Seagrass beds play a crucial role in coastal ecosystems, providing various ecosystems services that are unique to each seagrass species. It is important to have spatial information on the composition of seagrass species in order to effectively manage and utilize seagrass beds. Unfortunately, the lack of such data hinders the optimal management of seagrass beds in Indonesia, as exemplified by the case of Pari Island in the Thousand Islands. To address this issue, WorldView-2 imagery, a multispectral remote sensing image with high spatial and spectral resolution, can be utilized to map composition of seagrass species. Therefore, the objectives of this study were twofold: (1) to map the composition of seagrass species and (2) to assess the accuracy of the resulting seagrass species composition map for selected areas of Pari Island, using WorldView-2 imagery. To achieve these objectives, a combination of image segmentation approaches and multispectral classification employing the random forest algorithm was employed. The findings revealed that the seagrass species composition in the designated areas of Pari Island, based on a life-form classification scheme, comprised Enhalus acoroides (Ea) covering an area of 0.04 km2, Enhalus acoroides-Thalassia hemprichii (EaTh) covering 0.21 km2, and Thalassia hemprichii-Cymodocea rotundata (ThCr) covering 0.28 km2. The overall accuracy (OA) of the seagrass species composition map was determined to be 60.76%.
Seagrass beds are ecosystems that are sensitive to anthropogenic pressures. The restrictions imposed on human activities during the Covid-19 pandemic provide a unique opportunity for coastal ecosystems to recover. This situation presents a chance to monitor seagrass ecosystems on Panggang Island and Karang Congkak Island, aiming to observe any changes in the seagrass area. The method employed to ascertain the community structure involved photo transects, which were then processed using a machine learning supervised classification approach utilizing the random forest algorithm. The classification process categorized the area into five classes: seagrass, macroalgae, coral, bare substrate, and dead coral algae. Confusion Matrix Test was chosen to assess mapping accuracy, setting an accuracy threshold of ≥ 60%. The findings reveal notable alterations in seagrass areas across both islands. Panggang Island exhibited a considerable increase in seagrass coverage by 21,05 hectares. Conversely, Karang Congkak Island experienced a decrease of 2,88 hectares in seagrass coverage. These observed changes are likely not influenced by the anthropause phenomenon but rather by natural occurrences between 2019 and 2022, including the La Niña Triple Dip phenomenon and variations in sampling time during the transition season. The Overall mapping accuracy (OA) results for Panggang Island in 2019 and 2022 were 52,63% and 57,89%, respectively. Similarly, for Karang Congkak Island, the Overall Accuracy results were 60% in 2019 and 60,66% in 2022.
Seagrass, a marine angiosperm, plays a crucial role in providing significant ecosystem services. Due to its highly dynamic nature, seagrass cover can exhibit monthly or seasonal fluctuations. This research aims to investigate the dynamics of seagrass cover changes on Gili Lawang Island, East Lombok Regency throughout the period of 2022-2023, utilizing timeseries PlanetScope images. To develop a model for estimating seagrass cover percentage, we employed a stepwise regression approach that integrated sunglint-corrected Planetscope level 3B bands with field seagrass data. Training and accuracy assessment samples were collected using the photo-quadrate method, spatially distributed across various coastal characteristics of Gili Lawang Island. The obtained time-series seagrass percent cover maps were further analyzed in conjunction with climatic data to discern the underlying patterns governing seagrass cover dynamics. The novelty of this study lies in its potential to serve as a foundation for future research endeavors, such as the analysis of carbon stock dynamics in seagrass fields, and as a basis for establishing seagrass conservation zones in Gili Lawang.
