this location, such as in the short term: extreme winds, heavy rains, and in the long term: sea level rise, mangrove conversion, dyke and irrigation failure, drainage system damage, groundwater extraction, construction load, and an increase of impervious sur-faces. Urban development, which has proceeded without consider-ation of flood risk, has resulted in an increase of coastal inundconsider-ation and erosion that have led to substantial socio-economic losses such as the loss of damaged buildings, settlements, coastal structures, and agricultural production (Marfai, 2011; Winterwerp et al., 2014). Permanent coastal inundation has also caused rapid changes of shoreline in the area (Ervita and Marfai, 2017). 2.2. Satellite images and data pre-processing
Table 1lists the images that were used in this study, consisting of Landsat and ASTER images with 30 and 15 m resolution recorded at the low tide dating from 1988 to 2017. Atmospheric path radiance effect was reduced by applying histogram minimum adjustment (Hadjimitsis et al., 2010). Landsat 8 OLI/TIRS of 27 August 2013 was used as a base image and it was rectified by using a 2013 Pleiades image recorded on 27 February 2013 at the low tides.
For classification, seven bands of the Landsat OLI/TIRS, six bands of the Landsat TM, and six bands of the 2006 ASTER images were used. All bands are in visible, near infrared (NIR) and short wave near infrared (SWIR) bands. The ASTER image was resampled to 30 m using the nearest neighbour resampling method. Reference data from several images including Sentinel-2, ASTER, and Landsat with 10, 15, and 30 m spatial resolution, respectively, were used for accuracy assessment purposes. Those images were obtained from USGS EarthExplorer (USGS, 2017). Except for the Landsat images of 1994 and 2000, all images were recorded during the low tide, and the variation in water level was very small.
2.3. Shoreline model and change detection
To identify shoreline positions, FCM classification was per-formed and shorelines were generated based on the method pro-posed by Dewi et al. (2016). FCM as a clustering method was used to estimate the membership value of each pixel to water and non-water classes (Bezdek et al., 1984). Following unsuper-vised clustering for two classes, a combination of infrared bands (near or short-wave infrared) was used to decide which of the two membership images belonged to the water class. Readers are advised to refer to Dewi et al. (2016)for detailed descriptions regarding FCM classification for estimating the membership value of water pixels for shoreline detection. In this paper, the term shoreline, water, non-water and other land use/cover classes are in italics when they denote class names during an image classification.
Based on Dewi et al. (2016), shoreline locations were deter-mined by generating a margin or transition zone between both classes (water and non-water) based on a threshold range of the water membership (
a
) obtained from optimization of the FCM parameter. The threshold range when generating shoreline margin was decided based on kappa (j
) values via the estimation of lower (a
1) and upper (a
2) thresholds. This method considers theshore-line as an area because the method argues that the shoreshore-line posi-tion is imprecise. It is hardly possible to delineate shorelines as a single line due to gradual transition between water and land. Moreover, the shoreline position changes through time associated with the height of the sea level influenced by the tide. Therefore, the shoreline margin was developed by creating a crisp area deter-mined by
a
1¼ 0:3 anda
2¼ 0:7 as lower and upper thresholds.Each water class image was converted into polygon features in a GIS (Geographic Information System) environment. Three areas were identified: 1) water area if
a
0:7, 2) non-water area ifa
< 0:3, and 3) shoreline margin area if 0:3a
< 0:7. In fact, changes in shoreline position are caused by the exchanges between the shoreline margin and water or non-water areas.To detect the changes along the shoreline margin, two classified images were overlaid. Post classification comparison was used to extract detailed ‘‘from-to” change information. The shoreline posi-tion was highlighted to infer erosion and accreposi-tion sectors along the shore area, and shoreline changes were calculated. Six types of change were identified, namely shoreline margin to water, non-water to non-water, non-water to shoreline margin, non-non-water to shoreline
Fig. 1. Sayung sub-district in Central Java Province, Indonesia was selected as the study site. Topographic map from 2000 is used to visualize the area, which are dominated by agriculture areas and fishponds.
Table 1
Images used in this study and their reference data for accuracy assessment purposes. Images Acquisition Date Astronomical
Tide Level (m)
Reference Data
Landsat TM 23 Sep 1988 0.03 Landsat TM, 23 Sep 1988 Landsat TM 08 Sep 1994 0.19 Landsat TM, 08 Sep 1994 Landsat TM 06 Jul 2000 0.19 ASTER, 06 Sep 1999 ASTER 12 May 2006 0.08 ASTER, 12 May 2006 OLI/TIRS 27 Aug 2013 0.09 ASTER, 19 Aug 2013 OLI/TIRS 19 Jun 2017 0.04 Sentinel-2, 28 Jun 2017
margin, shoreline margin to non-water and water to non-water.Fig. 2 illustrates the process of shoreline change analysis.
