Underground projects all over the world draw attention from the public and the media not only for their complexity and high profile; but more importantly for the reservations surrounding the cost of the projects.
It is not a myth that underground projects generally suffer from cost and time overruns.
It does not, however, necessarily follow that a project suffering from cost overrun must also suffer from time overrun or vice versa.
Researchers, authors, public agencies, investors and investment banks have studied the subject of overruns in publicly funded projects including underground projects. An extensive literature on this subject exists and it expands over time as cost and time difficulties increase too.
It is therefore important to examine the cases from the literature in order to gain an understanding of the level of the overruns and the reasons for them.
Many, if not all, construction projects suffer from cost and time overruns in one way or another. There are many reasons ranging from ground conditions to escalation of material prices, changes in speci-fications to changes in legislation. Flyvbjerg, et. al. (2002) assert that high-speed rail is the most prone to cost underestimation, followed by urban and conventional rail respectively. Similarly, cost underestimation appears to be more frequent for tunnels than for bridges. This suggests that the complexities of technology and geology might have an effect on cost underestimation. Hence, it is clear that the first criteria to check in claiming for an overrun are the estimations or budgets allowed for a particular project. Although, many government agencies and investment bodies have different and complex estimation structures for estimating and budgeting the projects.
Large numbers of agencies, contractors, consultants and experts are involved at all stages of the project and have extensive experience in their fields. A large number of cost and time estimates are carried out before the contract is signed. The client and his experts and consultants, or the contractors, carry out estimates for the same project using all available tools such as historical data, local experience, methodologies, schedules, productivities, innovative solutions, commercial and also any business incentives.
And yet projects frequently run over time and over budget. Flyvbjerg, et. al. (2002) state:
a. In nine out of 10 transportation infrastructure projects, costs are underestimated.
b. For rail projects, actual costs are on average 45 per cent higher than estimated costs
c. For fixed-link projects actual costs are on average 34 per cent higher than estimated costs
d. For road projects, actual costs are on average 20 per cent higher than estimated costs
e. For all project types, actual costs are on average 28 per cent higher than estimated costs
f. Cost underestimation exists across 20 nations and five continents; it appears to be a global phenomenon.
g. Cost underestimation appears to be more pronounced in developing nations than in North America and Europe (data for rail projects only).
h. Cost underestimation has not decreased over the past 70 years. No learning that would improve cost estimate accuracy seems to take place.
i. Cost underestimation cannot be explained by error and seems to be best explained by strategic misrepresentation, i.e. lying.
j. Transportation projects do not appear to be more prone to cost underestimation than are other types of large project.
Table 1, below shows several transportation projects from Europe. The table clearly shows cost overruns between eight per cent and 116 per cent. However, it also shows that unit cost for three railway projects, the Paris to Lille TGV, the Madrid to Seville AVE and the Lyon to Marseilles TGV are between EUR 8.57M (USD 10.8M) and EUR 11.82M (USD 14.94M); In contrast for the ICE Frankfurt to Cologne line is EUR 33.98M (USD 43.8M). Although the cost difference may be due to many other factors, the effect of the ratio of tunnel length to total length (the ICE Frankfurt to Cologne line is 12.3 per cent and while the other three projects range between 3.2 per cent and 3.6 per cent) may also impact on total cost.
Florio (2009) lists the main causes of errors in cost estimation, which include delays in implementation, as changes in project specification and design, changes in quantity and prices, technological risks, underestimation of expropriation costs, changes in safety requirements and changes in environmental requirements. Although geological risk or overruns due to unforeseen conditions are not included in the main causes of errors in this classification, both of these are mentioned as major factors in Flyvbjerg, et. al. (2002).
Another example is from the Neue Eisenbahn-Alpentransversale (NEAT) project in Switzerland, which involves extraordinary tunnels such as Gothard Base Tunnel, Ceneri Base Tunnel and Lotscherg Tunnel. The project also suffered from cost overruns of 53 per cent of which nine per cent is accounted for by geology according to a report published by Swiss Office of Transport (2009).
