Forecasting cargo throughput in Portuguese ports Andrea Mainardi Dissertação para obtenção do Grau de Mestre em Engenharia e Arquitectura Naval Orientador: Prof. Tiago Santos Júri Presidente: Prof. Carlos Guedes Soares Orientador: Prof. Tiago Santos Vogal: Prof.a Regina Salvador Julho 2016
108
Embed
Forecasting cargo throughput in Portuguese ports · The aim of this thesis is to forecast the cargo throughput in Portuguese ports using a mix of Multiple Linear Regression (MLR)
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Forecasting cargo throughput in Portuguese ports
Andrea Mainardi
Dissertação para obtenção do Grau de Mestre em
Engenharia e Arquitectura Naval
Orientador: Prof. Tiago Santos
Júri
Presidente: Prof. Carlos Guedes Soares
Orientador: Prof. Tiago Santos
Vogal: Prof.a Regina Salvador
Julho 2016
ii
iii
Acknowledgments
I would like to thank my coordinator, Professor Tiago Santos for the countless hours he dedicated to
me and for the precious help when it looked like there was no chance of solving problems.
I would like to thank also my parents, for the endless support they gave me in all these years of
studying. Thank you Teresa for helping me to maintain a fully functioning mind and body.
And finally thanks to my friends who taught me in these years in Lisbon much more than any
university ever could, you know who you are!
iv
Abstract
Reliable port throughput forecasts are of the utmost importance for ports. Given the high investment
and long time needed to improve the port infrastructure and superstructure, a good balance is
required between port development and expected throughput. An over-dimensioned port will lead to
revenues not covering the capital and operating costs, while an under-dimensioned port will introduce
delays in the process of cargo loading and unloading, discouraging ship-owners to come back to the
port. Adequate port development thus requires reliable cargo throughput forecasts.
The aim of this thesis is to forecast the cargo throughput in Portuguese ports using a mix of Multiple
Linear Regression (MLR) and qualitative considerations. Port throughput data from the last 15 years
is obtained from the various Portuguese port authorities and from Instituto Nacional de Estatìstica
(INE) and is analysed. Economic and industrial indicator are collected from different sources, namely
INE, Banco do Portugal, OECD and PorData, aiming at identifying explanatory variables for observed
cargo throughput in ports. Cargo throughput is split into categories and compared with the economic
and industrial indicators to find similarities. Then a forecast of port throughput over the next 10 years
is presented. Considerations are made about the relation between the various ports of the country
and how they interact, as well as about the interaction between port throughput and economy.
Conclusions are drawn regarding main drivers of cargo throughput increase in Portuguese ports and
forecasts are presented for individual ports, the entire port range and different cargo types.
Keywords: Forecast, cargo throughput, ports, linear regression.
v
Resumo
Dado o elevado investimento a longo prazo necessário para melhorar as infra-estruturas e
superestruturas portuárias, previsões do tráfego confiáveis são de extrema importância para os
portos. É necessário um bom equilíbrio entre o desenvolvimento portuário e o rendimento esperado.
Um porto sobredimensionado conduzirá a receitas que não cobrem os custos operacionais e de
capital, enquanto um porto sob-dimensionado atrasará o processo de carga e descarga,
desencorajando os armadores a voltar ao porto. Um desenvolvimento portuário adequado requer,
assim, previsões do tráfego fiáveis.
O objetivo desta tese é a previsão da movimentação de carga nos portos portugueses, usando uma
mistura de Regressão Linear Múltipla (MLR) e considerações qualitativas. Dados sobre o tráfego dos
últimos 15 anos são obtidos das várias autoridades portuárias Portuguesas e do Instituto Nacional de
Estatística (INE). Esses dados estatísticos são analisados em detalhe. Indicadores económicos e
industriais são recolhidos a partir de diferentes fontes, como o INE, o Banco do Portugal, a OCDE e
Pordata, com o objectivo de identificar as variáveis explicativas para a movimentação de carga
observada nos portos. O trafego é depois dividido em categorias e comparado com os indicadores
industriais e económicos, procurando semelhanças. A seguir, uma previsão do tráfego ao longo dos
próximos 10 anos é apresentada. Considerações são feitas sobre a relação entre portos, assim como
sobre a interação entre a movimentação de carga e a economia. Sao retiradas conclusões sobre as
forças motrizes do aumento da movimentação de carga nos portos, e previsões são apresentadas
para os portos individualmente, para grupos de portos e para os diversos tipos de carga.