Oil spills frequently occur on the sea surface due to heightened vessel activities. Oil spills can be detected by applying supervised and unsupervised classification methods to satellite images using radar sensors. Supervised classification methods such as visual interpretation are widely used, but the results are very subjective. Conversely, unsupervised methods, while less subjective, necessitate parameter tuning for accurate results. This study's primary goal is to assess the impact of parameter tuning on unsupervised K-Means and Clustering Large Applications (CLARA) algorithms for detecting sea surface oil spills. It can be concluded that the area of identified oil spills is closely related to the iteration parameters and the number of cluster centers. The results of identification using the unsupervised method with these two algorithms will be compared with reference data from Indonesia National Institute of Aeronautics and Space (LAPAN) as the official institution that provides information regarding oil spills pollution on the sea surface in Indonesia. The main conclusion from this study, parameter tuning is highly required before carrying out the process of identifying oil spills on sea level using the unsupervised method especially related to the number of iterations executed, the desired number of cluster centers, and the clustering type of the algorithm used. Using the tuned parameters, the K-Means algorithm is able to identify oil spill areas that are quantitatively close to the reference data area, but the CLARA algorithm is able to provide identification results that have fewer errors in terms of oil spills look-alikes.
Paddy fields are complex land-use entities with various surface covers depending on the timing of the planting stages. Therefore, the best practice to map paddy fields using remote sensing has benefited from the availability of multi-temporal data which were used to characterize the phenology related to the paddy fields. However, this practice may require more RS data to be obtained and processed. Other mapping methods by capitalizing the spatial configuration, such as image segmentation in Object-Based Image Analysis (OBIA) and object recognition in Deep Learning using Convolutional- Neural Network (CNN) architecture has been used in the mapping application. This study aims to assess the accuracy from using mean-shift image segmentation and Random Forests and Extreme Gradient Boosting as the classifiers, with the accuracy from simple CNN architecture, by using Worldview-3 (WV3) full-spectrum image (16 bands). The image segmentation and deep learning analysis were conducted by using 16-bands from the WV3 image and classified by using RF and XGB, and CNN. The results showed that RF was able to identify the paddy fields with an accuracy of 88.09 % (User’s accuracy (UA)) and 81.61 % (Producer’s accuracy(PA)), while XGB produced an accuracy of 85.71 % (User’s accuracy (UA)) and 82.44 % (Producer’s accuracy (PA)), respectively. While CNN produced the accuracies of 49.5 % (PA), 96.3 % (UA) and 82.9 % (OA). The lower producer’s accuracy indicated the higher omission error where more paddy fields were classified as non-paddy fields. CNN produced promising accuracy results for identifying paddy field tiles with 82.9 % accuracy without using data augmentation, although it will be needed to increase the accuracy and more complex CNN architecture such as U-net is needed to determine the boundary of the mapped objects.
Coral reefs is are an important community in coastal and marine ecosystems. Currently, they are under high environmental pressures and suffer damages from human activities and increased sea surface temperature, narrowing the live coral cover. This study aimed to assess the mapping accuracy of the live and dead coral covers using PlanetScope satellite images around Mandangin Island, Madura, Indonesia. Minimum Noise Fraction (MNF) was applied to the bands corrected for the effect of energy attenuation by the water column using the Depth Invariant Bottom Index method, and Random Forest (RF) algorithm was used for mapping. The classification results showed five classes of benthic habitat 2021, namely live coral, dead coral, rubble, seagrass, and sand. Using the confusion matrix, it was found that the live and dead coral cover models had 72.5% accuracy. The mean live coral and dead coral covers were 18.87% and 36.40%, respectively.
Vegetation is the key to the ecological conditions of an area, especially the highlands, which function as protected areas. The mapping of vegetation cover percentage is essential considering that highlands in Indonesia have massive changes, particularly in areas affected by eruptions. This study aims to map the changes in vegetation cover percentage of the area around Mount Agung after the eruption in 2017. Pre-eruption and post-eruption multi-temporal remote sensing data were used to extract the percentage of vegetation cover using an empirical model built from regression of NDVI values and visual observation of vegetation cover percentage based on high-resolution imagery. The estimated error value is 9.67% of pre-eruption condition cover and 14.45% of post-eruption condition cover, used as a threshold value to determine the area and location of percentage changes of vegetation cover. The area of 1.93 km2 decreases vegetation cover percentage due to the eruption on the southern and southwest slopes of Mount Agung. The area that has not changed because of the eruption dominates at 24.92 km2. The area that experienced increasing vegetation cover percentages was minimal due to vegetation growth during the temporal difference of imagery (5 months).
