Risk Analysis of Offshore Oil Development Projects : A Case Study

Development of offshore oil production projects entails a huge capital outlay. The various aspects of the project are filled with uncertainties. If these are not considered correctly, an initially attractive venture can become uneconomic. It is thus necessary to carry out analyses that can identify and quantify the inherent risks of the project so that they can be mitigated as much as possible. This work presents a method of quantitative risk analysis that can serve as a basis for decisions regarding projects of this nature. We use a case study to exemplify the proposed risk analysis methodology.


INTRODUCTION
Deciding on an investment when there is no uncertainty is relatively simple.One need only calculate the net present value (NPV) and internal rate of return (IRR) of the various alternatives and choose the one with the best return.However, in practice decisions nearly always are filled with uncertainties.Companies' success depends on economic feasibility studies before deciding to embark on a project.These studies must consider the risks and uncertainties present in the project.
The high degree of uncertainty in offshore petroleum projects heightens the importance of using various methods to help identify where the greatest uncertainties lie and what variables most influence the expected economic result.The impact of these uncertainties must be quantified and the risks fully assessed, under penalty of generating undesirable economic results for the firm.
In the case of oil exploration and development projects, the investment decisions are affected by both technical and economic uncertainties.The technical uncertainties are intrinsic to the project and are not correlated with general market movements, but the economic uncertainties are.
One of the most commonly used forms of risk analysis is numerical simulation by the Monte Carlo method, which according to Prado (2000) is a way to transform a set of random numbers into another set of numbers (random variables) with the same probability distribution as the original set.
There are some computer programs that can be used to carry out risk analysis, such as @Risk and Crystal Ball, among others.In the case study presented here, we use the @Risk software.
We carried out 10,000 iterations and present here some of the main indicators to better identify the project risk, along with graphs to better illustrate these indicators.

RISK ANALYSIS
Before discussing the uncertainties and risks associated with these projects, it is necessary first to draw a theoretical distinction between the two words.According to Simonsem (1994), "risk is when the random variable considered has a known probability distribution, while uncertainty is when this distribution is unknown."Therefore, it is desirable, to the extent possible, to convert uncertainty into risk, calculated by determining the probability distributions for the uncertain variables.
There are various methods used to assess the risks to which offshore oil projects are subject: sensitivity analysis, analysis of scenarios, Monte Carlo simulation and decision trees.
Sensitivity analysis consists of changing the value of each of the variables individually to enable analysis of the effect of this alteration on the project's cash flow, identifying the variables that most influence its economic result.
Scenarios analysis permits correcting one of the errors of sensitivity analysis, by considering the interdependencies between the project's variables, for example: the optimistic, pessimistic and most probable (or expected) scenarios, considering different but consistent combinations of the variables.
Monte Carlo simulation involves the impact the variables have on the project's result.In summary, this method associates probabilities of the occurrence of these variables, considering all the possible combinations of the variables, making this method more robust that the other two.
Decision trees can be used to analyze projects that involve sequential decisions.Projects are not treated as black boxes, in which strict decisions such as acceptance or rejection are considered, ignoring the effect of earlier decisions on subsequent investment decisions.