Acknowledgments ...................................................................................................................................... iii
Abstract ....................................................................................................................................................... iv
Resumo ....................................................................................................................................................... v
List of Figures ............................................................................................................................................... ix
List of Tables ............................................................................................................................................... xii
List of Acronyms ........................................................................................................................................ xiii
2.2 Literature Review .................................................................................................................................... 8
2.2.1 Applications of traditional methods ...................................................................................................... 8
2.2.2 Comparison of performances and new methods .................................................................................. 9
4. ANALYSIS OF CARGO THROUGHPUT IN PORTUGUESE PORTS ............................................. 21
4.1 Port of Lisbon ........................................................................................................................................ 22
4.2 Port of Leixões ...................................................................................................................................... 24
4.3 Port of Sines .......................................................................................................................................... 27
vii
4.4 Port of Setúbal ...................................................................................................................................... 28
4.5 Port of Aveiro ........................................................................................................................................ 31
4.6 Port of Figueira da Foz .......................................................................................................................... 32
4.7 Port of Viana do Castelo ....................................................................................................................... 33
4.8 Port of Faro ........................................................................................................................................... 35
4.9 Port of Portimão ................................................................................................................................... 36
4.10 Port throughput recap .......................................................................................................................... 36
5. METHODOLOGY FOR CARGO THROUGHPUT FORECASTING ............................................... 44
5.2 Multiple Linear Regression ................................................................................................................... 47
5.3 Linear Interpolation .............................................................................................................................. 49
6.5 Wood, cork and paper products ........................................................................................................... 55
6.6 Coal and oil products ............................................................................................................................ 56
6.7 Chemical products ................................................................................................................................ 57
6.8 Non-metallic mineral products ............................................................................................................. 58
6.10 Transport material ................................................................................................................................ 61
6.11 Secondary raw materials ...................................................................................................................... 62
FIGURE 4 - MAIN PORTS AND INDUSTRIES OF PORTUGAL ................................................................................... 13
FIGURE 5 - MAP OF THE PORT OF LISBON ............................................................................................................ 16
FIGURE 6 - MAP OF THE PORT OF LEIXÕES ........................................................................................................... 16
FIGURE 7 - MAP OF THE PORT OF SINES ............................................................................................................... 17
FIGURE 8 - MAP OF THE PORT OF SETÚBAL ......................................................................................................... 18
FIGURE 9 - MAP OF THE PORT OF AVEIRO ............................................................................................................ 18
FIGURE 10 - MAP OF THE PORT OF FIGUEIRA DA FOZ .......................................................................................... 19
FIGURE 11 - MAP OF THE PORT OF VIANA DO CASTELO ...................................................................................... 19
FIGURE 12 - CRUISE QUAY IN PORTIMÃO ............................................................................................................. 20
FIGURE 13 - PORT OF FARO .................................................................................................................................. 20
FIGURE 14 - DRY BULKS THROUGHPUT IN THE PORT OF LISBON ......................................................................... 22
FIGURE 15 - LIQUID BULKS THROUGHPUT IN THE PORT OF LISBON .................................................................... 22
FIGURE 16 - GENERAL CARGO THROUGHPUT IN THE PORT OF LISBON ............................................................... 23
FIGURE 17 - CONTAINER THROUGHPUT IN THE PORT OF LISBON ....................................................................... 23
FIGURE 18 - CRUISE PASSENGERS THROUGHPUT IN THE PORT OF LISBON ......................................................... 24
FIGURE 19 - DRY BULKS THROUGHPUT IN THE PORT OF LEIXÕES ........................................................................ 24
FIGURE 20 - LIQUID BULKS THROUGHPUT IN THE PORT OF LEIXÕES ................................................................... 25
FIGURE 21 - GENERAL CARGO THROUGHPUT IN THE PORT OF LEIXÕES .............................................................. 25