2.4. Land use/cover generation and change detection
To have a better understanding in which LUC class(es) shoreli-nes have changed over the three decades, an existing LUC map from 2000 was compared with a new LUC map extracted from the 2017 Landsat image.
2.4.1. Extracting the LUC map from the topographic map
A digital topographic map was obtained from the Indonesian Geospatial Agency (BIG, 2016). The topographic map consists of several classes, namely irrigated and rain-fed paddy field, mixed gar-den/plantation, forest, non-irrigated field/palawija farm, grass, scrub, settlement, building, mangrove, fishpond, lake and river. Some LUC classes were combined so that a comparable class between LUC map produced from this source and classified Landsat image of 2017 was obtained. The combination of these LUC classes can be seen inFig. 3. The final classes were water body, fishpond, paddy field, other crops, mangrove, and built-up.
2.4.2. Extracting the LUC map from the 2017 Landsat image
To extract the LUC map from the 2017 Landsat image, a maxi-mum likelihood classification (MLC) was performed. Spectral classes were defined and spectral reflectance statistics were calcu-lated. For each class at least 1000 pixels were collected. Based on visual interpretation and analysis of the feature space, spectral characteristics of six land cover classes were sampled as inTable 2. A training set was defined and the spectral separability was com-puted by using the Jeffreys-Matusita distance (J-M), and subse-quently, MLC was computed.
To differentiate paddy field from other vegetation, the existing 2017 rice field map from the Ministry of Agriculture of Indonesia (MoA, 2017) was used. Rice field locations were identified and sep-arated from other types of cropland. Afterwards, cropland and mixed garden classes were merged, and a new label was given, i.e. other crops (OC). From this step, two new classes were obtained, namely paddy field (PF) and other crops.
Similarly, fishpond was separated from other water bodies by using the 2014 aerial photo and World Imagery obtained from Arc-GIS online (ESRI et al., 2017). Fishpond was interpreted visually from these data. Fishponds have certain shape characteristics; they
gen-erally show as regular objects in remote sensing images. Fishponds are periodically emptied for maintenance purposes. When that was the case at the moment of image recording, it does not mean that they were abandoned. For inundated fishponds, nets are fixed on top of the dykes to separate the fishponds (seeFig. 4a,b). This fish
Fig. 2. (a) Shoreline margin at timet1; (b) shoreline margin att2; (c) shoreline change estimation with six changed areas (A-F). While solid lines represent shoreline margin att1,
dashed line represents shoreline margin att2(modified fromDewi et al. (2016)).
Fig. 3. Combination of LUC classes from the 2000 topographic map. (a) Final LUC classes; (b) LUC classes of topographic map.
Table 2
Definition of land use/cover (LUC) classes in this study. LUC class Definition
Water body 1 (W1) Water bodies such as open sea, rivers and lakes; Water body 2 (W2) Water bodies with a large accumulation of sediment
such as muddy area, fishpond, and cultivated area at the beginning of the planting period;
Mixed garden (MG) Vegetated area close to settlements like shrubs, herbaceous plants and fruit orchards; Cropland (C) The area where paddy fields and other crops are
planted;
Mangrove (M) Trees that grow in brackish water;
net allows free water exchange. Each fishpond is separated by dykes and very often, mangroves grow on these dykes or inside fishponds (Fig. 4c-d). In this case, fishpond dykes could be identified from images. Only if the water level increased and these dykes were sub-merged completely, this fishpond could no longer be identified from images (Fig. 4e). After separating fishponds from other water bodies, water body 1 and water body 2 were merged, and a new label was given, i.e. water body (W). Therefore, the final classes for the 2017 LUC map were water body (W), fishpond (F), paddy field (PF), other crops (OC), mangrove (M), and built-up (BU).
2.4.3. Change analysis of the LUC maps
Both LUC maps from 2000 and 2017 were converted into vector layers by using GIS software. To detect the changes from the LUC maps, post classification change detection was performed (Jianya et al., 2008; Lambin and Strahler, 1994). The advantage of this method is that there is no need for radiometric correction of
images involved (Raja et al., 2013), however, this technique does not allow to detect changes within land cover classes (Dewi et al., 2017; Lambin and Strahler, 1994).
Finally, based on the results, post classification comparison was implemented to gather ‘from-to’ change information. The area of a specific change category was estimated by multiplying the number of pixels belonging to that specific change category and the area of a pixel. Change category and change percentages in terms of gain, loss and net changes were estimated. Gain refers to the changes into a class, for example gain of water body means changes from other classes such as paddy field into water body. On the other hand, loss indicates changes from a class into other classes, for instance loss of water body means the changes from water body into other land cover classes. Afterwards, net changes were estimated by calculating the difference between loss and gain. Fig. 5 shows the work flow of the LUC change detection analysis.