-Monte Carlo simulation (MCS) to analyze projects
The Monte Carlo method utilizes random numbers to compute quantities that are not necessarily random.It seeks to generate random values in a model with the objective of producing hundreds or thousands of scenarios.Due to its simplicity of application, this method has proved to be an excellent tool to deal with financial problems, such as how to calculate options prices, measure market and credit risk, calculate value at risk (VaR), analyze investment projects and solve real options.
For project analysis, the Monte Carlo method, according to Brealey and Myers (1998), consists of three phases: Phase 1: Construction of the project mode.consists of supplying the computer with a precise model of the project; Phase 2: Specification of the probabilities of forecasting errors; Phase 3: Selection of the numbers for the forecasting errors, calculation of the resulting cash flows of reach period and their registration, as long as the model is precise.
These authors suggest that MCS not only be used for the NPV distributions, but also to raise the level of comprehension of the project, investigate its future cash flows and evaluate the risk.It can serve as a basis for calculating the NPV, discounting the expected cash flows by an appropriate discount rate.
Despite its ease of application, MCS has some disadvantages.Among them are, in complex problems, the high number of interactions to obtain a result.Various techniques to reduce the variance have been developed to increase the result's precision without unduly raising the computation time.Another way to hasten the simulation's conversion process is to modify the way the random numbers are generated.The Latin hypercube sampling method, which uses the Monte Carlo principle, generates random numbers in a more efficient way.In BBR, Braz.Bus.Rev. (Engl.ed., Online), Vitória, v. 3, n. 2, Art. 6, p. 223-237, jul.-dec. 2006 www.bbronline.com.br this article, we follow this methodology together with the @Risk program to resolve the problem of the case study proposed later.
Another point that should be considered in the Monte Carlo method is the extreme difficulty of estimating the relationships among the variables and the probability distributions on which the simulation is based.For this reason, a project's risk analysis team should be very diligent in generating the inputs.

-Main uncertainties
The main uncertainties of offshore oil projects regard the production curve, investments, operating costs, project delays and prices of the products (oil and gas).

-Production curve
According to Ligero, Costa and Schiozer (2003), one of the main uncertainties in developing oilfields relates to the static variables (geological model) and dynamic variables (flow parameters) of the reservoir.These reflect directly on the production curve, and consequently on the project's net present value (NPV).Steagall and Schiozer (2001), in turn, propose a method to analyze and quantify the uncertainty and risk in production forecasts.Their method is based on the work of Loschiavo (2001), in which numerical simulation of the flow under different scenarios of a reservoir is used, combining the uncertainty attributes.
The production curve should represent the set of main existing uncertainties.For this, it is advisable to obtain as many production curves as possible, considering the parameters that cause the most impact on the reservoir.For each scenariocombination of parameters under analysisa new production curve is generated, which consequently alters the projects expected economic result.All the curves generated should consider the same design concept (capacity of the stationary production unit -SPU) and the number and location of the wells, because the risk of the concept envisioned should be analyzed in relation to the uncertainty about the reservoir, according to the level of information available about it.The big obstacle in doing a risk analysis of the reservoir is that it generates a huge number of production curves, which are normally obtained using a numerical flow simulator.It should be stressed that the time to generate a curve is generally lengthy, making the generation of many curves expensive.For this reason, this topic is frequently discussed in industry forums and is an area of intensive research and advances to improve the desired analyses.

Investments
Regarding investments, there are both technical and market uncertainties.
The quantification of the technical uncertainties includes finding the probability distribution that best represents a determined investment item, through historical data.Some programs, such as BestFit, can be used to find the best distribution from the data fed into the software.If such data are not available, a distribution for the investment costs can be used based on the experience of the technical staff in charge of budgeting.The market uncertainties referring to the investments should be assessed by modeling market costs using stochastic processes, such as Brownian geometric motion (BGM), mean reversion (MR), etc.This is because these costs vary stochastically over time.For example, the rate for a pipe-laying boat can be higher or lower over time and its variance can increase in time.Figure 1 presents samples of random walks for BGM and MR.Further regarding technical uncertainties, operating costs have two parts: one that varies with output and another independent of output.Thus, equation 1 can be used for operating costs: where, OCFix = Fixed operating cost; OCVar = Variable operating cost; Q = Oil flow of the project.
To consider the uncertainty in the fixed and variable costs, a probability distribution must be adopted for them based on historical data, when applicable, or if none are available, then on data obtained from the experience of technical staff in the area.
The consideration of the market uncertainties in the operating costs follows the same method described for investments, namely, considering stochastic processes.

Oil and gas prices
The uncertainties regarding oil and gas prices are market uncertainties and should be modeled using stochastic processes.Choosing the best stochastic process involves analyzing historical data and the project's proposed time horizon.The most popular stochastic process is BGM.However, Dias and Rocha (2001) use the mean reversion stochastic process combined with jumps to model the price of oil.
According to Dias (1996), econometric tests of the price of oil, over an interval of 30 years (or less) do not reject the hypothesis of BGM.However, when a longer interval is considered, of 117 years, the BGM model was rejected in favor of MR.So, it can be important to consider this latter model for long-duration projects.