FIGURE 22 - CONTAINER THROUGHPUT IN THE PORT OF LEIXÕES ...................................................................... 26
FIGURE 23 - CRUISE PASSENGERS THROUGHPUT IN THE PORT OF LEIXÕES ........................................................ 26
FIGURE 24 - DRY BULKS THROUGHPUT IN THE PORT OF SINES ........................................................................... 27
FIGURE 25 - LIQUID BULKS THROUGHPUT IN THE PORT OF SINES ....................................................................... 27
FIGURE 26 - GENERAL CARGO THROUGHPUT IN THE PORT OF THE PORT OF SINES ........................................... 28
FIGURE 27 - CONTAINER THROUGHPUT IN THE PORT OF SINES .......................................................................... 28
FIGURE 28 - DRY BULKS THROUGHPUT IN THE PORT OF SETÚBAL ...................................................................... 29
FIGURE 29 - LIQUID BULKS THROUGHPUT IN THE PORT OF SETÚBAL ................................................................. 29
FIGURE 30 - GENERAL CARGO THROUGHPUT IN THE PORT OF SETÚBAL ............................................................ 30
FIGURE 31 - CONTAINER THROUGHPUT IN THE PORT OF SETÚBAL ..................................................................... 30
FIGURE 32 - RORO THROUGHPUT IN THE PORT OF SETÚBAL .............................................................................. 31
FIGURE 33 - DRY BULK THROUGHPUT IN THE PORT OF AVEIRO .......................................................................... 31
FIGURE 34 - LIQUID BULK THROUGHPUT IN THE PORT OF AVEIRO ..................................................................... 32
FIGURE 35 - GENERAL CARGO THROUGHPUT IN THE PORT OF AVEIRO .............................................................. 32
x
FIGURE 36 - DRY BULKS THROUGHPUT IN THE PORT OF FIGUEIRA DA FOZ......................................................... 33
FIGURE 37 - GENERAL CARGO THROUGHPUT IN THE PORT OF FIGUEIRA DA FOZ............................................... 33
FIGURE 38 - DRY BULK THROUGHPUT IN THE PORT OF VIANA DO CASTELO ....................................................... 34
FIGURE 39 - LIQUID BULK THROUGHPUT IN THE PORT OF VIANA DO CASTELO .................................................. 34
FIGURE 40 - GENERAL CARGO THROUGHPUT IN THE PORT OF VIANA DO CASTELO ........................................... 35
FIGURE 41 - DRY BULK THROUGHPUT IN THE PORT OF FARO .............................................................................. 35
FIGURE 42 - GENERAL CARGO THROUGHPUT IN THE PORT OF FARO .................................................................. 36
FIGURE 43 – CRUISE PASSENGER THROUGHPUT IN THE PORT OF PORTIMÃO .................................................... 36
FIGURE 52 - ENERGY CONSUMPTION IN PORTUGAL (TEP)................................................................................... 45
FIGURE 53 - PORTUGUESE GDP AND DOMESTIC CONSUMPTION (CONSTANT PRICE 2014) ............................... 46
FIGURE 54 - GDP OF THE MAIN ECONOMIC PARTNERS OF PORTUGAL (CONSTANT PRICE 2014) ....................... 46
FIGURE 55 - COMPARISON OF TEU THROUGHPUT BETWEEN SINES AND THE MAIN TRANSSHIPMENT PORTS OF
THE WESTERN MEDITERRANEAN. THE NUMBER IN THE LEGEND INDICATES THE THROUGHPUT CAPACITY
OF THE PORT IN TEUS/YEAR. ........................................................................................................................ 47
FIGURE 56 - FORECAST OF THROUGHPUT IN PORTUGUESE PORTS - LOADED TONS OF PRODUCTS OF FOREST
AND AGRICULTURE ....................................................................................................................................... 51
FIGURE 57 - FORECAST OF THROUGHPUT IN PORTUGUESE PORTS - UNLOADED TONS OF PRODUCTS OF FOREST
AND AGRICULTURE ....................................................................................................................................... 51
FIGURE 58 - FORECAST OF THROUGHPUT IN PORTUGUESE PORTS - UNLOADED TONS OF CRUDE OIL AND LNG
FIGURE 59 - FORECAST OF THROUGHPUT IN PORTUGUESE PORTS – LOADED TONS OF MINERALS ................... 53
FIGURE 60 - FORECAST OF THROUGHPUT IN PORTUGUESE PORTS – UNLOADED TONS OF MINERALS .............. 53
FIGURE 61 - FORECAST OF THROUGHPUT IN PORTUGUESE PORTS – LOADED TONS OF FOOD PRODUCTS ........ 54
FIGURE 62 - FORECAST OF THROUGHPUT IN PORTUGUESE PORTS – UNLOADED TONS OF FOOD PRODUCTS .. 54
FIGURE 63 - FORECAST OF THROUGHPUT IN PORTUGUESE PORTS – LOADED TONS OF WOOD, CORK AND
PAPER PRODUCTS ......................................................................................................................................... 55
FIGURE 64 - FORECAST OF THROUGHPUT IN PORTUGUESE PORTS – UNLOADED TONS OF WOOD, CORK AND
PAPER PRODUCTS ......................................................................................................................................... 56
FIGURE 65 - FORECAST OF THROUGHPUT IN PORTUGUESE PORTS – LOADED TONS OF OIL PRODUCTS ............ 57
xi
FIGURE 66 - FORECAST OF THROUGHPUT IN PORTUGUESE PORTS – UNLOADED TONS OF COAL AND OIL
Neural networks have some advantages over other prediction methods, first of all they’re self-training,
this means that what happens in the hidden layer is not defined by the operator, and the network
refines himself iteration after iteration. This also brings a flexibility not present in other forecasting
techniques, on every set of data on which the network is used the method tailors itself to the needs of
the time series.
Once a prevision has been made there is the need for a benchmarking tool to attest the quality of the
prevision. This tool is the mean forecasting error. Usually in the development of a new method a test
time series is analysed, the method is then applied to the first 80% of the data, and the remaining
20% is used to assess the goodness of the forecast.
There are different formulas to calculate errors, here is a short overview on the most used ones:
Mean percentage error, MPE, the average of the percentage errors:
𝑴𝑷𝑬 = 𝟏𝟎𝟎
𝒏∑
𝒂𝒊 − 𝒇𝒊
𝒂𝒊
𝒏
𝒊=𝟏
(2)
where n is the sample size, a the actual value and f the forecasted value;
Mean absolute percentage error, MAPE, same as above, just the absolute value of the error
is taken:
𝑴𝑨𝑷𝑬 = 𝟏𝟎𝟎
𝒏∑ |
𝒂𝒊 − 𝒇𝒊
𝒂𝒊| (3)
𝒏
𝒊=𝟏
Mean square error, MSE, the average of the square error
𝑴𝑺𝑬 = 𝟏
𝒏∑(𝒂𝒊 − 𝒇𝒊)
𝟐 (4)
𝒏
𝒊=𝟏
8
Root of the mean square error, RMSE, the square root of the above, one of the most used
𝑹𝑴𝑺𝑬 = 𝟏
𝒏√∑(𝒂𝒊 − 𝒇𝒊)
𝟐
𝒏
𝒊=𝟏
(5)
In 2006 Rob Hyndman [4] proposed a new error calculation that should overcome some conceptual
problems in other error calculations, the Mean Absolute Scaled Error (MASE)
𝑴𝑨𝑺𝑬 = 𝟏
𝒏∑ (
|𝑒𝑡|
1𝑛 − 1
∑ |𝑌𝑖 − 𝑌𝑖−1|𝑛𝑖=2
)
𝒏
𝒊=𝟏
=∑ |𝑒𝑡|𝑛
𝑡=1
𝑛𝑛 − 1
∑ |𝑌𝑖 − 𝑌𝑖−1|𝑛𝑖=2
(6)
Where et is the forecast error, and at the denominator there is the error of the naïve series, if this error
is bigger than 1 then the forecast is less precise than the simple naïve method.