Figure 1 shows the cost overruns for the tunnels in Seattle Area. According to a report by the Straightline Institute one in four recent tunnels in the Seattle area was completed within budget. The report concludes that the savings for the Mount Baker Tunnel can be attributed largely to two factors. First, labour and materials were cheaper than expected because the US economy experienced a recession after the project costs had been estimated. Second, the soil conditions were relatively well understood in advance of the cost estimates, in part because an adjacent tunnel (the roadway for the original Interstate 90) had already been dug beneath the Mount Baker neighbourhood in 1940.
Many other examples regarding cost overruns can be provided but they all lead to same question: why are the same mistakes are being continually repeated? There are three main possible reasons (Flyvbjerg, et. al. 2003): Inadequate data on which the project is based, optimistic bias, and cost analyses are systematically and significantly deceptive. The discussion up to this point has only been about comparisons between the original budgets prepared by the clients and the final cost after the project completion. When the project is tendered and the bids received from the contractors, prices may be over or under that which is allowed. However, this is only a temporary situation and the most important is the price at the final completion of the project. The difference between the initial and final price and the reasons for this difference will further be analysed in this paper.
Further discussion on the low budgets allowed by the clients is not within the scope of this project. However, the following factors should also be taken into account when low budgets prepared by the clients are discussed:
a. Clients may intentionally approve tight and low budgets and schedules in order to make the consultant and the contractor believe that funds are more limited
b. Clients, particularly those influenced by politicians, may try to minimise the budget and schedule in order to secure the approval of the project.
c. Historical data may be highly biased since the prices for each contract is based on the current economical situation when the bid price is placed, economical situation of the wining contractor, competition, business decisions and fluctuations in the material prices.
Romero, et. al. (1997) pointed out that the differences between the ‘cost’ as calculated by the consultants and the clients and the ‘price’ as calculated by the contractors are the constructability issues, risk and competition since cost estimation based on production type calculations should not differ. However, it is the author’s opinion that there should be other factors that mainly drive the differences between the cost and the price. The main factor is uncertainty, which causes either excessive or insufficient figures when preparing budgets, or changes project price based on different perceptions.
Uncertainty and different perceptions of Uncertainty
Many famous and ‘scary’ terms are involved in underground work contracts such as unforeseen, unforeseeable, foreseeable, uncontrollable, unpredictable, disclaimer and risk. They are all the end products of uncertainty which is generally unavoidable in geology and geotechnics. Uncertainty is categorised as either epistemic or aleatory. Der Kiureghian, et.al (2007) provides definitions for those terms:
The word aleatory derives from the Latin alea, which means the rolling of dice. Thus, an aleatoric uncertainty is one that is presumed to be the intrinsic randomness of a phenomenon. Interestingly, the word is also used in the context of music, film and other arts, where a randomness or improvisation in the performance is implied. The word epistemic derives from the Greek episteme, which means knowledge. Thus, an epistemic uncertainty is one that is presumed as being caused by lack of knowledge (data).
Epistemic uncertainty and aleatory uncertainty can be referred as subjective uncertainty and randomness respectively. According to Der Kiureghian, et. al (2007) uncertainties are characterised as epistemic, if the modeller sees the possibility of reducing them bygathering more data or by refining models. However, if this possibility does not exist, uncertainties are categorised as aleatory.
Examples of an aleatory uncertainty are: the uncertainty in timing of activation or movement of a fault or landslide (regardless of their locations are whether known or not); the exact profile of the tunnel after blasting (regardless the geology and blasting technique); the exact number and size of boulders along the tunnel alignment; the fluctuation in the cost of construction materials, and changes in legislation. Examples of epistemic uncertainty include the uncertainty in the dip and strike of a fault crossing the tunnel (which may become known if further investigation is carried out); the amount of water discharged into the tunnel when a water bearing strata is penetrated (the amount can be identified if permeability is known accurately); excessive rates of corrosion due to groundwater chemistry (due to improper or missing testing). It is clear that more knowledge, information and data reduces the epistemic uncertainty whereas the same does not apply to aleatory uncertainty. Other factors in reducing epistemic uncertainty are the probability of occurrence and the scale effect.
Although there are many controversial opinions regarding the distinction between uncertainty, and that risk and uncertainty have been used as if equal terms for many years, many researchers have separated the terms. Brooke, (2010) commented on one of the well known but disputed works by Frank H. Knight (1921) on risk and uncertainty.
