The Chemical Engineering Journal, 55 (1994) 61-68 61

Evaporation loss from solvent tanks

H. Schmidli and C. Weihs

Ciba-Geigy Ltd., Mathematical Applications, CH-4002 Base1 [Switzerland)

(Received June 15, 1993)

Abstract

A simple thermodynamic model is derived linking the daily evaporation loss from solvent tanks to easily
available meteorological data. The model is checked by comparison with temperatures measured at two
CIBA tanks for almost 2 years. The distribution and statistical summaries of the daily evaporation loss
can be calculated approximately using the so-called delta method. Since the model is valid for many
types of tank and solvent, it facilitates the evaluation of antipollution devices and methods for the reduction
of solvent loss, and allows the quantification of mean and extreme daily evaporation loss for many types
of existing tank.

1. Introduction fluid temperatures are known, a model for the

evaporation loss depending only on meteorological
data, tank and solvent parameters is obtained. From
Evaporation emission from solvent tanks causes
this simple model, the distribution of the daily
air pollution, which is regulated in Switzerland by
evaporation loss, based on meteorological mea-
the Swiss ‘ ‘Luftreinhalte-Verordnung (LRV)” [ 11.
surements from 1981 to 1989, is calculated ap-
Since the amount and distribution of daily evapo-
proximately using the so-called delta method.
ration loss was only crudely known, measurements
Previous attempts to calculate the mean daily
at two tanks at CIBA Base1 over a period of almost
evaporation loss are described for example in refs.
2 years were used to give more information about
2-4; for more references, see ref. 5.
the evaporation loss from solvent tanks.
The main goal of this study was to derive a simple
model for the distribution of the daily evaporation
loss for many types of tank and solvent by linking 2. Measurements
the evaporation loss to easily available meteoro-
logical data. This makes it possible to calculate Prom May 1988 to November 1989 fluid, gas and
statistical summaries of the daily evaporation loss three wall temperatures were measured at 10 min
for existing tanks or to evaluate methods for the intervals for two 100 m3 tanks at CIBA in Base1
reduction of solvent loss. Statistical summaries are, (Switzerland) containing the solvents dichloro-
for example, the mean of the daily evaporation loss methane (CH,Cl,) and isopropyl alcohol (C,H,O)
or the 95%-quantile of the daily evaporation loss, respectively. Additionally, the air temperature near
i.e. the value such that the daily evaporation loss the tank was measured.
is smaller than this on 95% of days. Of these measurements, only the fluid and gas
The mean of the daily evaporation loss can be temperatures are used for modelling. The three wall
used to quantify the yearly evaporation loss; the temperatures might be of use for precise modelling
95%-quantile of the daily evaporation loss helps us of the gas temperature, especially with respect to
to design antipollution devices. engineering aspects such as the compression re-
After a short description of the available mea- sistance (see ref. 6). The air temperature near the
surements, a thermodynamic model is presented tank is mainly measured for comparison with mea-
relating easily available meteorological data (air surements by the official weather station.
temperature and global sun h-radiance) with gas and In addition, hourly measurements at the weather
fluid temperatures for given tank parameters. Since station in Binningen, only 5 km from Basel, of air
the evaporation loss can be calculated if gas and temperature and global sun h-radiance were available

0923-0467/94/$07.00 0 1994 Elsevier Science S.A. All rights reserved

SSDI 0923-0467(93)02822-E
62 H. Schmidli, C. We& / Evaporation jkom solvent tanks

ISOPROPYL FiLCOHOL : GAS, FLUID, f?IR TEMPERATURE RN0 SUN IRR.