Each error is used in different applications, however one of the most used ones is the RMSE.
2.2 Literature Review
Literature is full of examples of different application of forecasting methods, this section shows the
main applications to port cargo throughput and freight rates forecasting. The technique used for the
forecast in this thesis will then be explained in chapter 5. Such technique will consist of a long time
forecast of ports throughput in Portugal, considering the different classes of cargo (container, general
cargo, ro-ro, dry and liquid bulk), categories (crude oil, cement, metals…) as well as differentiating
export and import (when possible).
This literature review will be structured as follows. Firstly an overview of specific applications of the
methods shown in the previous subchapter (Time series, MLR and ANN). Then the focus will be
moved on papers that apply different methods and study their relative performance, as well as
publications that employ a mixture of methods to reach a result. Afterwards there will be some general
suggestions, insights and in-depth considerations provided by various sources (private publications,
books). Finally a resume is made.
2.2.1 Applications of traditional methods
Regression analysis is one of the most used methods, even so some things can be done to improve
it, for example it is noted in [5] that the relation can change itself during the years, different stages in
the economic development of a country can change the nature of the imports and exports, in their
case the economic development of Taiwan changed the content of the containers throughout the
years, moving from bulky and cheap basic resources to highly refined expensive ones, thus the
amount of TEUs moved by each unity of GDP changes, this nonlinearity can be taken into account to
create a more precise forecast.
Artificial neural networks are broadly used. There are many examples in literature in which some
advice for the work can be found. In [6] it is shown that the performance of neural networks is higher
on longer term forecast, the researchers also suggest that a short term forecast should use a more
9
complicated model with less variables (less input nodes and more hidden nodes) while a long term
forecast should take into account more variables, but analyse them with a simpler network.
References [7], [8] show one of the few application of neural networks to forecast the traffic in ports
while most researches are about freight rates, like for example [9]. Their application is solid, and the
confrontation with linear regression analysis shows how neural networks are more precise when
making long term forecast. ANNs are also used to forecast traffic, like for example in Suez canal [10].
Or to forecast the traffic for a specific terminal, like in [11].
In [12] researchers made a study on the applicability of neural networks, checking literature about
traffic forecast in the past 20 years, it can be seen how neural networks are having a really good
reception even if they are relatively new methods.
Time series forecast is also used, in [13] different methods are applied to the same problem to check
their relative performance, these models are: decomposition model, regression model with seasonal
dummy variable, grey model, hybrid grey model and SARIMA. ARIMA methods are also used in [14]
to forecast freight rates in the dry bulk market.
2.2.2 Comparison of performances and new methods
As shown in [8], ANNs have an advantage over regression analysis when considering container
throughput in Bangkok port, when comparing the actual data with the forecast they get a correlation
coefficient of 0.8620 with linear regression and 0.9509 with ANN. It is worth noting though that linear
regression needs much less data to perform a good forecast, and is also much simpler and faster.
Some authors take another different approach, like in [15], where instead of just focusing on the port
economy the whole macro economy of the hinterland of an extended complex of ports is modelled,
including a model of the logistic layer of the chain. Such a comprehensive approach has some
advantages regarding the overall economy of the hinterland, but it is very scarce regarding the
situation in each and every port of the range.
When using causal methods it is important to choose wisely the variables. Reference [16] presents a
forecast of port traffic in Portugal using GDP to explain most of the cargo throughput, and the
country’s energy consumption to forecast the import of fossil fuels (oil and coal). Some other authors,
like in [17] developed a specific mathematical tool to relate the demand of import and export to
macroeconomic variables.
Many paper use also the so called System Dynamics method, which is a mixture of Linear Regression
with Bayesian networks, examples of these can be found in [18] where it is used to forecast the freight
rates of capsize dry bulk carriers. Another alternative way is the commodity based approach, where
the economics of the hinterland are also considered and modelled, like for example the South African
example in [19]. This approach is similar to the CDE-MPR used in [20] to forecast container
throughput in the ports of the region of the Pearl River delta in China. Another different approach
(error correction model) is used in [21] to forecast the container throughput in Hong Kong.
Considering the incredible amount of variables acting on the shipping system it should be
acknowledged, as in [22], that this is a chaotic system, and thus give up on the idea of using linear
10
models to explain it and instead focus on models that take this chaotic nature into account. Another
way of taking this chaotic behaviour into considerations is by using fuzzy time series like in [23].
2.2.3 In-depth considerations
Some authors focused on the interaction between some parameters of the model, namely the number
of observations used to fit the model and the forecast horizon. Nielsen et al. shown [24] that model fit
and forecast performance cannot be achieved at the same time. Specifically, when trying to make a
model that fits a higher amount of past data the quality of forecast decreases, while a model that
provides a quality forecast comes with an inferior correlation to past data. Thus when forecasting it is
of primary importance to use a well-known and stable method, and blindly increasing the amount of
past data thinking that this will lead to a more precise long-term forecast is counterproductive.
