Presented in Sao Paulo, Brazil, 1995, supported by the UK's Royal Institution





The custody transfer of hydrocarbon fluids are discussed focusing on the custody transfer of liquids and refrigerated liquefied petroleum gas (LPG) in particular. There is presentation of several years ship loading and metering figures for LPG and analysis to reveal interesting intercomparisons between the metering system’s normal and alternative, informal traceability.


C3 = propane
C4 = butane
C3 + C4 = total LPG cargo taken on ship
m = volume measurement by flow metering system, in barrels (bbls)
s = volume by ship tankage measurements, in barrels (bbls)


Although product quality, and its method of determination is a vital aspect of any custody transfer agreement, the determination of the amount of material changing hands has always been of prime importance. Accordingly, since their introduction in the late nineteenth century modern industrial flow meters have been responsible for measuring an ever growing proportion of the world's annual wealth transfer. Today, they perhaps measure more of this than any other type of instrument. This has lead to a greater interest in their accuracy and the accuracy of flow meter calibration.

Because of the nature of equipment to measure fluid quantities they have, in the past, usually been traded as volumes. Since product volume is temperature dependent greater exactitude in measurement required the specification of measurement temperature. More recently development of reliable and accurate industrial densitometers has lead to trade in mass which allows easier, more fundamental inter-comparison at different locations without the need to refer to a particular temperature.

Large quantities of gas have been, and still are, measured by square edged orifice plate, but within the last two decades gas turbine meters have been increasingly used in those locations where they can be easily, reliably and accurately verified. This movement started in western Europe but has since been taken up in North America.

As a relatively high value product large volumes of oil were, for a long time, measured almost exclusively by tankage measurements. Venturi, and orifice plate flowmeters were not seen as being sufficiently accurate, and although various forms of positive displacement meter had been invented fairly early in the development of the oil industry they hadn’t the capacity to be used in a practical way to handle the large quantities of liquid being moved. It took until the early 1950’s before a sufficiently rugged and accurate version of the, by then, well known turbine meter, began to result in their use for this application.

Although -- for products with a viscosity of up to 20 centistokes -- most modern turbine meters exhibit better than 0.25% uncertainty of measurement over a 10:1 range their inferential nature does not promote immediate confidence. They are therefore regularly ‘proved’ on site by an easily believable, reference proving device to empirically establish the volume of liquid passing through the meter during the time interval from any one meter output electrical pulse to the next. Most commonly the reference is a 'pipe prover', a length of roughly flow meter diameter pipe through which flow pressure pushes a tight fitting elastomer ball between two axially well separated switches. By an appropriate system of valving the flow, having just passed through the meter, pushes the ball to trip first one switch, then the other. An electronic counter totalizes the number of turbine meter pulses produced between the two switch operations so that, with the volume of pipe separating the two switches accurately measured in a separate 'off-line' operation, the volume 'per meter pulse' can be calculated. Ideally the volume per pulse should be invariant, but in reality it varies, although by less than the meter’s linearity within the meter’s range.

In more recent years a meter’s inherent linearity has become of less importance because of the development of special purpose flow computers with fully automated prover control. These require only a single signal to enter an automatic routine which proves the selected meter over its entire range. Calibration is measured at, typically, up to ten different flow rates and the new calibration factors stored internally for automated use correcting future flow rates. As such each meter’s calibration is known and used over its full range to almost the very small uncertainty of the prover itself, commonly better than 0.02%.

Even though turbine meters are commonly made in sizes up to 600 mm diameter the large volumes that many metering stations are required to meter, and the need for operational flexibility and redundancy, typically result in piping arrangements in which several meters can be operated in parallel. This aspect, together with the requirement to prove each meter individually, results in complicated valving configurations. Primary considerations then become; the ability of closed valves to seal effectively and, to a lesser extent, to open and close relatively quickly since while traveling near the closed position flow rate is below a meter’s linear range.

Product quality is also of importance if it is so ill defined that time varying changes in the liquid’s physical properties reflect as significant uncompensated meter factor variations. In this respect even changes in physical conditions can be a matter for concern. It is also necessary to ensure that the liquid is free from contaminants which can cause meter impairment and hence inaccurate metering. Large solids cause jamming or impact damage. Sticky particulate solids can accumulate to inhibit rotor movement, adversely upset the flowmeter’s dynamic forces and promote instability. Small quantities of vapour cause inaccuracy by displacing liquid volume within the meter barrel. Fortunately some of these potential problems can often be removed by the inclusion of strainers and vapor eliminators upstream of the meter. Waxes, however, one of the common contaminates in some oils, are not so easily dealt with.