Fig. 4. (a-e) Aerial images of 2014 and related photos taken during 2015 fieldwork were presented and used to differentiate various characteristics of fishpond in the study area. Yellow points with numbers show the location where photos were taken: (1–2) fishponds with netted structures; (3–4) fishponds separated with natural dykes; and (5) inundated fishponds.
2.5. Accuracy assessment
The accuracy of the shoreline model and the LUC maps were assessed by generating a conventional error matrix. For the shore-line image and LUC map, 140 and 300 points were collected ran-domly from each reference image to perform the accuracy assessment of the shoreline model and the LUC map, respectively. A visual interpretation approach was performed to distinguish a land cover class for each selected point. Afterwards, kappa (
j
) val-ues were estimated by generating the confusion error matrix (Congalton and Green, 2009).3. Results and discussion
3.1. Shoreline margins and their changes
The water membership and the classified images are presented in Fig. 6. Dark blue colour pixels in water membership images
(Fig. 6a–d and i–j) represent a high value of water membership indicating water area. Meanwhile, light blue colour pixels show a low value of water membership which indicates non-water area (land).Table 3presents the accuracy assessment results in terms of kappa (
j
) values when generating shoreline margin via thresh-olding. Values from 0.1 to 0.9 were set to estimate a threshold interval to generate shoreline margin. From the results, it can be seen that the highestj
values were obtained when thresholds were set to 0.4–0.5. Meanwhile,a
1=0.3 anda
2= 0.7 obtained lowerj
values. Threshold values below 0.3 or higher than 0.7 show rather badj
values. Given these results, a threshold interval between 0.3 and 0.7 was chosen as the interval for generating shoreline margin.Fig. 6(e-h and k–l) present the classified images showing the spatial distribution of shoreline margin, water and non-water from 1988 up to 2017. Meanwhile,Fig. 7shows the comparison of area percentages of each class. From the figure, there was a trend of the increase of the water area from 14% in 1988 up to 54% in 2017.
Fig. 6. Water membership and classified images. Dark blue pixels in water membership indicate a high membership to water while a light blue colour represents low membership to water. Classified images consist of three classes: water, non-water and shoreline margin. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
Over the same time, the non-water area was decreasing from 78% in 1988 down to 36% in 2017.
The changes in shoreline positions are shown inFig. 8, while the extents of erosion and accretion are presented in Table 4. The
locations of major erosion and accretion areas were identified for each study period including some unaffected locations. The largest coastal erosion and inundation, which have caused a substantial loss of coastal land, occurred in the period 1994–2000 (Fig. 8b,g). FromTable 4, it is obvious that during this period the study area experienced the maximum erosion of 18 km2, with the largest
con-tribution of the changes from non-water into water for 11 km2. The second largest erosion is also obvious in the period 2006–2013 (see Fig. 8d,i). The net area change denoting erosion was 14 km2which
was mainly due to the change from non-water to water for 8 km2. Despite the fact that coastal erosion and inundation have affected some areas and caused loss of land, new land has appeared at a few locations (see red circle inFig. 8i). Moreover, in the period 2013– 2017, the land area has increased for 5 km2.
The overall change of shoreline positions during the twenty-nine-year time period (between 1988 and 2017) is shown in Fig. 9. Massive change of non-water to water implying erosion has occurred in the area. The amount of erosion due to this change was 25 km2(seeTable 4). In total, the overall erosion (the changes
of non-water to shoreline margin, non-water to water, and shoreline margin to water) that has occurred was 34 km2, whereas the overall
accretion (due to changes of shoreline margin to non-water, water to
Fig. 7. Comparison of the respective extents of shoreline margin, water and non-water by percentage with the total area equal to 69 km2.
Fig. 8. Shoreline change of Sayung, based on image analysis from (a) 1988 to 1994, (b) 1994–2000, (c) 2000–2006, (d) 2006–2013, and (e) 2013–2017. (f-j) The spatial distribution of erosion and accretion along the shore inferred from the change in shoreline positions. Red circle shows the position of accretion in mangrove areas. Table 3
Kappa (j) values from the accuracy assessment results by settinga= 0.1–0.9 for generating shoreline margin. Image Threshold values (a)
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1988 0.50 0.58 0.74 0.79 0.8 0.77 0.71 0.58 0.39 1994 0.54 0.59 0.82 0.84 0.82 0.82 0.76 0.62 0.51 2000 0.61 0.65 0.77 0.85 0.83 0.81 0.77 0.73 0.64 2006 0.57 0.7 0.81 0.87 0.91 0.89 0.81 0.68 0.6 2013 0.46 0.65 0.78 0.81 0.81 0.79 0.77 0.52 0.51 2017 0.50 0.74 0.78 0.84 0.85 0.79 0.67 0.51 0.29
non-water and water to shoreline margin) was only 2 km2. Hence,
the net area change was 32 km2.