In [3] two interesting ideas are presented, the first is that shipping time series do not follow normal
distribution, when analysing long series of past data it can be seen how the values deviate from the
norm more than 3 standard deviations many more times than expected from a Gaussian distribution.
Statistical data presents “fat tails” which implies the presence of kurtosis and skewness. Also in this
paper short cycle (periodic and non-periodic) variations in freight rate are explained with V-statistic. It
is worth saying that freight rate and port throughput are completely different (even if related) set of
data.
The variables they depend on are different so it is not sure that these conclusions can be applied to
port forecasting.
One of the main problems when forecasting a long-term situation is the inevitable presence of strongly
nonlinear, non-predictable economic shocks (such as the crisis of 1974 and 2008). Many other
variables enter in the picture, such as laws and political decisions. An exhaustive report [25] from
MDS Transmodal about the forecast of traffic in UK ports is available and allows good insights about
what to consider when forecasting long term-situations.
A deep knowledge is needed about the country’s energetic policy, including the amount of oil,
coal and gas used for various purposes (energy generation, refineries), the situation of
national reserves as well as the existence of environmental protective laws and the impact
that they will have in the future. A balance has to be made forecasting the need to import.
Market studies about the situation of the car market, one of the main drivers of Ro-Ro traffic.
The situation of agriculture, to have a better overview on the dry bulk traffic.
The economic agreement with neighbouring countries, to know how much of traffic directed in
other countries can be handled by the country’s ports.
An interesting study has been carried out [26] about the evolution of neighbouring ports in East Africa,
it is shown that in modern times one of the most influencing variables on the growth of one port
instead of another is simply the inland connectivity, ports which are more easy to access via a
different array of transports (road, train, inland waterways) have a much higher development potential.
When forecasting it is important to achieve a balance between the amount of past data used and the
forecasting horizon, a number of papers have been analysed.
11
In a private study on the port of Vancouver , [27], 23 years of past data are used to forecast 35 years
in the future. A scenario-based study about the ports in the Baltic region , [28], uses 12 years of data
to predict 8. In a Vietnamese study using ARIMA, [29], 20 years of past data are used to forecast
traffic for the 6 years to come. Another paper, [5], uses a modified regression analysis, in which 13
years of past data are used to forecast 5 future years.
A couple of studies analysing the throughput of the Hamburg-La Havre range, [15], [30], use a
combination of methods to make previsions, in both cases the time series extend in the past as much
as they do in the future.
Even with these differences, most or the studies analysed forecast the time window analysed is
symmetrical, the number of years ahead is equal to the years abaft.
Two classic books from the 80’s analyse the process of port traffic forecasting and port planning, [31]
and [32], they both insist on the importance of having a deep knowledge of the hinterland of any port
before attempting any forecast, knowing exactly where the goods are coming from and where they are
going is fundamental to attempt a decent forecast. They also note that any port that is trying to
develop a forecast usually uses a combination of basic methods, tailored on the data available and
the characteristics of the hinterland. Given the high sensibility of port throughput to one-of-a-kind,
unpredictable economic events it is wise to prepare some different future scenarios, guessing the
most important socio-politic developments of the future years. These scenarios will then influence the
forecast in different ways.
2.2.4 Summary
Summarizing, the field of publications about forecasting looks wide and fragmented.
Many studies however focus on forecasting freight rates, or other variables which are much more
volatile than port throughput, for example [3], [9], [10], [14], [18], [22]–[24].
Looking at the studies who concentrate on port throughput, most of them are taking into account only
containerized cargo, [5], [7], [8], [13], [27], [29], [19]–[21], [11], [33]–[35]. This focus on containerized
cargo is due to the homogeneity of it. Single container act as unitary cargo, and most authors don’t
even take into consideration the content of each container, the only exception being [5]. This
approach, however simplistic, is effective due to the big variety of cargos shipped via container.
A core group of techniques are widely used and remixed, this includes regression, ARIMA and neural
networks, and each method is applied preferably on series with different characteristics. Most
research papers focus on taking one of these well-known techniques and tailor them to certain
situations (for example [5]), or compare the performance of different methods when applied to the
same problem (for example [8], [13]).
Overall the publications analysed belong mostly to two main groups:
Private firm studies, usually commissioned by port authorities, shipping companies or terminal
managements. These studies are usually not clear about the techniques used to make the
forecast, and focus instead on the surrounding conditions to give a context to the forecast,
this is most likely due to a blend of qualitative and quantitative methods, making the
explanation of the method difficult.
12
Research papers, these studies are very clear about the mathematics behind the forecast.
They focus on modifications of well-known forecasting techniques and their validation and
comparison with different other approaches. Usually these studies don’t give forecasts per-se,
instead the time series analysed are used both to create and to validate the method.
The technique used for this work will be explained in chapter 5, after giving an overview of the
situation of the sector in Portugal.
13
3. GEOGRAPHICAL AND INDUSTRIAL OVERVIEW
The geographical and industrial conformation of Portugal is analysed, to have a deeper understanding
of where the ports are, how they relate to each other and with their hinterlands.
Ports located in the Portuguese west coast have been collectively called “Portuguese Range” [36],
they constitute a multi-port gateway region, situated at the far west end of Europe. They have the
potential to be the gateway for cargo directed towards Western Europe, sitting at the extreme of the
European rail freight corridor n°4. In the past years the throughput of the Portuguese ports grew,
together with the connections between Portugal and Spain.