The liquefied petroleum gases (LPG’s) are, as the name implies, liquefied industrial alkanes which are gaseous at atmospheric temperatures and pressures. The term is usually limited to cover industrial grade propane and butane, which only require a moderate change in physical conditions to become liquefied.

LPG’s are typically rendered such by either the application of only pressure or a reduction of only temperature. An idea of the deviations from atmospheric conditions to produce these commodities is given by the fact that if only temperature is changed butane liquefies at about -4 deg.C and propane at -43 deg.C. Although bulk storage and movement of LPG’s is done in both pressured and reduced temperature regimes, the latter is by far the most common. For this reason, and the fact later results in this paper relate to it, further discussion focuses towards the refrigerated form.

One of the problems which LPG’s exhibit more strongly than oils is their greater tendency to release and carry vapor. Ideally, if the process and facilities are properly designed and operated, none should be entrained in a metered flow but this is not always the case. The difficulty arises with either pressurized or refrigerated forms for which excessive, localized pressure drop, such as caused by unsuitable pipe or vessel design, partially closed valves, or partially fouled strainers, can easily cause gasification which downstream pressure recovery rarely reliquifies. Hot weather can aggravate the tendency, particularly if predomenantly normal butane is left with significant quantities of higher boiling point iso-butane, either by design or blunt “cut-off” liquifaction. A vapor eliminator, just upstream of each meter, does not always have sufficient capacity and efficiency to fully cope.

LPG’s low viscosity promotes valve leakage to endanger the metering accuracy of the whole system in a way that is often very difficult to detect. Soft seals, are inappropriate, especially for propane, which rapidly degrades otherwise suitable, known materials.

For light oils turbine meters usually have a sufficiently smoothly varying meter (factor) calibration curve for flow computer interpolation routines to accurately evaluate flow rates intermediate between those at which ‘proof-run’ calibrations were made. This is not always so for the more ‘erratic’ calibrations curves often yielded by LPG’s.

The low viscosity, non oily nature of LPG also promotes rotor bearing ware directly, by almost eliminating a lubricating film, and indirectly, by encouraging the release of fine pipeline rust to act as bearing abrasive which promotes instability.

LPG’s coefficient of thermal expansion is considerably greater than that of oils which increases the need for stable, accurate temperature measurement and compensation. Stability is important because thermal measurements are slow so that, at any moment, the temperature used for automatic compensation may be different to that of the small increment of volume being registered.

Large vapor pockets cause damaging, unbalanced fluid dynamic forces. The problem is usually greatest when a meter is first brought into service for a particular transfer. The reason for this is associated with the system to maintain low temperature. To prevent LPG’s substantial, problematic vaporization within pipelines and maintain the low temperature of pipelines to meters, it is common practice to include as much as possible of the liquid filled pipework of a ship loading system within a ‘loop’ to allows the LPG’s continuous recalculation back to the liquefaction plant. If loading is not in operation this recirculation is allowed, by an appropriate valve configuration, to extend via the meters as far as the base of the loading arms. When loading is in progress the continuous supply of fresh, cool liquid removes the need for recirculation for other than keeping the liquid return line cool. For this more limited application a small fraction of the loading flow is bled off just upstream of the meters. It is during the change-over from non-loading to loading operational mode that temperatures in different sections of pipework change most and need to be allowed to stabilize before, proving, metering and loading commence.

In spite of best efforts to maintain proper process conditions during loading it is inevitable that the combined effects of heat influx and pressure drop cause substantial quantities of vapor to be released into a ship’s cargo tanks. If not properly accommodated this leads to substantial error in the bill of lading. Proper methods of accommodation include: accepting excess vapor back on shore, provided it is properly metered and back calculated to liquid equivalent for debit from that loaded (1). Loading at a sufficiently slow rate that the ship’s own refrigeration system can condense all boil-off vapors.


Most marine LPG tankers are fitted with accurate level and temperature gauges and have had the volume of their tanks carefully measured and certified by an independent authority. The information to then allow accurate measurement of cargo volume is contained in their ‘ship’s tables’, which relate tank dimensions to product level and temperature, and the ship’s trim. Well calibrated and instrumented ships should be able to calcultate typical cargo volumes to within 0.2%, but this does not often seem to be the case. Ship’s measurements are seldom used for the sale, although they are commonly referenced as a check on metering figures.