Random forest is a machine learning algorithm that can be used to improve the classification accuracy of mapping using remote sensing, especially for seagrass mapping in a complex optically water shallow. This research is aimed to map seagrass species composition and percent cover using random forest classification and regression using PlanetScope image. Optically shallow water around Labuan Bajo was selected as the study area. Sunglint and water column corrections were applied to the surface reflectance image. Principle Component Analysis (PCA) transformation was applied on surface reflectance bands, deDeglint bands, and depth-invariant index bands. These bands were used as the input band for random forest classification and regression algorithm, using field data to train the algorithm. Benthic field data was collected by the photo transect and seagrass field data was collected by the photo quadrat transect technique. Benthic habitat classification scheme was constructed based on the variation of benthic habitat insitu, which consisted of coral reefs, seagrass, macroalgae, and bare substratum. Seagrass species composition classification scheme was constructed following the variation of seagrass species insitu, which consisted Enhalus acaroides (Ea), Enhalus acaroides mixed Syringodium isoetilolium (EaSi), Enhalus acaroides mixed Thalassia hemprichii (EaTh), Halodule uninervis (Hu), Mixed species class, Thalassodendron ciliatum (Tc), Thalassodendron ciliatum mixed Enhalus acaroides (TcEa), Thalassia hemprichii (Th), Thalassia hemprichii mixed Cymodocea rotundata (ThCr), and Thalassia hemprichii mixed Syringodium isoetilolium (ThSi) class. Accuracy assessment using independent field data showed that random forest algorithm produced 63.57%- 72.09% overall accuracy for benthic habitat and 83.52%-85.71% overall accuracy for seagrass species composition. Random forest regression for seagrass percent cover produced R 2 between 0.78-0.81 with the error of prediction between 14.59-15.26.
Oceanographic conditions, physical development, cultivation, and sedimentation in river estuaries are dynamic trends occured in Jepara Regency. These dynamics need to be understood so it is necessary to determine the position of the shoreline as an impact of morphodynamic to see the latest variations of the shoreline in Jepara Regency. Landsat imagery can be an alternative source of data for shoreline mapping, while shoreline extraction methods can be conducted using water index, which is easy to perform. Regulation published by the Head of the Geospatial Information Agency Number 6 of 2018 can be used as a standard for shoreline maps accuracy obtained from remote sensing imagery. The research objective is to map the Jepara shoreline using NDWI, MNDWI, and AWEI transformations and compare the water index performance. Shoreline data is extracted from Landsat 8 OLI imagery, while the reference shoreline for accuracy assessment is obtained from visual interpretation of PlanetScope imagery. Threshold 0 and subjective threshold based on experiments per coastal physical typology samples are used to separate land-sea. The difference in the shoreline length on the eight shorelines are due to the limited capability of the water index in obtaining the shoreline. MNDWI shoreline with a threshold of 0 gives the lowest RMSE value (RMSE= 25,33 m) among another index, while the NDWI shoreline with a threshold of 0 gives the highest RMSE value (RMSE= 43,77 m).
Seagrass meadows have many ecosystem services to coastal areas and adjacent ecosystems, these services include nursery area for marine organisms, sea turtle feeding ground, and blue carbon sequestration. Therefore, it is important to protect seagrass in order to preserve their functions. Seagrass percent cover is one of the parameters to asses seagrass condition. Several approaches have been developed to map seagrass in optically shallow waters and one of them is by using remote sensing. This approach is more effective and efficient compared to field survey alone. The aim of this study is to produce seagrass spatial distribution and percent cover map using high resolution image. In this research, Support Vector Machine (SVM) classification and regression, one of the machine learning algorithms, was used to classify PlanetScope image using field data as training area to map seagrass spatial distribution and percent cover. The result show that SVM produced 73.98% overall accuracy for benthic mapping, with seagrass class producer’s accuracy and user’s accuracy is 93.71% and 85.35% respectively. Meanwhile, for seagrass percent cover, the SVM algorithm produced map with 26.48% standard error.