In Figure 4 a map of Portugal with the analysed ports and industries is shown. The industries
presented on the map are just the main ones of continental Portugal, in each port section a more in-
depth description of the industries present in the hinterland of each port is given.
Figure 4 - Main ports and industries of Portugal
14
It is evident how the range can be divided in 3 groups: Northern ports, Leixões, Viana do Castelo,
Aveiro and Figueira da Foz; Central ports, Lisbon, Setúbal and Sines; Southern ports, Portimão and
Faro.
3.1 Industrial Overview
The main industries present in Portugal are: refineries, cement factories, steel mills, paper factories,
one automobile factory and various power plants.
A list of the power plants running on fossil fuels is shown in Table 1.
Table 1 - List of Portuguese power plants (source:Wikipedia)
Station District Capacity Primary fuel
Lares Power Station Coimbra 826 MW Natural gas Pego I Power Station Santarém 576 MW Coal Pego II Power Station Santarém 837 MW Natural gas Ribatejo Power Station Lisbon 1176 MW Natural gas Sines Power Station Setúbal 1180 MW Coal Tapado do Outeiro II Power Station Porto 990 MW Natural gas Tunes Power Station Faro 165 MW Diesel Barreiro Cogeneration Station Setúbal 64.5 Fuel oil (Cogeneration)
The two refineries of the country are located in Sines and Matosinhos, they are both managed by
Galp. The one in Sines produces: Gasoline; diesel; LPG (liquefied petroleum gas); Fuel oil; naphtha
(used in the petrochemical industry to produce polymers from which plastic, fibres for textiles and
even bubble gum is produced); jet fuel (fuel for airplanes); bitumen (for asphalt and insulate); sulphur
(for pharmaceutical products, farming and pulp whitening). It has a distilling capacity of 10.9 million
tons per year, or 22 thousand barrels per day. The refinery in Matosinhos produces: Fuel oil; base oil;
aromatics; solvents; greases; paraffin; bitumen and sulphur. It has a production capacity of 4.46
million tons per year. (source: galpenergia.com)
Close to the port of Setúbal there is a Volkswagen Autoeuropa factory, the factory was previously
owned by a joint venture between Ford and Volkswagen, eventually in 2008 Ford left the factory and
production declined. Nowadays the factory produces cars almost up to it capacity of 172500 cars per
year. (source:Wikipedia)
An important industrial sector for Portugal is cement. There are various companies producing cement
in the country, the main ones are Secil and Cimpor. Secil has 3 production complexes, one in Setúbal,
producing 2 million tons per year, one in Leiria producing 1.35 million tons per year and one in
Alcobaça producing 380 thousand tons per year. (source:secil.pt) Cimpor has several factories
around the country and abroad, one in Loule producing 350 thousand tons per year, and other three
in Lisbon, Figueira da Foz and Coimbra.
Also paper production plays an important role in Portugal, for the economy in general as well as for
ports. There are 2 big producers of paper in the country: Portucel Soporcel (now called Navigator
Company) and Altri. Portucel Soporcel is one of the biggest paper producer in Europe, it manages 3
15
big factories in Setúbal (510,000 tons/year), Figueira da Foz and Cacia. The total production of the
company is around 1.6 million tons of paper per year (source:thenavigatorcompany.com). Altri has 3
subsidiaries companies that manage paper factories in Portugal: Celbi, Celtejo and Caima. Together
they produce 790 thousand tons of paper and pulp per year. (source:altri.pt)
Steel production in Portugal is done by recycling metal scraps, the procedure is convenient because it
demands much less energy than the operation of a traditional blast furnace. The main steel mills of
the country is managed by Lusosider, it is located in Seixal (between Lisbon and Setúbal), it produces
550 thousand tons of laminated steel per year (source:lusosider.pai.pt).
3.2 Lisbon
Lisbon is the capital of Portugal, the city is located along the northern shore of the estuary of the Tejo
River, and the estuary is very broad, creating a natural bay where the port terminals are distributed. It
is a landlord port, the port authority owns the land on which the terminals are located, but the day-to-
day management of the activities is carried by private companies, every terminal is governed by a
different company. In the northern side there are container, RoRo and general cargo terminals, plus 2
cruise terminals and recreational docks. On the southern side of the river there are several dry and
liquid bulk terminals, as well as some small general cargo ones. Today the expansion of Lisbon’s port
is hindered by the city around it, most terminals are completely surrounded by urban development so
there is no space left where to expand the quays and superstructure. Even so Lisbon is still one of the
main ports of Portugal, handling 14% of the national cargo.
Two of the container terminals of the port have been recently bought by a Turkish company, Yıldırım.
The container throughput in these two terminals have been stagnating for the past 10 year, so a
possible evolution is now in the hands of this company.
16
Figure 5 - Map of the port of Lisbon
Lisbon works as a hub for the many industries located along the two sides of the Tejo River. Most
heavy industries are located on the southern side of the river (steel mills, chemical factories) while on
the northern side and along the course of the river there are paper and fertilizers factories, as well as
several power plants.
3.3 Leixões
Leixões is the main port of northern Portugal, it is situated on the Atlantic coast 4 km north of the
estuary of the Douro River, where the city of Porto is. It is an artificial landlord port, in the bay all the
terminals are distributed: dry and liquid bulks, containers, general cargo, RoRo and cruise. Leixões is
the second national port, handling 24% of the national cargo. Leixões’s container terminal is now
been used at its maximum capacity, expansion work are already being carried out. Leixões is one of
the Portuguese ports that has been growing more in the past years.