3.2. Land use/cover (LUC) maps and their changes
Fig. 10b shows the 2000 LUC map obtained after merging some classes of the 2000 topographic map (Fig. 10a), whereasFig. 10c presents LUC map from the classified Landsat image. From Fig. 10a,b, it can be seen that the study area was dominated by fish-pond (in blue) at the sea-side of the land and a largely agricultural area (in green) dominated by paddy field to the landward direction. Coastal urban area, represented in red coral, was located along the river side.
After LUC maps from 2000 and 2017 were overlaid, thirty types of change were identified. For further analysis, the changes were merged into six types: i) changes into water body representing all the changes of LUC class (fishpond, paddy field, other crops, man-grove, and built-up) into water body; ii) changes into fishpond repre-senting all the changes of LUC class (water body, paddy field, other crops, mangrove, and built-up) into fishpond; iii) changes into paddy
field; iv) changes into other crops; v) changes into mangrove; and vi) changes into built-up (seeFig. 11).
FromFig. 11a,b, it is obvious that the largest LUC change was changes into fishpond for 16 km2(denoted by a dark blue colour),
whereas the second largest LUC change was changes into water body for 12 km2. In this case, fishpond slowly changed into water
body when the water level in the fishpond areas was getting higher. Hence, these fishponds can no longer be used to cultivate fish (Marfai, 2011).
By looking at information provided inFig. 12, it is clear that the agricultural area -paddy field and other crops- was the largest con-tributor to the changes into fishpond category for 11% and 12% of the areas, respectively. In this case, substantial changes of paddy fields and other crops were due to the conversion of these agricul-tural areas into fishpond. Fishpond, in turn, suffered substantial changes into water body, namely 12% (seeFig. 12a). The change of fishpond areas gave the largest contribution to the changes into water category.
In this location, land subsidence was identified as a factor that aggravates coastal inundation. The subsidence rates in this area
Table 4
Changed area (in km2
) in the period 1988–1994, 1994–2000, 2000–2006, 2006–2013 and 2013–2017. The largest erosion occurred in the period 1994–2000, and the largest accretion occurred in the period 2013–2017.
Change classes Change area in km2
1988–1994 1994–2000 2000–2006 2006–2013 2013–2017 1988–2017
non-water to shoreline margin 1 3 4 2 1 5
non-water to water 1 11 4 8 1 25
shoreline margin to non-water +1 +0.3 +1 +1 +1 +0.5
shoreline margin to water 2 5 2 6 1 4
water to non-water +1 +0.2 +1 +1 +2 +0.5
water to shoreline margin +2 +1 +2 +1 +5 +1
Net changes 1 18 7 14 +5 32
Note: + gain of non-water (accretion); loss of non-water (erosion).
vary from place to place, with rates larger than 3 cm yr1 (Chaussard et al., 2013). The subsidence may be caused by the combination of natural consolidation of alluvium soil, ground water extraction and the load of buildings (Abidin et al., 2013; Marfai and King, 2007b). The Asia-Pacific Network for Global Change Research (APN, 2013) after Marsudi (2001) reported that
the ground water table decreased at rates of 1.2–2.2 m yr1 between 1990 and 1996. Ground water extraction occurs for indus-trial purposes and for household needs as the consequences of the population growth. Furthermore, an excessive ground water extraction not only triggers land subsidence but also salt water intrusion. Even though the inundation as a result of subsidence is
Fig. 10. Land use/cover maps; (a) LUC map with an original class from the 2000 topographic map, (b) 2000 LUC map with combined classes, and (c) 2017 LUC map from Landsat image.
Fig. 11. (a) Spatial distribution of LUC changes in the study area from 2000 to 2017; and (b) Changed areas in km2
much larger than that of sea level rise, a combination of land sub-sidence and sea level rise makes the shoreline more vulnerable to erosion by increasing the wave height, in particular in a muddy coastal environment (Wesenbeeck et al., 2015). The rise of the sea level in Semarang and surroundings was reported as approxi-mately 2.65 mm yr1(Wirasatriya et al., 2006). Once erosion is ini-tiated, the water and salinity level start to rise in remaining fishponds and further affect the agricultural area. This explains the changes from paddy field into fishpond and from fishpond into water body in our results. As mentioned byMarfai (2011), paddy and crops were exposed to a lethal salinity level, which caused a decline in the productivity of both paddy and crops (see Fig. 13a). Furthermore, inundation of fishponds (seeFig. 13b) not only causes a decline in fish productivity but can also lead to aban-donment of the fishpond area as mentioned byWesenbeeck et al. (2015); when fishponds get inundated, people revert to small-scale off-shore fishery.