The following thermodynamic model for the gas
temperature, depending on air temperature and
global sun h-radiance on the tank can be derived
(see for example ref. 6):
E/(Y,q+T,=Tg+a,T,+a,i;, (1)

where T,, Tg and q are air and gas temperatures

and the global sun h-radiance on the tank, iii, and
i;s are the first and second derivatives with respect
to time of Tg, a, and E are the air-heat transfer
coefficient and the absorption coefficient, and a,,
a2 are tank and solvent specific constants.
From the meteorological stations, only the global
sun h-radiance Q on a horizontal plane is available
and not the global sun n-radiance on the tank q.
The global sun irradiance on the tank for a freely
standing tank is given approximately by
0.0 6.0 12.0 18.0 24.0

TIME [HOURI q= TRIH + Wan(+) Q

21rRfH+ 27~
(2)
Fig. 1. Daily profiles of temperatures and sun &radiance for
isopropyl alcohol (November 12, 1988): F fluid temperature, where R and H are the radius and height of the
G gas temperature, A air temperature, S sun irradiance.
tank and 4 is the angle of incidence of the sun.
Note that the factor 6= q/Q is the ratio of the
(measurements from 1981 to 1989). Additional mea- tank surface projected along C#J
on a horizontal plane
surements available such as the amount of rain, to the total tank surface.
wind conditions, etc. proved to be of relatively Simple formulae for three basic situations can be
limited use for the present purpose.
derived:
Typical daily profiles of temperature and sun
it-radiance are shown in Fig. 1. no sun on tank S=O
rRIH
sun on top of tank S = (2?rR,H + 2*)
3. Thermodynamic basis
rR/H + 2/tan(90° - $)
To derive a simple model for the evaporation fullsunontank6=
2gRIH-k 2~7
loss, three basic assumptions are made: the vapor
behaves like an ideal gas; the vapor concentration where $ is the geographic latitude of the tank
is proportional to the vapor pressure, which is a position.
function of the fluid temperature; the gas temper- If only the daily evaporation loss from solvent
ature increases steadily between daily minimal and tanks is of interest, we need to know only the daily
maximal gas temperatures and does not increase minimal and maximal gas temperatures and the
otherwise. mean fluid temperature. Using model (1) and the
Based on these assumptions, it can be shown facts that temperatures are nearly constant at night
that the daily evaporation loss from solvent tanks and derivatives vanish at minima and maxima, these
only depends on the daily mean fluid temperature can be calculated approximately by
T,““” and the daily minimal and maximal gas tem-
peratures Tgminand Tgmm(see for example ref. 2). Vi” = Tamin
The goal is now to predict the daily evaporation (3)
Tg’- z ~/a aSQ”” + T,“”
loss from easily available meteorological measure-
ments, i.e. the daily mean, minimum and maximum Since a model like (1) holds for the fluid tem-
of air temperature and the daily mean and maximum perature as well, by integrating over 1 day one
of the global sun irradiance. Thus, first a simple obtains an initial estimate of the daily mean fluid
model is derived for prediction of the daily mean temperature, using the fact that the fluid temperature
fluid temperature Tfmemand the daily minimal and is nearly constant for 1 day:
maximal gas temperatures Tgmhand Tgmaxfrom the
meteorological measurements. (Ja)
 H. Schmidli, C. Weihs / EvaporatiolL from solven.8 tanks 63