Machine learning classification in remote sensing imagery is considered capable of producing classification results with high accuracy in short processing times. This research was conducted with the aim of mapping the spatial distribution of benthic habitat on different types of coral reefs in the waters of Flores Island, NTT using PlanetScope image using Random Forest (RF) and Support Vector Machine (SVM) classification algorithm. Benthic habitat information from field surveys were used to train the RF and SVM algorithm and validate the classification results. The classification results indicated that Mesa Island, the Northern and the Western side of Labuan Bajo are dominated by seagrass beds, and on Bangkau Island is dominated by coral reefs and bare substratum. The highest overall accuracy of the RF classification results is 71.88% from West Labuan Bajo (fringing reef) result. Meanwhile, the highest overall accuracy of the SVM classification is 76.74% from Bangkau Island (patch reef) result.
The green mussel cultivation by fishermen in Pasaran Island is influenced by nature and uses simple technology without regarding water conditions. In fact, site selection considering the water condition is one of the important factors in determining the success of quality green mussel cultivation. High market demand but not supported by modern technology, good marketing strategies, price stability, and appropriate cultivation site can reduce the production of green mussels. This research was conducted to determine the optimal location for the green mussel cultivation around Pasaran Island, in Lampung Bay and to formulate a management strategy based on the map. Modeling parameters measured on the field include depth, salinity, pH, temperature, current velocity, dissolved oxygen, water clarity, and chlorophyll-a. Data processing methods include inverse distance weighted (IDW) interpolations and fuzzy overlay. The study result in the form of raster-based physical water suitability maps for green mussel cultivation are intended to refine the uncertainties in the vector-based data presentation on water quality data so that it is expected to provide additional information to avoid a less optimal cultivation environment so it maintain the quality of green mussel products and support to accelerate aquaculture production raising program (minapolitan) in Lampung Bay.
Seagrass beds with various species are widespread in Indonesia, where one of them is on Parang Island. The important role of seagrass as a blue carbon sink makes the composition of species and carbon stocks of seagrass on Parang Island need to be mapped. PlanetScope image is one image that is expected to be able to map biophysical information on seagrass beds. The objectives of this study are (1) to map the distribution of composition of seagrass species and (2) to map the seagrass above-ground carbon stock (AGC) on Parang Island, Karimunjawa Islands using PlanetScope. The composition of seagrass species was obtained through multispectral classification (maximum likelihood, random forest, support vector machine) and AGC seagrass through empirical modeling. The class composition of seagrass species obtained was Cymodocea rotundata (Cr), Enhalus acoroides (Ea), Thalassia hemprichii (Th), and EaThCr, with the accuracy of 32.16%. Seagrass AGC empirical modeling has an R2 0.086. The DII23 water column corrected band has the highest accuracy for seagrass AGC mapping, which is 66.90% with a Standard Error of Estimate (SE) value of 4.78 gC/m2 . The total estimated AGC of seagrass found on Parang Island is 50.15 t C.
Coral reef live percent cover (LPC) mapping has always been a challenging application for remote-sensing. The adoption of machine-learning algorithm in remote-sensing has opened-up the possibility of mapping coral reef at higher accuracy. This paper presents the application of machine-learning regression in the empirical modeling of coral reef LPC mapping. Stepwise regression, Support Vector Machine (SVM) regression, and Random Forest (RF) regression were used model the percentage of live coral cover in optically shallow water of Parang Island, Central Java, Indonesia using field photo-transect data to train the PlanetScope image. PlanetScope multispectral bands were transformed into water column corrected bands, Principle Component bands, and Cooccurrence texture analysis bands to be used as predictors in the regression process. The results indicate that the accuracy of machine learning algorithm to map coral reef LPC is relatively low due to the radiometric quality issue in the PlanetScope image (RMSE = 15.43%). We could not yet fairly justify the performance of machine learning algorithm until we applied the algorithms in other images.