Massive erosion and coastal inundation in this area might be influenced by massive reclamation in the neighbourhood area i.e. Semarang city, which is located in the western side of the study area (seeFig. 14). The reclamation activity in Semarang, as the cap-ital of Central Java, has a long history dating back to the 1980s (Miladan, 2016). Large scale coastal reclamation occurred because
of a high demand for space for housing and economic activity. Fish-pond and marshes turned into urban areas including settlements, commercial- and business areas, recreational areas and industrial zones, thereby anticipating urban growth. Land reclamation was carried out starting from 1979 to develop Tanah Mas real estate, followed by Puri Anjasmoro and Semarang Indah real estates in 1985, and Marina real estate in 2004 (Pratiwi, 2012). Moreover, Tanjung Mas Port, Ahmad Yani Airport and the industrial zone were also extended through coastal reclamation since the 1980s and 1990s, respectively. Many of those reclamation projects are still ongoing (ANTARANews, 2017; Miladan, 2016).
The largest part of the changes into mangrove category was due to the changes of fishpond and other crops into mangrove, for about 2% and 1% of the area, respectively. Despite the fact that substantial erosion occurred in the study area, which led to a massive retreat of the shoreline position, a gradual accretion leading to an advanc-ing shoreline can also be identified (see the orange pixels in Fig. 11a for e.g., grid cells A2, B1 and B2). Replanting of mangrove trees (seeFig. 15a) as a response to coastal erosion was one of the causes for this accretion, for example in Bedono (Fikriyani and Mussadun, 2014) and Timbulsloko villages (Astra et al., 2014). Another coping strategy to coastal erosion was hybrid-engineering, combining mangrove rehabilitation with
conven-Fig. 13. (a) Paddy field experienced crop failure due to floods, (e) flooded fishpond with net.
Fig. 12. (a-f) Detailed changed areas in percentage of each LUC class between 2000 and 2017 (W = water body, F = fishpond, M = mangrove, OC = other crops, PF = paddy field, BU = built-up).
tional engineering measures (seeFig. 15b) such as sediment traps, so mangroves can grow again (Wesenbeeck et al., 2015).
Changes of fishpond into other crops, paddy field and built-up covered only small parts of 0.3%, 0.2%, and 0.4% of the area, respec-tively (seeFig. 12). These changes might be due to land reclama-tion and sedimentareclama-tion that occurred in this locareclama-tion especially in the north-eastern part of the area. As has been mentioned, many rivers run across the study area, which brought sediment from upstream to the downstream area and caused silting up at the river mouth (Subardjo, 2004). Over longer time, this caused severe coastal inundation and erosion since some of the sediment was deposited along the river bottoms and estuaries. This led to the narrowing and to the silting up of the river, and the reduction of sediment supply to the near shore which further induced coastal inundation and erosion (Astra et al., 2014). Meanwhile, in Semar-ang (bordering the study area at the western side), inundated fish-ponds were reclaimed to develop industrial and residential areas (Miladan, 2016). Despite the fact that land reclamation expanded the space available for economic purposes, this activity came at a price in terms of its negative impact on environment. The construc-tion of urban areas increased the surface runoff and reduced the ability of the ground to absorb rainfall. Furthermore, when there
were major land use changes in the coastal area, for example fish-ponds, swamps and paddy fields turned into built-up areas; conse-quently, floods, land subsidence, and erosion leading to coastal inundation occurred not only in the urban zones that were devel-oped on the marsh areas but also in adjacent areas.
4. Conclusion
This study successfully developed a method to investigate shoreline changes in the north of Java. The method is based on the combination of fuzzy classification and post classification com-parison. The results show that massive changes in the shoreline position could be observed in the area, visible as changes from agriculture area (paddy field, other crops) into fishpond and changes from fishpond into water body. These changes are mainly caused by land subsidence, coastal erosion and flooding.
Information regarding the spatial dynamics of shoreline change, as provided by this research, is an important aid to coastal man-agers and coastal planners to prioritise actions related to disaster risk reduction. By knowing areas that are at risk to be affected by the changes of the shoreline, the local government may modify
Fig. 15. (a) Replanting of mangroves, and (b) hybrid structure as shoreline restoration projects in the study area.
Fig. 14. Reclamation projects in the nearby area (Semarang city) started around 1980s to develop housing complex (points a,b,e), harbour (point c), and industrial area (point d). Source: images made available via Google Earth.
land use planning and move citizens and economic activities away from hazardous areas. Moreover, in areas at risk of inundation, the government can implement coping strategies, such as constructing flood defences and planting mangroves to stabilize shoreline areas. Declaration of Competing Interest
The authors declare that they have no known competing finan-cial interests or personal relationships that could have appeared to influence the work reported in this paper.