Because of the slow change in fluid temperature TABLE 1. Parameters

the final estimate of the daily mean fluid temperature

is taken to be a weighted average of the initial General parameters
M,=O.O224 m3 mol-’ molar volume
estimate of the mean fluid temperature and its
‘I’,,=273 K normal temperature
predecessor, based on empirical evidence, i.e. for pn= 1.013 bar normal pressure
day d
Tank parameters
T,““““[d] = 0.3i;;l”“[d] + 0.7?fmean[d- I] (4b) * geographic latitude of the tank position (deg)
s, saturation factor of gas, i.e. the ratio of actual
To check the quality of these approximations, concentration to saturation concentration,
the temperature measurements at the two tanks and typically 1.0 for storage tanks and 0.6 for
other tanks (see ref. 2)
from the nearby weather station were used. The
f gas volume proportion, i.e. the proportion of
mean prediction errors are small for both tanks and the tank volume filled by gas, typically 0.075
all three temperatures (less than 5%). for storage tanks and 0.5 for other tanks (see
ref. 2)
R radius of tank (m)
H height of tank (m)
4. Daily evaporation loss heat transfer coefficient between air and tank
ffa
(W m-’ K-r)
The daily evaporation loss L (kg) based on ref. absorption coefficient
2 is a function of the minimal and maximal gas : sun exposure factor (0 no sun on tank, 0.5 sun
on top of tank, 1 full sun on tank)
temperatures Tgmi”,Z’gma and the mean fluid tem-
6 according to eqn. (2)
perature Z’,“‘” (K) given by
Solvent parameters
M molar weight (kg mol-‘)
a, b, c, d Reid constants (called VP A, VP B, VP C, VP D
= Ic&(T;“““)/T;“““( l/T,“‘” - l/,gm”“) (5) in ref. 10)
P, critical pressure (bar)
where k =s~~R~HM/M~(T,,~/~,,) is a constant and T, critical temperature (K)
~47’) is the vapor pressure (bar) at fluid temperature
T (K) (see below). The parameters are explained
in Table 1.
by E= 0.2 for the tanks on which the measurements
Substituting the gas and fluid temperatures in
were performed. The heat transfer coefficient (be-
eqn. (5) by their approximations eqns. (3), (4a)
tween air and tank) CX,is assumed to be constant
and (4b), the daily evaporation loss can be expressed
and approximately equal for tanks which are not
as a function of the meteorological measurements
insulated. It is set to a conservative value of (Ye=
as follows:
5 W rnp2 K-r (see, for example, refs. 8 and 9).
The tank parameters for the two observed tanks
are given by radius R = 2 m, height H= 10 m and
geographic latitude t,!r=40”. The sun can shine only
on the top of the tank containing dichloromethane
(S = 0.5), while the other tank is almost fully exposed
to the sun (S= 1). Neither tank is a storage tank,
where the meteorological measurements are de&red so the saturation factor is s, = 0.6 and the gas volume
as x1 = Tamean[d], x2 = Tamin[d], x3 = Tamax[d], proportion f= 0.5.
x4 = (Qmem[d])1’2, x5 = Q”[d], x6 = T,“‘“[d - 11, The vapor pressure is calculated according to
x7 = (Q”‘“[d - l])“? T,“““[d], T,“‘“[d], Tamax[d], Reid et al. [lo] by
Q”‘“[d] and Q”“[d] are the extreme and mean air
temperature and global sun h-radiance at day d. pr(T)=exp[(uz+bz’.5+cz3+dz6)/(1-z)] (7)
The reason for taking the square-root of Q”“”
with
is that the variables xi, . . . . x7 were identified to
be approximately normally distributed based on x=1-T/T,
historical data (1981-1989). The normality of the
variables x1, . . . , x7 is used in the next paragraph. The Reid constants, the critical temperature, the
To calculate the daily evaporation loss, the pa- critical pressure and the molar weight can be found
rameters listed in Table 1 are needed. The absorption for many solvents in ref. 10, Appendix A, Property
coefficient E is tabulated for different tank wall Data Bank. For the two considered solvents, these
materials in ref. 7 Chapters 3 and 4, and is given constants are listed in Table 2.
64 H. Schmidli, C. Weihs / Evaporation from solvent tanks