According to Brooke, (2010); Knight’s commonly accepted definition is that risk refers to outcomes that can be insured against, and uncertainty to outcomes that cannot be insured against. Brooke, (2010) also adds that uncertainty refers to all instances where only subjective estimates of future outcomes are possible.
According to Luce, et. al. (1957), reproduced after Riabacke (2006), ‘risk’ occurs where each action leads to one of a set of possible specific outcomes, each outcome occurring with a known probability whereas ‘uncertainty’ occurs when actions may lead to a set of consequences, but where the probabilities of these outcomes are completely unknown. A risky situation is thus a situation where the outcome is unknown to the decision-maker. Although they are used interchangeably in practice, the terms risk and uncertainty will be used as separate terms, and as defined by Luce, et. al. throughout this paper. The main reason for this is that the risks can be defined since they exist as a probability, while uncertainty cannot be measured in terms of probability.
An example of this is that probabilistic analysis of a TBM project under certain conditions can provide probabilistic information and knowledge about a new project for the contractor. Hence, certain aspects of the new project are risky but not uncertain. Conversely, perched aquifers over a tunnel which is constructed by conventional methods are an uncertainty since the behaviour and impact of perched aquifers based on their potential occurrence, size, location and extent are unknown by the contractor. However, uncertainty can be interpreted with a certain ‘degree of belief’ or subjective probability based on experience. Degree of belief, however, does not imply being arbitrary but should be based on some knowledge. Figure 2 provides examples to visualise the difference among the terms since the perception of uncertainty is the core subject of this paper.
Assume that two tribes named A and B living on either side of the mountain decide independently (since they do not have any form of communication and they are unaware of each other) to tunnel the mountain to reach the other side. They start at the same time and drive tunnels in opposite directions as shown in 2a. Although they do not know that the mountain has a syncline, they are able to record and track the geology as they keep driving. Hence, based on already completed sections of the tunnel, Tribe B recognises that cyclically encountered strike and dip of the strata will continue in the rest of the tunnel and they can identify the risks based on that knowledge, although it is a total uncertainty.
The risk area which is actually an uncertainty is shown as hatched area in Figure 2a. Now assume that Tribe A temporarily stops tunnelling while Tribe B continues and hits the centre of the structure where they probably faced water, then continues and observes that a strike and dip of the same strata changes in a way that indicates they may be connected as Figure 2b. At this point, Tribe B has a degree of belief in the strike and dip of the strata as well as excavation conditions and behaviour of the rock mass for the remaining tunnel length. When Tribe B continues excavation further (Figure 2) they observe and confirm that their belief has been correct and the rest of the tunnel length becomes a risk which can be identified and quantified. Interestingly, for Tribe B this has become an area of risk based on data, information and knowledge, but for Tribe A, it remains an uncertainty. Therefore, it is clear that someone’s uncertainty can be another’s risk. This is the essential point in the perception and pricing the projects. Higher profits are earned where there is uncertainty and conventional profit margins are earned where there is risk. This theory has been established by Frank H. Knight (1921), reproduced by Brooke (2010), who stated that "if the future is risky, no profit can be earned if all the alternative possibilities are known and the probability of occurrence of each can be accurately ascertained". Hence, the contractors first try to identify the uncertainties and the risks when they obtain the contract documents including geological and geotechnical reports.
Aleatory uncertainties remain aleatory for every party. However, epistemic uncertainties are further analysed by the contractors based on their knowledge, past projects, data available to them and local experience. As a result, some epistemic uncertainties become risks according to the degree of belief for the contractors; and they use that advantage in pricing the project.
Interdependence of geologIcal and geotechnIcal parameters
The initial stage for checking the tender documents for a tunnel is obviously the contract delivery method and contract conditions because these two items affect the whole process. The next stage is to consider all geotechnical reports, which may include geotechnical factual reports, geotechnical design reports, geotechnical interpretive reports and geotechnical baseline reports. Either some or all of these may be provided as part of the contract documents. However, it is not unusual to encounter tenders with only a simple geological summary and a bill of quantities. The contractors may choose to get the reports checked, interpreted and evaluated by external consultants in addition to their own in-house checks. If the construction method has not been defined by the client in the tender documents, the contractor’s first action is to identify a resemblance between the new project and previous projects. This process is considerably less difficult if the contractor has worked in the same geography and geology in previous projects. This initial conceptual stage also guides the process of analysing the reports with respect to the importance and complexity of parameters which vary for different methods of construction.