References
Abidin, H.Z., Andreas, H., Gumilar, I., Sidiq, T.P., Fukuda, Y., 2013. Land subsidence in coastal city of Semarang (Indonesia): characteristics, impacts and causes. Natural Hazards and Risk 4, 226–240.
Addo, K.A., Larbi, L., Amisigo, B., Ofori-Danson, P.K., 2011. Impacts of coastal inundation due to climate change in a CLUSTER of urban coastal communities in Ghana, West Africa. Remote Sens. 3, 2029–2050.
Ahn, Y., Shin, B., Kim, K.-H., 2017. Shoreline change monitoring using high resolution digital photogrammetric technique. J. Coastal Res., SI, 204–208.
Alharbi, O.A., Phillips, M.R., Williams, A.T., Thomas, T., Hakami, M., Kerbe, J., Niang, A.J., Hermas, E., Al-Ghamdi, K., 2017. Temporal shoreline change and infrastructure influences along the southern Red Sea coast of Saudi Arabia. Arab. J. Geosci. 10.
ANTARANews, 2017. Pelindo II are planning to extend Tanjung Mas Port through land reclamation. In. Semarang: ANTARA News.
Aouiche, I., Daoudi, L., Anthony, E.J., Sedrati, M., Ziane, E., Harti, A., Dussouillez, P., 2016. Anthropogenic effects on shoreface and shoreline changes: input from a multi-method analysis, Agadir Bay, Morocco. Geomorphology 254, 16–31. APN, 2013. A study on loss of land surface and changes to water resources, resulting
from sea level rise and climate change. In: Final Report Submitted to APN (p. 201). Kobe, Japan: Asia-Pacific Network for Global Change Research. Astra, A.S., Sabarini, E.K., Harjo, A.M., Maulana, M.B., 2014. Mangrove-based and
community-based coastal protection in Timbulsloko village, Demak. In: Wetland Conservation Magazine (pp. 4–5). Bogor, Indonesia: Wetland International Indonesia.
Bezdek, J.C., Ehrlich, R., Full, W., 1984. FCM: the fuzzy c-means clustering algorithm. Comput. Geosci. 10, 191–203.
BIG, 2016. Ina-Geoportal. In: Cibinong, Indonesia: BIG (Indonesia Geospatial Information Agency).
BIG, 2017. Online Tidal Prediction. In: Cibinong, Indonesia: BIG (Indonesia Geospatial Information Agency).
Boak, E.H., Turner, I.L., 2005. Shoreline definition and detection: a review. J. Coastal Res. 21, 688–703.
Bouchahma, M., Yan, W., 2014. Monitoring shoreline change on Djerba Island using GIS and multi-temporal satellite data. Arab. J. Geosci. 7, 3705–3713.
Brown, S., Barton, M., Nicholls, R., 2011. Coastal retreat and/or advance adjacent to defences in England and Wales. J. Coast. Conserv. 15, 659–670.
Cenci, L., Disperati, L., Persichillo, M.G., Oliveira, E.R., Alves, F.L., Phillips, M., 2018. Integrating remote sensing and GIS techniques for monitoring and modeling shoreline evolution to support coastal risk management. GIScience & Remote Sensing 55, 355–375.
Chaussard, E., Amelung, F., Abidin, H., Hong, S.-H., 2013. Sinking cities in Indonesia: ALOS PALSAR detects rapid subsidence due to groundwater and gas extraction. Remote Sens. Environ. 128, 150–161.
Cheng, T., Molenaar, M., Lin, H., 2001. Formalizing fuzzy objects from uncertain classification results. Int. J. Geogr. Inf. Sci. 15, 27–42.
Clark, J.R., 1927. Coastal Zone Management Handbook. CRC Press, Florida.
Congalton, R.G., Green, K., 2009. Assessing the Accuracy of Remotely Sensed Data: Principles and Practices. CRC Press, Boca Raton.
Crossland, C.J., Baird, D., Ducrotoy, J.-P., Lindeboom, H., 2005. The coastal zone – a domain of global interactions. In: Crossland, C.J., Kremer, H.H., Lindeboom, H.J., Crossland, J.I.M., Tissier, M.D.A.L. (Eds.), Coastal Fluxes in the Anthropocene. Springer, Berlin, p. 231.
Damaywanti, K., 2013. Dampak Abrasi Pantai terhadap Lingkungan Sosial (Studi Kasus di Desa Bedono, Sayung Demak). In: Prosiding Seminar Nasional Pengelolaan Sumberdaya Alam dan Lingkungan Semarang.