TABLE 2. Solvent properties be transformed to a normal distribution, where the

transformations are the so-called power transfor-
Property Isopropyl alcohol Dichloromethane mations (y = (z”- l)/ar, crareal number; for example,
M (kg mol-‘) 60.1 x 1O-3 84.93 x 1o-3
a=0.5, y=2(&-1); (Y= 0, y = In(z), by limit con-
: -8.16927
-9.43213x 1o-2 2.17546
- 7.35739 siderations, (Y--) 0). Unfortunately, the optimal trans-
formations to normality generally depend on both
: 7.85
-8.10040 3.50701
- 4.07038 tank and solvent parameters. To apply the delta
method, a fixed transformation is needed which is
PC (bar) 47.6 63.0
T, (K> 508.3 510.0
independent of tank and solvent parameters. Hence
a compromise transformation has to be found. An
investigation of the optimal transformations for dif-
ferent tanks and solvents suggests that for summer
days, the evaporation loss is approximately normally
5. Distribution of the evaporation loss distributed (CLI= l), and for winter days the logarithm
of the daily evaporation loss is approximately nor-
mally distributed (a= 0) (for a definition of summer
For the evaluation of methods for the reduction and winter days, see below). The effect of taking
of solvent loss, the evaporation loss for a specific these fixed transformations instead of optimal power
day is of limited use. What is needed is the dis- transformations is illustrated below.
tribution of the daily evaporation loss over a large
A normal distribution is uniquely defined by its
period of time or corresponding statistical sum-
mean and its variance. Therefore the distribution
maries such as the mean. There are at least two
of the daily evaporation loss is specified if for summer
ways of obtaining this distribution of the daily
days the mean and the variance of the daily evap-
evaporation loss for given tank and solvent param-
oration loss can be calculated, and if for winter
eters. The first method gives the “exact” distribution,
days the mean and the variance of the logarithm
the second so-called delta method gives a good
of the daily evaporation loss can be calculated. The
approximation.
delta method can be used to calculate these means
The “exact” method consists in calculating, ac-
and variances.
cording to the preceding section, the evaporation
The basic idea of the delta method can be sum-
loss for every day of the years 1981-1989, say
from the daily mean, minimum and maximum air marized as follows. Assume that a p-variate random
temperature and the daily mean and maximum global variableX=(zi, . . . . x,) has a p-dimensional mean
sun irradiance (data from the nearest meteorological m and a p Xp covariance matrix C. Let the random
station). The distribution of the daily evaporation variable Z=L(X) be a function of X. Using Taylor’s
loss can then be summarized by, say, the mode, theorem to linearize the function L, the mean and
the mean and the 95%-quantiIe, i.e. the most frequent variance of 2 are approximately given by
value, the arithmetic average and the value such
that the evaporation loss is smaller than this in 95% p =EQ =L(m)
(8)
of cases. The main disadvantage of this approach a2 = VARQ -D(m)‘CD(m)
is computational since the calculations are slow and
require a large amount of storage. Therefore, the where the p-dimensional vector D(m) is the first
delta method is discussed in the following, which differential of L at m, i.e. itsjth coordinate is defined
gives a good approximation of the distribution of as (Wm))j = (aLIaXj>lm.
the daily evaporation loss, without the computational The delta method is now applied, using the loss
burden of the “exact” method. function L@,, . . . , z,) and its logarithm from eqns.
The delta method does not require storage of all (6) and (7) separately for summer and winter days,
the data from the meteorological station. Instead, where summer and winter days are defined as in
these data are condensed in such a way that not ref. 2 (153 summer days from May to September
too much information is lost. From these condensed and 212 winter days from October to April).
meteorological measurements of the years
1981-1989, one can derive the distribution of the
daily evaporation loss or summaries such as, say, 5.1. Summer daps
:he mode, the mean and the 95%-quantile. One first needs the mean m, and the covariance
The delta method wilI give a good approximation matrix C, of the variables xi, . . . , x7 for summer
f the distribution of the daily evaporation loss can days; these are calculated from the historical summer
 H. Schmidi, C. Weihs / Evaporation from solvent tanks 65

data of the years 1981-1989, obtained by the nearby

Predicted evaporation loss : isopropyl alcohol 1381-83 Minter
weather station.
The mean of variable xi is calculated as mSi=
IlnCIz_ Ixik, and the covariance between variables
xi and xj is calculated as c,,. =
l/~Z;_I(x~-mmsl)(x~-m,) where xi’, . . . , xi” and
xj ) ..‘, xjn are the measured historical summer
data for variables xi and xj respectively.
The mean m, = (m,,, . . . , m,,) of the variables xl,
. . . , x7, a 7-dimensional vector, is then computed
as