Although there are many parameters involved in geological and geotechnical reports for any underground project, some of them have more weight than others in terms of method selection and costing. All models and parameters will be used to calculate the productivities as well as any short and long term effects. However, this is not an easy task since the parameters are interrelated and complex. Furthermore, different methods will require different parameters to be checked. As an example, particle size distribution is important for a tunnel excavated by slurry machine whereas it has much less importance if the same tunnel is excavated by conventional methods. Another example is the effect of strain bursting which can be controlled by using the distressing effect of blasting in a rock tunnel whereas this approach may not be that easy when a TBM is used.
Table 2 shows the development of productivity based parameters together with the parameters that have long term effects. This is only one example showing some possible parameters but a range of other parameters and models may differ for each case. It only shows where parameters are originating from and which parameters directly affect the productivity, with long term risks. The meaning of the table’s letters are given below:
A. Related with geology
B. Related with geotechnical parameters and rock mass classification
C. Related with excavation and support system
D. Directly affects productivity
E. Directly affects long terms risks
It is clear from the table that while some parameters have direct impact on productivity, others only have indirect effects. Therefore, the cost estimation will mainly be based on the productivities (directly affected by those parameters) and a premium will be allowed for long term affects if those effects are not offset either by insurance or by the client.
The example in Table 1 shows only the main part of the process. Method and procedures adopted during investigation stage, scale effect, inherent pitfalls and limitations of classification systems were also analysed as part of this process.
Case Studies
Several case examples are summarised in this article in order to show how geological and geotechnical reports are processed, and uncertainties and risks are identified by the contractors during the tender and project execution stages. Names and details including costs of the projects have been intentionally modified. Two case studies in this paper explain the concepts of uncertainty, degree of belief and risk. There are many other projects in the literature that may be analysed with the same concept. Riemer (2006) reports that Q-rating to the phyllites of the Kali Gandaki headrace tunnel in Nepal had given a rating of 2<Q<50 for 84 per cent of the tunnel and led the engineer to consider an unlined tunnel which was rejected by the owner on the basis of experience with other tunnels in similar rock. Eventually, rock mass performance observed during construction of the tunnel mainly corresponded to a rating in the range of 0.4<Q<1.0.
This example also shows the importance of the understanding of epistemic uncertainty and degree of belief.
The geological environment in which the tunnel was constructed was modelled by the designer, with certain classification but the degree of belief of the client based on its past experience led to decision to line the tunnel which proved to be the correct decision during tunnel construction.
Conclusion
Although this paper cannot address all possible causes of time and cost overruns, it may be helpful to consider another perspective of costing tunnel projects from the point of view of the contractors, the mostly silent party.
Although there are very standard conclusions that are known to the industry such as realistic geotechnical reports, good site investigation, risk sharing, no disclaimer clauses and partnership, the following factors should also be considered:
a. Clients should disclose past project experiences to the bidders and should not hide the risks.
b. Parameters and information in the reports should be classified according to their effects on productivity considering potential construction methods. This will assist in providing more knowledge and data.
c. Clients should give the bidders an opportunity to comment on the gaps, omissions and insufficiencies in geotechnical reports.
d. Commentary on possible rock zones should be provided.
e. If excessively high baselines compared to available data are selected, the bidders will think that either the tests are faulty or the client has certain information that it does not want to disclose. This will cause the bidders to add premiums on their prices.
f. If rock mass classification systems such as RMR and Q are used, it is important to provide all data used to calculate the index values rather than reporting the index value only.
g. Commentary on rock mass behaviour is very important.
h. Baselines should be measurable and verifiable otherwise they will not have significant bearing on pricing
i. If unconventional techniques are proposed as a part of the project a detailed description including risks and potential construction techniques should be provided.
j. A bill of quantities or pricing should be selected in a way to reflect the complexity of the geology. Single pricing under one rock class with a wide range rock mass quality will lead to a higher price than narrow ranged rock mass classification.
k. Scale effect should always be considered even evaluating the response of the tunnel to excavation.
l. Clients should take into account other factors, such as market conditions, business drivers and competition when tendering the projects.