The results presented below were obtained from a equatorially located system for loading up to two 500,000 barrel (80,000 m3) ships at once with the two LPG’s, propane and butane. It is large by world standards in that each product is supplied via 900 mm loading lines, and the total LPG metering station comprises : four 300 mm, two 200 mm, two 150 mm and two 75 mm turbine meters. Each LPG has dedicated to it; two of the 300mm, one of the 200 mm and one of the 150 mm meters. The remaining meters can be ‘lined up’ for either propane or butane.

As always for such facilities the metering station is close to the berth, which in this case means at the end of an 11 Km. jetty. Continuously recirculated loading lines are thus about 12 Km long, which offers considerable scope for a large solar heat influx and its variation with weather conditions.

The loading system’s design and operation includes features to minimize most of the potential problems discussed above. Pipeline blinds were introduced, after the initial commissioning, to reduce dependency on the integrity of some valves -- at the cost of reduced operational flexibility. A computerised interlock system forces the incident flow stream, meters and provers to be cooled to within 0.25 deg.C of each other before proving commenced. Meters are used at almost exactly the same flow rate at which they were proved, to minimise flow rate interpolation errors. Loading rate is restricted, differently at different parts of the loading cycle, to allow ships to process all boil-off vapors with their own refrigeration systems. Although valves move fairly slowly most cargoes are large so take a relatively long time to load. As a result valve speeds are not seen as being problematic from the metering accuracy point of view.

Cargoes taken on board at the site usually total in excess of 60,000 m3 though split, fairly equally, between the two products, propane and butane. Sometimes a full 80,000 m3 of only one of the products is taken. At the other end of the scale ships occasionally call ‘hot’ for only coolant, in which case they are given just sufficient, typically circa 200 m3, to be able to use their own refrigeration equipment to cool down. Within any year about 100 different ships visit the terminal, many of them repeatedly. All have measured tanks, are fitted with level and temperature gauges and are furnished with tables to relate tank measurements, product temperatures, levels and also ship’s trim to the quantity of LPG contained within the tanks.


The results presented in this paper refer, in total, to about 250 loadings, i.e. the loading of approximately 20,000,000 m3 of LPG onto 25 different ships. The ships visited the terminal over a 51 month period, during which interval the most frequent customer called 17 times, the least frequent, 4 times.

In all of the following ‘average’ percentage discrepancies between ship measurements and flow meter measurements are based on the total quantity of appropriate product, whether a single LPG or the two LPG’s combined -- not the average of the individual discrepancy readings. The purpose of this was to avoid the perhaps larger percentage errors produced in loading small cargoes from causing a disproportionate weighting to the overal average.

Table 1. Loading of One Ship over a Four Year Period


Table 1 shows individual loading figures for a total of ten loadings of one particular ship. As can be seen their was, on average, very good agreement between ship measurements and flow meter measurements in that, over the ten separate loadings, there was in total only -0.027% disagreement between the two measurements. The corresponding averaged discrepancy for propane and butane measurements can be seen to be 0.0059 and -0.07, respectively. Larger differences usually occurred for individual loadings.

Discrepancy seemed fairly independent of cargo size, and showed no noticeable trend over the four year period. Interestingly the multiple loading averaged discrepancies for propane and butane were of opposite sign, i.e. the ship records the loading of more propane but less butane than the meters. The difference between ship and meter figures for propane was an order of magnitude better than for butane.

Some of the results of Table 1 are illustrated in the histogram of Graph.1. This perhaps shows better than the table that the discrepancy between ship and metering figures tends to be slightly smaller for butane measurements than for propane measurements, and the difference for the two different products individually tend to track one another with time. The fairly equal scatter of bars above and below the absicca reflects the four year averaged agreement of the two measurement types. The separate loadings; 1, 2, 3 etc. are in chronological order.



Graphs 2, 3 and 4, over page, show separately the averaged loadings of twenty five ships with propane, butane and the total cargo combination, propane and butane. Although the total scatter in all three cases seems, from these graphs, to be fairly even distributed about the y axis the calculated averages show that ships recorded propane loads 0.71% higher than the metering system, butane loads 0.46% higher than the metering system and total cargoes 0.42% higher than the metering system. In these graphs there is no significance in the order of the bars, they merely represent the logging of ships in alphabetical order.