mS, msL ms3 %r -5 %e %,

289.52 284.92 294.49 13.49 647.32 289.51 13.49 0.3 0.4 0.1
a.0 (1.1 0.2

The 7 X 7 covariance matrix C, has in row i and Predicted loos Ckg/ORVI Cull sun (actual)

column j the covariance C,, between variable xi

and xj. It is computed as

15.11 11.55 18.59 5.96 341.63 13.05 6.17

Predicted svaporstion loss : isopropyl alcohol 1981-89 sumer
11.55 12.12 11.94 1.09 71.41 11.74 3.06
18.59 11.94 25.43 10.16 595.43 14.54 8.68
5.96 1.09 10.16 9.70 578.91 2.83 4.49
341.63 71.41 595.43 578.91 38901.35 162.92 259.43
13.05 11.74 14.54 2.83 162.92 15.30 6.04
6.17 3.06 8.68 4.49 259.43 6.04 9.74

The mean daily evaporation loss for summer days

is estimated by

fiS =L(m 51, *-*9 ms7) (91

according to eqns. (6), (7) and (8).

To calculate approximately the variance of the
daily evaporation loss for summer days, one needs a.0 11.1 8.2 a.3 1.4 a.5

the derivatives of L(x,, . . . , x7) with respect to x1, Predicted loss [kg/DRY1 full run (aclual)
. . ., x7 at m,,, . . . , mS7_ For the formulae for
(D(m,))i= (aL(xl, . . . , x7)/3xi)lm, see ref. 11. The
estimate of the variance of the daily evaporation
loss hS2 is then calculated following eqn. (8). Predicted evaporation loss : isopropyl alcohol 1381-83

5.2. Winter days

A similar procedure has to be applied for winter
days. Again, one first needs the mean m,+.and the
covariance matrix C,., of the variables x1, . .., x7
for winter days (see ref. 11). Then, the mean of
the logarithm of the daily evaporation loss for winter
days is estimated by

iL=logL(?v,, ***, 1))2\N,) (IO)

according to eqns. (6), (7) and (8).

To estimate the variance of the logarithm of the 0.. 1.1 e.2 8.1 a.4 8.5

daily evaporation loss for winter days, one needs Predicted loss [kg/ORVl full sun (actual)

the derivatives of log L(x,, . . . , x7) with respect to

x1, . . . , x7 at m,, . . . , w,. For the formulae for Fig. 2. Distribution of evaporation loss: histogram, observed
(D(m& = (alog L(x,, . . . , X,)/&Q)/_, see ref. 11. data; -, density approximation by the delta method; - - -,
Again, the estimate eW2 of the variance of the density approximation by optimal power-transformation.
66 H. Schmidli, C. Wa&s / Evaporation from solvent tanks

logarithm of the daily evaporation loss is calculated mode = exp(& - 6,“) (12)
using eqn. (8).
The mean is given by
5.3. Winter plus surnm mean = Id, fiS+ d,,. exp(iL, + &,‘/2) l/Cd,+ A) (13)
The estimates of mean and variance of the daily
evaporation loss for summer days, fiS and GS2,and The 95%-quantile, i.e. the value such that daily
the estimates of mean and variance of the logarithm evaporation loss is smaller than this on 95% of
of the daily evaporation loss for winter days, $,+. days, is given approximately by
and &,“, can now be combined to calculate the $5 = q& -t fiS (14)
whole distribution of the daily evaporation loss p(L)
as with

p(L) = {d, /(27r6S2)1’2exp[ -(L-- iQ2/(2&2)1 q = @- ‘{10.95(d, + dJ - dJ/d,}

+ dJ[L(2?&,.2)‘n 1 where @is the standard normal distribution function;