Davidson-Arnott, R., 2010. An Introduction to Coastal Processes and Geomorphology. Cambridge University Press.
Dewi, R.S., 2018. Multitemporal Image Analysis for Monitoring Fuzzy Shorelines. In: Faculty of Geo-Information and Earth Observation (ITC) (p. 197). Enschede, The Netherlands: Faculty of Geo-Information and Earth Observation, University of Twente.
Dewi, R.S., Bijker, W., Stein, A., 2017. Change vector analysis to monitor the changes in fuzzy shorelines. Remote Sens. 9, 147.
Dewi, R.S., Bijker, W., Stein, A., Marfai, M.A., 2016. Fuzzy classification for shoreline change monitoring in a part of the northern coastal area of Java, Indonesia. Remote Sens. 8, 1–25.
Dewi, R.S., Bijker, W., Stein, A., Marfai, M.A., 2018. Transferability and upscaling of fuzzy classification for shoreline change over 30 years. Remote Sens. 10.
Duru, U., 2017. Shoreline change assessment using multi-temporal satellite images: a case study of Lake Sapanca, NW Turkey. Environ. Monit. Assess., 185 Ervita, K., Marfai, M.A., 2017. Shoreline change analysis in Demak, Indonesia. J.
Environ. Prot. 8, 940–955.
ESRI, Digital-Globe, Earthstar-Geographics, CNES/Airbus-DS, GeoEye, USDA-FSA, USGS, Getmapping, Aerogrid, IGN, IGP, GIS-Community (2017). World Imagery. In. USA: ESRI.
Feyisa, G.L., Meilby, H., Fensholt, R., Proud, S.R., 2014. Automated water extraction index: a new technique for surface water mapping using landsat imagery. Remote Sens. Environ. 140, 23–35.
Fikriyani, M., Mussadun,, 2014. The evaluation of mangrove rehabilitation program in the coastal area of Bedono village Sayung sub-district Demak regency. Ruang 2, 381–390.
Foody, G.M., Atkinson, P.M., 2002. Uncertainty in Remote Sensing and GIS. John Wiley & Sons Ltd, England.
French, P.W., 2001. Coastal Defences: Processes, Problems and Solutions. Taylor and Francis, London.
García-Rubio, G., Huntley, D., Russell, P., 2015. Evaluating shoreline identification using optical satellite images. Mar. Geol. 359, 96–105.
Ghosh, M.K., Kumar, L., Roy, C., 2015. Monitoring the coastline change of Hatiya Island in Bangladesh using remote sensing techniques. ISPRS J. Photogramm. 101, 137–144.
Hadjimitsis, D.G., Papadavid, G., Agapiou, A., Themistocleous, K., Hadjimitsis, M.G., Retalis, A., Michaelides, S., Chrysoulakis, N., Toulios, L., Clayton, C.R.I., 2010. Atmospheric correction for satellite remotely sensed data intended for agricultural applications: impact on vegetation indices. Nat. Hazards Earth Syst. Sci. 10, 89–95.
IoW-Council, 2006. Monitoring and Modelling the Shoreline. In: Practical Guide: Messina Project.
Jianya, G., Haigang, S., Guorui, M., Qiming, Z., 2008. A review of multi-temporal remote sensing data change detection algorithms. In: The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Beijing
Kay, R., Alder, J., 2005. Coastal Planning and Management. Taylor & Francis, USA.
Kuleli, T., 2010. Quantitative analysis of shoreline changes at the Mediterranean Coast in Turkey. Environ. Monit. Assess. 167, 387–397.
Kumar, M., Chauhan, R., 2010. Present status, challenges and management of the japanese coastal zone environment. In: Ramanathan, A., Bhattacharya, P., Dittmar, T., Prasad, M.B.K., Neupane, B.R. (Eds.), Management and Sustainable Development of Coastal Zone Environments. Springer-Verlag Berlin Heidelberg, Germany, p. 275.
Lambin, E.F., Strahler, A.H., 1994. Change-vector analysis in multitemporal space: a tool to detect and categorize land-cover change processes using high temporal-resolution satellite data. Remote Sens. Environ. 48, 231–244.
Li, R., Di, K., Ma, R., 2004. A comparative study of shoreline mapping techniques. In: Bartlett, D., Smith, J. (Eds.), GIS for Coastal Zone Management. CRC Press, Florida, U.S..
Marfai, M.A., 2011. Impact of coastal inundation on ecology and agricultural land use case study in Central Java, Indonesia. Quaestiones Geographicae 30, 19–32. Marfai, M.A., King, L., 2007a. Potential vulnerability implications of coastal inundation due to sea level rise for the coastal zone of Semarang city, Indonesia. Springer: Environment Geology.
