Using Shipping Data to Improve Indonesia’s Port Network Connectivity
Part of Indonesia is favourably positioned in the Strait of Malacca, a maritime gateway for trade to and from Asia. The majority of Indonesia’s export commodities are transported by sea, which makes port network connectivity paramount for the archipelagic nation’s economic growth. Building on exploratory research that got going at our Research Dive on trade and competitiveness, we worked with Universitas Gadjah Mada to predict the evolution of Indonesia’s maritime network. Here we discuss the evolution in twofold — both with and without the Government of Indonesia’s Tol Laut maritime development master plan aimed at connecting ports in the archipelago to improve logistics efficiency.
Increased port network connectivity promotes efficiency in the transport of goods between different regions, which in turn lowers costs and provides access to new markets. Over the past few years, the Government has been making efforts to upgrade its maritime infrastructure, which includes the development of the Tol Laut master plan.
Automatic Identification System (AIS) is a tracking system used on ships and by vessel traffic services. Alongside its practical application to maritime safety, AIS data can be useful for research on a variety of topics. Analysing data for a 24-month period (January 2016 — December 2017), we set out to develop statistics of vessel activity at Indonesian ports, as well as predict how the maritime network in Indonesia and across the region would evolve in specific scenarios. To complement the AIS data set, the World Port Index which details seaports throughout the world was used to identify each port’s location.
Determining the influence of each port within the Indonesian maritime network, referred to as its betweenness centrality, is important for the development of Tol Laut. Betweenness centrality looks at the shortest path between every pair of ports within a network, whereby ports that have the shortest paths between them will have higher scores. The same method was applied to assess the connectivity between Indonesian ports and other ports in the Asia-Pacific region.
The analysis was conducted in three segments. First, raw AIS data was mined to model the maritime network in Indonesia and the region, creating an origin destination (O/D) matrix. The second segment of the research examined nuanced activities within Indonesia’s port network as well as the Asia-Pacific port network, also using the O/D matrix, to develop respective profiles that include information on the location of ports, the connections between ports and average path travel times.
Third, a predictive model was then developed to explore how Indonesia’s port network would evolve with and without Tol Laut, where the 2016 AIS data was used to train the model and the 2017 data was used for validation. The model displayed a 0.99 overall precision performance with a 0.92 AUC accuracy score (a score of 1 means absolute accuracy).
The analysis revealed that Western Indonesia has better port connectivity compared to Eastern Indonesia; almost two-thirds of the ports throughout the country’s archipelago are located on the western end. Based on a measure of betweenness centrality, the islands of Sumatra, Kalimantan and Java in particular have the best connectivity across Indonesia’s maritime network. The research also indicated that Western Indonesia has 81 per cent of the country’s total shipping frequency, with 64 per cent in Java alone. These findings confirmed the assumption that ports located in Java are relatively busier than ports on other islands, further highlighting the economic disparity between Western and Eastern Indonesia.
Indonesia’s maritime network has a high number of intermediate stops within, which means a vessel typically has to make multiple stops as it makes its way from its origin to destination. Fewer intermediate stops mean increased efficiency and decreased costs in maritime transport. After measuring Indonesia’s maritime network diameter, Jakarta, Surabaya, Ujung Pandang and Balikpapan were located at the centre of the maritime network, meaning these were the best connected ports — vessels from these ports can reach any port in Indonesia in fewer than four stops.
With regard to the number of port connections, ports located in the western part of Indonesia had 72 per cent of the total number of connections throughout the country, while ports that are located in Eastern Indonesian ports only account for 28 per cent of the country’s connections. These figures suggest that infrequency in shipments on the eastern end of the nation’s archipelago affect maritime connectivity. The best connected domestic ports mentioned above play a vital role in domestic port network connectivity as they facilitate much of the sea transit between both ends of the archipelago.
From the AIS data set, 646 ports in the Asia-Pacific region were identified. It was observed that only 362 of these ports (56 per cent) had regular weekly shipments, classified as at least one shipment per week. For the purpose of this analysis, the remaining 284 ports were excluded. Within the Asia-Pacific region, Indonesia was found to have 5 per cent network density, which points to the proportion of Indonesia’s potential connections within the network based on connections that are actually present. This figure was comparably lower compared to other archipelagic nations, for example Japan and the Philippines that had 9 per cent and 10 per cent network density respectively. China had the highest network density of 26 per cent. Indonesia’s relatively low density score means transporting goods in the region to and from Indonesian ports will involve multiple transit points. Neighbouring Singapore had a higher density percentage and the highest betweenness centrality score, which confirmed its dominance as a transhipment hub in the region.
Without Tol Laut
The model predicted that if the Tol Laut master plan is not implemented, Surabaya and Ujung Pandang would play a more significant role as the transport hubs within the domestic network, especially in integrating the eastern part of Indonesia. At the end of 2017, Surabaya had a betweenness centrality score of 20 per cent. The model predicted that Surabaya will experience a 6 per cent increase in betweenness centrality, which means 26 per cent of all domestic shipment would transit through Surabaya Port. Furthermore, the results from the model indicated that the existing maritime network configuration cannot drive the Tol Laut master plan due to several missing links between ports. Therefore, the Government of Indonesia would need to intervene with the development of port connections, especially in Eastern Indonesia to fully realise the Tol Laut master plan objectives.
With Tol Laut
The model predicted that the successful implementation of the Tol Laut master plan would play a vital role in improving domestic port connectivity. The plan would also create a significant shift in port intermediacy based on measure of betweenness centrality. For example, Port of Bitung’s (which is located in Sulawesi) betweenness centrality score would increase from 6 to 14 per cent, meaning that almost 14 per cent of domestic shipments would transit through Bitung Port. With regards to the Asia Pacific region, the implementation of Tol Laut would reduce network dependence on Singapore by 8 per cent and on Tanjung Priok by 24 percent. This suggests that successful implementation of the Tol Laut master plan should have the effect of balancing the maritime network within Indonesia. In terms of network efficiency, when Port of Bitung becomes a hub port for Eastern Indonesia, as foreseen by Tol Laut, results of the analysis suggested that this should lead to a reasonable increase in efficiency compared to the current network.
This analysis using AIS data confirms the relevance of Tol Laut to the Indonesian Government’s broader economic development agenda. Moving forward, Pulse Lab Jakarta plans to replicate the analysis covering a broader time frame (preferably using three years of AIS data as this is the typical time frame for maritime network research) and expand the scope of the research from regional to global. To generate more insights that can be useful for policy makers and that can be used to improve efficiency in port operations, we are also considering computing average and standard deviations of delay time (calculated as the time between vessels’ arrival and departure) at all domestic ports and including additional information on vessel classification and types of goods transported. As the global AIS data set already exists, it may prove beneficial for the Government of Indonesia to establish an AIS data unit with nationwide coverage to facilitate further research that can inform policy making related to economic development at the subnational and national levels.
This exploratory research has been undertaken by Pulse Lab Jakarta in collaboration with technical experts from Universitas Gadjah Mada in Yogyakarta using big data analysis and machine learning methods. The research was conducted based on a direct request from the Government of Indonesia, and this blog is based on the technical paper output by the team.
Pulse Lab Jakarta is grateful for the generous support from the Government of Australia.