for d,=153 and &=212, we have q=1.179.
X exp[ - (log L - ~&)“/(ZC?~~)]} As already noted, the mean is important for cal-
culation of the yearly evaporation loss, while the
x I/(&+&v) (II)
95%-quantile of daily evaporation is important in
where d, and d, are the number of days in summer connection with antipollution devices and with air
and winter respectively (see Fig. 2 for an illustration pollution restrictions. The mode, i.e. the most fre-
of this procedure). quent daily evaporation loss, is mainly informational
From Fig. 2 it can additionally be seen that the in that it helps to visualize the distribution of the
choice of fixed transformations of the evaporation daily evaporation loss.
loss instead of optimal power transformations seems In ref. 2, an alternative method for calculating
to be justified. The density approximation of the the mean daily evaporation loss is given, which
evaporation loss by the delta method is not much consists essentially in replacing the gas and fluid
farther away from the histogram of the observed temperatures in eqn. (5) by fixed values obtained
data than the density approximation obtained using from temperature measurements. It is not possible
the optimal power transformation. to evaluate the effect of different sun exposure, heat
The mode, i.e. the most probable value of daily transfer coefficients, climatic conditions and geo-
evaporation loss, is then given approximately by graphic locations on the mean evaporation loss by

TABLE 3. Distribution of the daily evaporation loss (kg per day)

Solvent VDI Delta method Historical data

mean
Mean Mode 95% Mean Mode 95%

Full sun
Isopropyl alcohol 0.10 0.13 0.03 0.31 0.11 0.03 0.29
Dichloromethane 1.88 2.23 0.88 4.74 1.97 0.71 4.33
Vinylacetate 0.46 0.55 0.19 1.24 0.49 0.16 1.15
Toluol 0.12 0.14 0.05 0.32 0.12 0.04 0.30
Methanol 0.18 0.22 0.07 0.50 0.19 0.07 0.47
Diethylether 2.08 2.47 0.99 5.20 2.17 0.91 4.79
No sun
Isopropyl alcohol 0.10 0.07 0.02 0.15 0.06 0.02 0.16
Dichloromethane 1.88 1.23 0.54 2.52 1.17 0.66 2.62
Vinylacetate 0.46 0.30 0.12 0.65 0.29 0.10 0.67
Toluol 0.12 0.07 0.03 0.17 0.07 0.03 0.17
Methanol 0.18 0.12 0.04 0.26 0.11 0.04 0.27
Diethylether 2.08 1.36 0.60 2.77 1.30 0.51 2.39

Note that VDI [2] only gives a daily mean value and neither mode nor 95%quantile. The effect of different sun exposure, heat
transfer coefficients, climatic conditions and geographic locations on the mean evaporation loss cannot be evaluated with the
VDI model.
Tank properties used: radius R = 2 m, height H = 10 m, sun absorption E = 0.2, heat transfer coefficient a = 5 W m-a K-i, saturation
factor s, = 0.6, gas volume proportion f= 0.5.
 H. Schmidli, C. Weihs / Evaporation from solvent tar&s 67

Predicted daily e*spw*tkm loss : YIYYL#ElRIE 81-89

Predmled daily wworat~on loss : IOLUOL

Fig. 3. Distribution of evaporation loss.

this method. However, the mean seems to be useful First, for six chosen solvents, with tank parameters
as an upper bound (see Table 3). Other summary as for the observed tank containing isopropyl al-
measures such as the mode or the 95% quantile cohol, the distribution of the daily evaporation loss
cannot be obtained using ref. 2. is calculated by the delta method and compared
The quality of the approximation to the distri- with the “exact” distribution (see beginning of the
bution of the daily evaporation loss obtained by the section). As seen in Fig. 3 and Table 3, the correlation
delta method is judged as follows. is quite good.
68 H. Schmidli, C. Weihs / Evaporation from solvent tanks