Mangrove species inventory and mapping is very important as an effort to preserve the ecosystem and biodiversity of mangrove forests. One way of efficient mangrove species inventory and mapping is to use remote sensing imagery, especially through the analysis of its spectral reflectance pattern. This study aims to map the fourteen mangrove species on Karimunjawa Island, Central Java, Indonesia by: (1) measuring the mangrove species spectral reflectance pattern in the field, (2) characteristic analysis of the mangrove species reflectance pattern, and (3) mapping the dominant mangrove species distribution. The spectral reflectance measurement of mangrove species objects in the field was done by using JAZ EL-350 VIS-NIR (ranges from 300 to 1100 nm). The JAZ field spectrometer was pointed at a distance of 2 cm from the target objects with 10 reading repetitions for each species. Field measurements results were then taken to the laboratory for analysis of spectral reflectance and absorbance patterns, which served as key object recognition in this study. To combine the field and image spectral reflectance patterns, the field reflectance patterns were resampled to the spectral resolution of WorldView-2 image (8 bands, 2 m pixel size). The spectral angle mapper (SAM) method was the used to locate and map the distribution of each targeted mangrove species. As expected, the results showed that the largest difference of spectral curves between species was at the NIR wavelength spectrum (700-900nm). Hence, it is potential to be used as the basis for identification of species mangrove from remote sensing imagery. However, the result of this mapping approach only showed a low accuracy of 62%. The low value of map accuracy was attributed to the inaccuracy in defining threshold in SAM for each class. This study provides a basic understanding of the use of spectral reflectance for mangrove species mapping from remote sensing imagery.
Characterization of seagrass spectral reflectance response is important to understand seagrass condition and for
the possibility of mapping activities using remote sensing data, which is important for the management,
monitoring, and evaluation of seagrass ecosystem. This paper presents the spectral reflectance response of
several tropical seagrass species. These species are Enhalus acoroides (Ea), Thalassia hemprichii (Th) and
Cymodocea rotundata (Cr). Spectral reflectance response of healthy seagrass, epiphyte-covered seagrass, and
damaged seagrass leaves for each species were measured using Jaz EL-350 field spectrometer ranged from 350 -
1100 nm. Repeated measurements were performed above water on harvested seagrass leaves. The results
indicate that there is a change in spectral reflectance response of damaged or epiphyte-covered seagrass leaves
compared to the healthy leaves. The results show similar pattern for the three species, where the peak
reflectance in visible wavelengths shifted toward longer wavelengths on damaged seagrass leaves. The results of
this research open up a possibility of mapping seagrass health condition using remote sensing image.
Mangrove forest is an important ecosystem located in coastal area that provides various important ecological and
economical services. One of the services provided by mangrove forest is the ability to act as carbon sink by sequestering
CO2 from atmosphere through photosynthesis and carbon burial on the sediment. The carbon buried on mangrove
sediment may persist for millennia before return to the atmosphere, and thus act as an effective long-term carbon sink.
Therefore, it is important to understand the distribution of carbon stored within mangrove forest in a spatial and temporal
context. In this paper, an effort to map carbon stocks in mangrove forest is presented using remote sensing technology to
overcome the handicap encountered by field survey. In mangrove carbon stock mapping, the use of medium spatial
resolution Landsat 7 ETM+ is emphasized. Landsat 7 ETM+ images are relatively cheap, widely available and have
large area coverage, and thus provide a cost and time effective way of mapping mangrove carbon stocks. Using field
data, two image processing techniques namely Vegetation Index and Linear Spectral Unmixing (LSU) were evaluated to
find the best method to explain the variation in mangrove carbon stocks using remote sensing data. In addition, we also
tried to estimate mangrove carbon sequestration rate via multitemporal analysis. Finally, the technique which produces
significantly better result was used to produce a map of mangrove forest carbon stocks, which is spatially extensive and
temporally repetitive.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.