Marfai, M.A., King, L., 2007b. Tidal inundation mapping under enhanced land subsidence in Semarang, Central Java Indonesia. Nat. Hazards.
McFeeters, S.K., 1996. The use of the Normalized Difference Water Index (NDWI) in the delineation of open water features. Int. J. Remote Sens. 17, 1425–1432.
Mentaschi, L., Vousdoukas, M.I., Pekel, J.-F., Voukouvalas, E., Feyen, L., 2018. Global long-term observations of coastal erosion and accretion. Sci. Rep. 8, 11. Miladan, N., 2016. Communities’ contributions to urban resilience process: a case
study of Semarang city (Indonesia) toward coastal hydrological risk. In: Architecture, space management (p. 547). France: Universit le Paris-Est. MoA, 2017. SIMOTANDI (Sistem Informasi Monitoring Pertanaman Padi). In.
Jakarta: Ministry of Agriculture of Indonesia.
Morton, R.A., 2002. Coastal geoindicators of environmental change in the humid tropics. Environ. Geol. 42, 711–724.
Morton, R.A., Leach, M.P., Paine, J.G., Cardoza, M.A., 1993. Monitoring beach changes using GPS surveying techniques. J. Coastal Res. 9, 702–720.
Nicholls, R.J., 2010. Impacts of and responses to sea-level rise. In: Church, J.A., Woodworth, P.L., Aarup, T., Wilson, W.S. (Eds.), Understanding Sea-Level Rise and Variability. Blackwell Publishing, UK, pp. 17–51.
Ongkosongo, O., 2010. Indonesia. In: Bird, E.C.F. (Ed.), Encyclopedia of the World‘s Coastal Landforms. Springer, Dordrecht.
Pardo-Pascual, J.E., Almonacid-Caballer, J., Ruiz, L.A., Palomar-Vázquez, J., 2012. Automatic extraction of shorelines from Landsat TM and ETM+ multi-temporal images with subpixel precision. Remote Sens. Environ. 123, 1–11.
Pratiwi, M.R.I., 2012. Impact of the Tidal Floods Dynamics to Social Ecological System of Semarang City (Case Study in Tanjung Mas District). In: Sekolah Pascasarjana (p. 95). Bogor, Indonesia: Institut Pertanian Bogor.
Raja, R.A.A., Anand, V., Kumar, A.S., Maithani, S., Kumar, V.A., 2013. Wavelet based post classification change detection technique for urban growth monitoring. J. Indian Soc. Remote Sens. 41, 35–43.
Sarwar, M.G.M., Woodroffe, C.D., 2013. Rates of shoreline change along the coast of Bangladesh. J. Coast. Conserv. 17, 515–526.
Shi, W., 2010. Principles of Modeling Uncertainties in Spatial Data and Spatial Analyses. Taylor and Francis Group, United States of America.
Stefano, A.D., Pietro, R.D., Monaco, C., Zanini, A., 2013. Anthropogenic influence on coastal evolution: a case history from the Catania Gulf shoreline (eastern Sicily, Italy). Ocean Coastal Manage. 80, 133–148.
Subardjo, P., 2004. Morphological study for rob mapping in the Sayung Coast, Demak Regency, Central Java. Ilmu Kelautan 9, 153–159.
Sunder, S., Ramsankaran, R., Ramakrishnan, B., 2017. Inter-comparison of remote sensing sensing-based shoreline mapping techniques at different coastal stretches of India. Environ. Monit. Assess., 189
Tamassoki, E., Amiri, H., Soleymani, Z., 2014. Monitoring of shoreline changes using remote sensing (case study: coastal city of Bandar Abbas). 7th IGRSM International Remote Sensing & GIS Conference and Exhibition. IOP Publishing. USGS, 2017. EarthExplorer. In: USA: U.S. Geological Survey.
Stefano, A.D., Pietro, R.D., Monaco, C., Zanini, A., 2015. Anthropogenic influence on coastal evolution: a case history from the Catania Gulf shoreline (eastern Sicily, Italy). Ocean Coastal Manage. 80, 133–148.
Winterwerp, H., Wesenbeeck, B.v., Dalfsen, J.v., Tonneijck, F., Astra, A., Verschure, S., Eijk, P.v., 2014. A sustainable solution for massive coastal erosion in Central Java. In, Discussion Paper (pp. 1-45). Delft, The Netherlands: Deltares – Wetland International.
Wirasatriya, A., Hartoko, A., Suripin,, 2006. Study of sea level rise as a base for rob problem solving in coastal region of Semarang City. Pasir Laut 1, 31–42.
Zhang, Y., 2011. Coastal environmental monitoring using remotely sensed data and GIS techniques in the Modern Yellow River delta, China. Environ. Monit. Assess. 179, 15–29.