Figure 6 - Map of the port of Leixões
The northern region of Portugal is densely populated. In the hinterland of Leixões there are several
industries: a steel mill, a refinery, paper factories, some caves and wood and cork harvesters.
Leixões is at the end of the navigable Douro waterway, thus acts as a hub for all the industries located
along its course, it is important noting that the Douro waterway comprises a good part of Spanish
territory.
3.4 Sines
Sines is a port located in the south of Portugal, along the Atlantic coast. It is an artificial port, it came
into operation only in 1978. Given the natural deep waters surrounding the port it is the port of choice
for the bigger vessels docking in Portugal. The container terminal, opened in 2004, is the most
important in Portugal, its importance as a transshipment hub is rapidly growing. Currently Sines
handles 45% of the national cargo.
17
Sines is the main energetic hub of Portugal, with the only coal power plant of the country located
close by and big refineries and chemical production industries also around the city. In its hinterland
there are several marble and copper caves, as well as paper factories.
Figure 7 - Map of the port of Sines
3.5 Setúbal
Setúbal is a city located 40 kilometres south of Lisbon, it sits at the estuary of the Sado river. The port
area develops between 2 protected natural parks, but even so this port has plenty of space to account
for future development. It is a mixed landlord/private port, with most terminals owned by the port
authority and some smaller ones privately owned. Upstream from the port there is also the Lisnave
shipyard, once the biggest European shipyard, today it is mainly used for reparation. The port area
extends over the northern shore of the Sado, except the SECIL cement terminal which is right at the
estuary of the river. This port has very good intermodal transport capability due to its position along
the north-south and east-west axes. In 2014 the port handled 10% of the national cargo.
Setúbal, given its closeness with the AutoEuropa factory, has the biggest RoRo terminal of the
country, which is able to dock the biggest car carriers existent today. Various cement factories are
present around Setúbal, as well as steel mills, paper factories and copper and marble mines in the
hinterland.
18
Figure 8 - Map of the port of Setúbal
3.6 Aveiro
Aveiro is a city located 70kms south of Porto, the port is located in the inland lagoon of Ria de Aveiro,
it is a protected natural area, and thus port development has to be extremely careful. It is the most
recent port infrastructure of the country, thus it is well organized without congestion issues.
Figure 9 - Map of the port of Aveiro
In the hinterland of the port there are mainly cement factories, chemical refineries, paper and glass
factories.
3.7 Figueira da Foz
Figueira da Foz is a city located in between Porto and Lisbon, the port develops around the estuary of
the Mondego river, it is mostly a local spoke port, dedicated to dry bulks and general cargo.
19
Figure 10 - Map of the port of Figueira da Foz
The main factories in the hinterland are paper, glass and cement producers.
3.8 Viana do Castelo
Viana do Castelo is a small town in the far north of Portugal, situated at the end of the Lima River it is
mostly a local port, handling mostly general cargo and dry bulks, as well as some liquid bulks.
The industries presents in the hinterland of Viana are a paper factory, a mine and a factory that
produces wind turbines.
Figure 11 - Map of the port of Viana do Castelo
3.9 Southern Ports
Algarve is the southernmost region of continental Portugal, there are two small ports along the south
coast, one in Portimão and one in Faro.
20
Portimão has a cruise quay, shown in Figure 12, given the high touristic value of the region it is a
heavily-used terminal.
Faro mostly uses its dry bulk terminal, shown in Figure 13, to export cement produces in the nearby
factory.
Figure 12 - Cruise quay in Portimão
Figure 13 - Port of Faro
21
4. ANALYSIS OF CARGO THROUGHPUT IN PORTUGUESE
PORTS
To perform the forecast a variety of data was gathered from different sources, in this chapter the
sources will be shown and an explanation of the data will be given, port by port, industry by industry.
Data gathered includes:
Tons loaded and unloaded, in the main ports of Portuguese mainland (Lisbon, Leixões, Sines,
Setúbal, Aveiro, Viana do Castelo, Figueira da Foz and Faro), subdivided in categories (dry
bulks, liquid bulks, general cargo, containers and ro-ro), as well as the cruise passengers,
where present;
The main categories of cargo handled in the different ports, to have a deeper understanding
of which are the main drivers of port throughput;
Econometric indicators, related to the Portuguese (and world) economy, like Portuguese
GDP, population, inflation, domestic consumption and finally world GDP;
Industrial indicators, the yearly performances of different sectors of the industry related to the
main goods traded in ports, this includes tourism, alimentary, metallurgic, cement and glass,
petroleum and chemical industry, as well as the production of vehicles and the yearly
production of electricity, subdivided by the different sources of energy.
Data was gathered from different sources:
Instituto Nacional de Estatistica (INE) [37], provides yearly publications called Estatisticas dos
Transportes e Comunicaçoes and Estatisticas da Produçao Industrial. From here it was
retrieved the amount of tons loaded and unloaded, for each port, for each category as well as
the industrial indicators, INE publications go from 2001 to 2014.