All efforts should be directed to change the attitude of ‘your uncertainty is my risk’ to ‘we share risks’. It is important to know that a greater amount of information will help to reduce epistemic uncertainty
Case one: project RT-01
This is a motorway project with a cost of more than USD 220M which includes two NATM tunnels with twin bores and each carrying three traffic lanes. Excavation cross sections range between 120m2 and 150m2 depending on the use of invert. The client undertook the preliminary design and geotechnical report. However, the contractor who was very experienced with similar tunnels was responsible for completing the design of the portals as well as shop drawings for the tunnels. The client provided rock mass classification, based on ÖNORM B 2203 (Austrian Norm). However, Q and RMR systems were also used by the designer to validate the distribution. The project used a measure and value contract and the rock mass classifications needed to be signed off by representatives of the engineer after an inspection with the contractor.
Table 2 shows the rock mass distribution provided in the contract, as estimated by the contractor and the as-built situation for a 500m long tunnel. It is worth noting that a major part of the client’s estimate was based on a class whereas the contractor’s was mainly based on B class, which proved to be the case during construction. The reason for the designer’s inaccurate estimate was mainly due their site investigation and interpretation of site results. Since the tunnel was under a hill and rock mass was exposed in outcrops, no borehole were made. This was also partly due to access issues in the densely forested land. The information gathered from out crops alone was enough to make an assessment. Q and RMR values have never been reassessed based on the effect of excavation. The behaviour of the rock mass was not considered and index based rock mass classifications (i.e. Q and RMR) were used only to assess ÖNORM B 2203 classes which are mainly behaviour based. The contractor visited the site and made visual investigations, collated all data together with the methodology and made its own estimation. Although the contractor estimated that the dominant class would be B1 (fractured rock with low deformations and loosening), actual rock conditions were more on class B2. The lineament observed was believed to be a fault zone by both the designer and the contractor and classified as C1 and C2. However, it was classified as B3 during the excavation. This example shows that the uncertainty was not in the parameters or the information used but rather it was the evaluation of those parameters and the information and the perception of the rock mass behaviour based on the past experience in similar tunnels.
Case two: project WS-01
Four pipeline tunnels with a cross section of about 40m2. Total length is more than 4,000m. Tunnels were mainly located in metabasalt, metagraywacke, schist and marble. However, the last tunnel was mainly within granite and granodiorite. The client did no site investigation but provided geotechnical profiles with the distribution of excavation classes. Excavation and support was designed by the client with three different classes. The contract was measure and value and the standard contract clause shifted all risk for the changes in the distribution of support classes to the contractor. Excavation and support were paid separately per linear metre of the tunnel in any particular support class. The design and construction was mainly following NATM. While Support Class One (SC1) has only rock bolts, wire mesh and shotcrete (out of cycle), SC2 has the same support only the number of bolts and shotcrete thickness increases while round length decreases about 35 per cent. However, SC3 is heavily supported using rock bolts, shotcrete, wire mesh and steel sets. Tunnel excavated in SC3 should also have an invert and forepoling as temporary support. SC3 represents 10 per cent of all tunnel length and mainly assigned to zones at either end of the tunnels and SC1 and SC2 represents 57 per cent and 33 per cent of the total tunnel length respectively. Approximate RMR values for the support classes have been shown in Table 3 in order to give an idea about the rock mass quality. It may be expected that different rock support classes will have different prices. Table 3 shows that the contractor allowed a significant premium to his prices in order to level prices for different support classes. One factor was the geology. The contractor visited the site and observed extensive jointing, folding and faulting which would affect the excavation and the progress. Hence, the contractor had the opinion that SC1 and SC2 do not differ practically. Due to the rock conditions, the contractor also estimated that the overbreak would be significant and this will not only increase his direct cost but would also have direct impact on shotcrete quantity and time since the specification called for a final shotcrete layer. The necessity of concreting the invert, leaving a rough surface would also lead to cost increases. Access conditions were also a factor.
The most important lesson is the necessity of having a complete geological and geotechnical report, which should identify areas with potential problems and include an evaluation of the rock mass behaviour. The pricing structure should allow contractors to address concerns without adding large premiums.