A feature of the results is that they show a noticeable variation in the calibration accuracy of different ships. This is perhaps most meaningfully illustrated in Graph 4, in which each bar represents the average, on average, of 10 separate loadings into one ship. Each loading was typically in excess of 60,000 m3 of LPG. As can be seen the scatter, suggestive of differences in ship calibration, varies by as much as 1.0%. This supports the use of a metering system for the subject loading facility but also, more especially, for those facilities which expected to often make smaller sales, (1). However, the metering system is of little value if it cannot consistently provide more accurate measurements than those obtained from tankage, be the tanks on shore or on the ship.

The results undicate that the subject metering system may not be as accurate as was thought. On average it exhibited a definite tendency to record lower quantities, by about 0.4%, than determined by ship measurements. Alternatively it may be that there is a universal tendancy for ship’s calibrations to be biased, perhaps because of the difficult measurement techniques required.

It would seem surprising if the metering system was much in error because most of the usual ‘problems’ had been avoided as outlined in Section 5, above. Leaking recirculation valves were unlikely to be the cause because they would have resulted in the ships reading lower than the meters, rather than higher as was found to be the case. It was also unlikely that there was leakage via. berth segregating valves when two ships were being loaded at the same time, since at such time appropriate leakage promoting pressure differences were at their lowest. It was, however, perhaps possible that there was some leakage via valves segregating the two berths when only one ship was being loaded, i.e. those times when pressures were high at the loading berth but low in the recirculated berth.

For the averaged ship’s figures to be biased seems as unlikely as the metering system to be in error. Although details were not available it is not unreasonable to expect that the ships were calibrated at a variety of different locations, by a variety of different agencies using a variety of different techniques. They were owned by many different shipping companies and their names suggested widely dispersed origins. This very diversity seems compatible with the scattered discrepancy between ship and meter measurements, but not the net difference.

Shore tankage measurements could have thrown some useful light on the disparity, especially since most discharges were large, but unfortunately this data was not available. Likewise, ship discharge figures related to receiving port measurements might have been illuminating, but these too could not be obtained.

The above results confirm the validity of using butane to prove meters for use on propane, in that the general apperance of Graphs 3 and 4 is very similar, as were the discrepancies between ship and meter measurements averaged over all 25 ships. All the differences between propane and butane that might be envisaged causing a difference in metering, e.g. different; temperature, propensity to produce vapour, viscosity, lubricating qualities and inclination to promote ‘erratic’ meter factor curves, seemed to be of little significance -- or at least coolectively tended to average.

An interesting feature of such a ship loading facility is that it should have a tracability which is independent of the route via the provers. The alternative tracability is via the ships which the facility loads. Although this second route is more tenuous and prone to errors in the different steps along the way to the relevant standards, it should at least yield results in fair agreement with the normal route tracability, via the provers. Obviously the disagreement between the two would be expected to be larger for one loading of one ship than for many loadings of the same ship. The averaging of errors caused by inaccurate allowance for ship’s trim being one obvious reason for better agreement when averaging over many loadings. Similarly, still better agreement would be expected when averaging many loadings of many different ships. The improvement in this case being caused by calibration error of any one ship being averaged out amongst the many others.

The fact multiple loadings of one ship did not result in much improvement of the two tracabiltiy paths was interesting, Ref. the most right column, individual loading figures, of Table 1. compared to the average ‘total’ figure at the bottom of that same column. More puzzling was the similarity between Graph 1, showing the many loadings of one ship, and Graph 4, showing the many loadings of many ships. The ultimate surprise was that the average of the many results in Graph 4 should be as large as 0.417%, since it reflects a surprisingly large discrepancy between the two different tracability paths. Clearly more investigation is necessary. A more extensive analysis of the data may provide better insight.

For future studies it would seem particularly desirable to accompany new data of the type used in this paper, with accurate shore tankage measurements and a fairly detailed account of the operational circumstances accompanying each loading. It would also be particularly interesting to discover if other, similar facilities revealed comparable results, likewise, how LPG loading facilities compare with those for oil. If the major part of the problem lies with the metering facility one might expect the oil facilities to show better result. If it lies with ship calibrations results may be very similar although they may be even worse since, on average, oil tankers are probably calibrated less accurately than LPG tankers.


1. Battye, J.S. et al, “The Design of a Refrigerated Hydrocarbon Metering System of Unprecedented Size”, International Conference on Advances in Flow Measurement Techniques, University of Warwick, England, 1981.