Timetable for implementation
Another very important factor to consider in analyzing risk and uncertainty is the project's timetable.A way to consider this uncertainty is to carry out a risk analysis for the timetable, using probability distributions for starting and ending dates of the activities that cause uncertainty.To do this, specific programs can be used, such as @Risk for MS Project from Palisade Corporation.This analysis generates a probability distribution for the starting date of production.The probability of delayed or early start-up can considerably affect the project's NPV.

Analysis of the project
To carry out a risk analysis, we recommend first considering the technical uncertainties, because these are manageable.According to Dixit and Pyndick (1994), technical uncertainty encourages the investment step-by-step as the variance of this uncertainty is reduced.For example, the manager can invest in information to improve the understanding of the project, and consequently, mitigate its risks.The probability distribution of the NPV resulting from this analysis will show the project's degree of risk, and the manager can know the variables that most affect its economic outcome.
After conducting the analysis only considering the technical uncertainties, the two uncertaintiestechnical and marketare mixed to verify the impact they can cause on the project's NPV.The NPV probability distribution is thus generated.Since market uncertainties are not manageable, the project manager cannot take mitigating actions in this respect.According to Dias (1996), economic uncertainty is exogenous to a firm's decision-making process.

A CASE STUDY
To illustrate the risk analysis methodology presented, we carried out a case study of a deepwater offshore oilfield.The data are adapted from those presented by Suslick (2001).

-The data considered
We considered a steeply rising production curve in the first years, until its peak, followed by an exponentially declining curve.This is the most commonly used curve and can be represented by the following equation: A typical production curve is shown in Figure 2.

Figure 2: Typical Oilfield Production Curve
We considered the production peak as a variable following a triangular distribution with minimum, most likely and maximum parameters equal to 137, 200 and 247 thousand bbl/d, respectively.The SPU is limited to 200 thousand barrels of liquid per day, so it is not possible to surpass this level in case of simulating a curve with greater potential output.
To consider the uncertainty in the model presented in equation [2], we assumed a normally distributed exponential rate of decline, , with mean 18% and standard deviation 0.9%.We considered the project length to be the concession period granted in Brazil by the National Petroleum Agency (ANP), 27 years.Therefore, the variation not only occurs in the production curve, but also in the field's total reserve (volume recovered), which for the most probable case would be approximately 500 MMbbl.
For investments to develop production, we assumed a minimum and maximum price of US$ 3.11 and 5.07/bbl for a field of around 500 MMbbl, as adopted by Suslick (2001).For giant fields, as is the case in this study, uncertainty is very significant, since normally they are complex projects and demand a high outlay.Therefore, to consider the technical uncertainty in this item we chose a triangular distribution with the minimum and maximum parameters cited just above, and for the most probable value we assumed the average of these two figures, namely US$ 4.09/bbl.The minimum, most probable and maximum parameters for this triangular distribution, converted into millions of dollars, works out to US$ 1555, 2045 and 2535 million, respectively, as shown in Figure 3.
The triangular distribution is widely used when there are no historical data available, and also because it is easy to understand.We did not consider market uncertainty referring to the investments.We used equation [1] to consider the technical uncertainty in the operating costs.Again we used a triangular distribution for the fixed operating cost, with minimum, most probable and maximum parameters of US$50, 45 and 40 million, respectively.The most probable figure corresponds to approximately 57% of the total expected operating cost, as used in the case study of Suslick (2001).For variable operating cost, we adopted as fixed US$1.55/bbl, meaning the variation occurs due to the production curve.Again, we did not consider market uncertainty in variable operating costs.
The cost of abandonment must also be considered.Here we assumed a triangular distribution with minimum, most likely and maximum parameters or US$ 80, 100 and 140 million, respectively.
We considered that the oil price follows Brownian geometric motion (BGM), with a mean of US$ 20/bbl, volatility of 30% and growth rate of 1%.The BGM equation is as follows: where, Pt = Oil price at time t; Pt-1 = Oil price at time t-1;  = Growth rate;  = Volatility; t = Difference between time t and t-1.