Port authorities publish a yearly account of the port throughput, all the available information
from the past years was gathered, given the individuality of each port authority the data is not
homogeneous, for some ports only 5 years of data is available, for some others there is no
distinction between loaded and unloaded cargo. The ports analysed are Lisbon [38], Leixões
[39], Sines [40], Setúbal [41], Figueira da Foz [42], Aveiro [43], Viana do Castelo [44], Faro
and Portimão [40].
Economic indicators were taken from 3 different websites: PorData [45] a Portuguese data
aggregator, OECD data [46] the database of OECD countries and the International Monetary
Fund [47].
In this section an overview of cargo throughput in each port is given. For extended data on the main
cargos within each type of transportation refer to Appendix B. For each port in consideration only the
most significant data is analysed.
22
4.1 Port of Lisbon
Lisbon is mainly a dry and liquid bulk importer, as well as an important container port. From Figure 14
it can be seen how dry bulk throughput has been more or less stable in the past years. The main
imported dry bulks are cereals (corn, soy, wheat and canola) and scrap metal, together accounting for
84% of the imports in 2014. The main exported dry bulks are cement, sand, fertilizers, forage and
malt, worth 91% of export in 2014.
Figure 14 - Dry bulks throughput in the port of Lisbon
Figure 15 shows the liquid bulk throughput in the port of Lisbon. The throughput decreased around
2004 and 2007 but overall it continued stable since 2001. The main imported liquid bulks are
ammonia, diesel and fuel oil, together accounting for 71% of the imports in 2014. The main exported
liquid bulks are fuel oil, biodiesel and vegetable oils, accounting for 94% of exports in 2014.
Figure 15 - Liquid bulks throughput in the port of Lisbon
4.8 Other food products n.e.c. and tobacco products (except in parcel service or grouped)
4.9 Various food products and tobacco products in parcel service or grouped
79
5 Textiles and textile products; leather and leather products
5.1 Textiles
5.2 Wearing apparel and articles of fur
5.3 Leather and leather products
6 Wood and products of wood and cork (except furniture); articles of straw and plaiting materials; pulp, paper and paper products; printed matter and recorded media
6.1 Products of wood and cork (except furniture)
6.2 Pulp, paper and paper products
6.3 Printed matter and recorded media
7 Coke and refined petroleum products
7.1 Coke oven products; briquettes, ovoids and similar solid fuels
7.2 Liquid refined petroleum products
7.3 Gaseous, liquefied or compressed petroleum products
7.4 Solid or waxy refined petroleum products
8 Chemicals, chemical products, and man-made fibers; rubber and plastic products ; nuclear fuel
8.1 Basic mineral chemical products
8.2 Basic organic chemical products
8.3 Nitrogen compounds and fertilizers (except natural fertilizers)
8.4 Basic plastics and synthetic rubber in primary forms
8.5 Pharmaceuticals and parachemicals including pesticides and other agro-chemical products
8.6 Rubber or plastic products
8.7 Nuclear fuel
9 Other non metallic mineral products
9.1 Glass and glass products, ceramic and porcelain products
9.2 Cement, lime and plaster
9.3 Other construction materials, manufactures
10 Basic metals; fabricated metal products, except machinery and equipment
10.1 Basic iron and steel and ferro-alloys and products of the first processing of iron and steel (except tubes)
10.2 Non ferrous metals and products thereof
10.3 Tubes, pipes, hollow profiles and related fittings
10.4 Structural metal products
10.5 Boilers, hardware, weapons and other fabricated metal products
11 Machinery and equipment n.e.c.; office machinery and computers; electrical machinery and apparatus n.e.c.; radio, television and communication equipment and apparatus; medical, precision and optical instruments; watches and clocks
11.1 Agricultural and forestry machinery
80
11.2 Domestic appliances n.e.c. (White goods)
11.3 Office machinery and computers
11.4 Electric machinery and apparatus n.e.c.
11.5 Electronic components and emission and transmission appliances
11.6 Television and radio receivers; sound or video recording or reproducing apparatus and associated goods (Brown goods)
11.7 Medical, precision and optical instruments, watches and clocks
11.8 Other machines, machine tools and parts
12 Transport equipment
12.1 Automobile industry products
12.2 Other transport equipment
13 Furniture; other manufactured goods n.e.c.
13.1 Furniture
13.2 Other manufactured goods
14 Secondary raw materials; municipal wastes and other wastes
14.1 Household and municipal waste
14.2 Other waste and secondary raw materials
15 Mail, parcels
15.1 Mail
15.2 Parcels, small packages
16 Equipment and material utilized in the transport of goods
16.1 Containers and swap bodies in service, empty
16.2 Pallets and other packaging in service, empty
17 Goods moved in the course of household and office removals; baggage and articles accompanying travellers; motor vehicles being moved for repair; other non market goods n.e.c.
17.1 Household removal
17.2 Baggage and articles accompanying travellers
17.3 Vehicles for repair
17.4 Plant equipment, scaffolding
17.5 Other non market goods n.e.c.
18 Grouped goods: a mixture of types of goods which are transported together
19 Unidentifiable goods: goods which for any reason cannot be identified and therefore cannot be assigned to groups 01-16.
19.1 Unidentifiable goods in containers or swap bodies
19.2 Other unidentifiable goods
20 Other goods n.e.c.
20 Other goods not elsewhere classified
XX Unknown goods
81
82
APPENDIX B – DATA
LOADED CARGO – PRODUCTS OF FOREST AND AGRICULTURE (TONS)