-Results considering technical uncertainties
First we carried out an analysis only considering the technical uncertainties, to better understand the effect of the manageable risks of the project under study and be able to take actions to mitigate them.Table 2 presents the main statistics related to the analysis considering only technical uncertainties.
Tabela 2: NPV Statistics for the Analysis Considering Technical Uncertainties Figures 4, 5 and 6 represent the probability distributions of the NPVG found in the simulations.Note that the probability of the NPV being less than zero grows sharply with increasing minimum project attractiveness, especially when going from 15% to 20%.This shows the need to carefully define the rate adjusted to the risk to be considered in this type of study.Another interesting aspect is to know which variables have the greatest impact on the project's NPV. Figure 7 shows a sensitivity analysis of the NPV, considering simulation 3.For the other simulations, the ranking was the same.Figure 7 shows that the production curve and investments are the items that have the most impact on the NPV.Hence, it is advisable to evaluate the possibility of investing in additional information to mitigate these risks.

-Results considering technical and market uncertainties
We also conducted an analysis considering both technical and market uncertainties to assess how much these combined uncertainties can affect the result.The results presented do not permit affirming which measures can be taken to reduce the risks, because the technical and market uncertainties have very different characteristics.The probabilities of a negative NPV were 37.0%, 46.5% and 56.6% for simulations 1, 2 and 3, respectively, which shows that market uncertainties can mean large gains, but also large losses.We also conducted a sensitivity analysis for the project's NPV.This analysis is depicted in Figure 11 and shows that the price of oil causes the greatest impact on NPV.This is due to the high oil price volatility in the world market.

CONCLUSION
Risk analysis of investment projects is a way to gain a better understanding of the risks involved, by identifying and quantifying the risk of the variables that most influence a project's economic outcome.This knowledge is extremely useful to assist decision-makers to choose projects to carry out.This is particularly important in large projects with high levels of uncertainty, such as development of oilfields.Risk analysis permits knowing the risks involved in projects to enable more effective mitigation measures.Thus, technical and market uncertainties should not be mixed, because they demand distinct mitigation actions.An analysis should first be conducted considering these uncertainties separately, and then they can be combined to get an idea of their possible influence on the project's NPV.
Conducting risk analysis requires a high level of diligence by the whole team involved in the project, because the precariousness of the input data can cause doubtful results and thus reduced reliability of the analysis.
The results of the risk analysis carried out in the case study here show that the production curve and investments were the items that had greatest impact on the NPV, considering only technical uncertainties.The indicator of the probability of negative NPV showed that, depending on the MAT adopted for the project, it may or may not have a probability of being loss-producing.An increase in the MAT has a very large impact on the

Figure 1 :
Figure 1: Brownian Geometric Motion and Mean Reversion Figure 3: Triangular Distribution for the Total Investment

Figure
Figure 4: NPV Probability Distributions for Simulation 1 Considering only Technical Uncertainties

Figure 5 :
Figure 5: NPV Probability Distributions for Simulation 2 Considering only Technical Uncertainties

Figure 7 :
Figure 7: Correlation Coefficient (Tornado) Graph for the NPV Considering only Technical Uncertainties

Figures 8 ,
Figures 8, 9 and 10 show the NPV distributions for the project under study.

Figure
Figure 8: NPV Probability Distributions for Simulation 1 Considering Technical and Market Uncertainties

Figure 11 :
Figure 11: Correlation Coefficient (Tornado) Graph for NPV Considering Technical and Market Uncertainties

Table 3
presents the results of the simulation carried out combining the technical and market uncertainties.Table3-NPV Statistic for the Analysis Considering Technical and Market Uncertainties 4 P10 = Optimistic, with only 10% probability of being surpassed. 5P50 = Most likely scenario, with 50% probability of being surpassed. 6P90 = Pessimistic, with 90% probability of being surpassed.