As an additional check, for the six chosen solvents, delta method would possibly have to be adapted.
72 very different tank scenarios were fixed and for A PC-program for calculation of the distribution of
each scenario statistical summaries (mode, mean, the daily evaporation loss is in routine use in CIBA.
95%-quantile) were calculated by the delta method This program also includes extensions of the model
and the “exact” method. The tank scenarios are presented here, in that different formulae for the
obtained by using each of the 72 possible com- vapor pressure p, (see eqn. (7)) are included which
binations of the following tank parameters: radius are valid for different kinds of fluid.
R= 2,4 m; height H= lo,20 m; sun exposure S= 0,
0.5, 1; heat transfer coefficient a,=5, 10 W rnp2
K-l; absorption coefficient ~=0.2, 0.4, 0.6. Acknowledgments
Typically, for the important summary statistics,
the mean and the 95%-quantile, an error of less We would like to thank J. Hediger for initiating
than 5% is obtained, but for some worst case and managing the project, C. Feder for his support
scenarios the error can reach 20% (95%-quantile) of the project at its final stages, and P. Vegh for
or 25% (mean). The mode is more difficult to estimate challenging the algorithm and implementing it in
and the errors are typically around lo%, in extreme a PC environment. Moreover, we would like to thank
cases up to 30%. For most purposes, these errors P. Miiller for developing the data recording concept,
seem to be acceptable. If greater precision is needed, J. Gautschi for installation of the measuring equip-
the “exact” method should be used. As an illustration ment and for systems calibration, and R. Stalder
of possible simulations with the model, the effect for daily supervising of the measuring equipment
of sun exposure on the distribution of the daily at work.
evaporation loss is demonstrated for the six chosen
solvents in Table 3, where the tank parameters are
as for the observed tanks, with the exception of References
sun exposure. It appears that the extreme values
(959/o-quantile) are considerably higher for full sun 1 Schweizerische Luftreinhahe-Verordnung (LRV), Schweizer
exposure and the mean evaporation loss is increased Bundesrat, December 16, 1985.
2 VDI-Richtlinien, Emissionsmin~~, R@inerief~
by 60%-709/o from no to full sun exposure. Note
Mineratilvertriebslager, VDI-Richtlinie 3479, VDI-Verlag,
that the error of the delta method is much smaller Diisseldorf, 1985.
in comparison with the effect due to different sun 3 E.H. Walker, R.M. Eltringham and A. Puttik, Evaporation
exposure. losses from petrol storage tanks in the United Kingdom -
a practical survey, J. Inst. Petrol., 41 (1955) 297-325, 382.
4 API Evaporation Loss Committee, Evaporation loss from
fixed-roof tanks, American Petroleum Institute Bull. 2518,
6. Discussion
Washington, DC, 1962.
5 J. Hediger, Tanklager-Emissimwn nach VDI 3479, Inteme
III this paper two methods have been presented, Dokumentation, Ciba-Geigy IVT, Apparate-Technik Engi-
allowing easy calculation of the distribution of the neering-Tools, 1991 (Ciba-Geigy, Basel).
daily evaporation loss for many tanks and solvents. 6 W. Gosslau, E. Miiller and R. Weyl, Thermodynamische
Grundlagen der Behah.eratmung aufgrund von Wittenmgs-
If the “exact” method is used to calculate this
einfliissen; Technische i@xrwachung Bd.26, I, 16-22, Ber-
distribution, then for each calculation, meteorolog- lin, 1985.
ical measurements (daily mean, minimal and max- 7 VVI-Wiirmeatlas, VDI-Verlag, Dilsseldorf, 1983.
imal air temperature, daily mean and maximum of 8 D. Fullarton, J. Evripidis and E.U. S&Kinder, InfIuence of
global h-radiance) must be available for every day product vapor condensation on venting of storage tanks,
Chern. Eng. Process., 22 (1987) 137-144.
of a 10 year period, say. If the delta method is
9 R. Sigel, B. Kuxdorf, R. Meiss and H. Schwarz, Praktische
used, only summaries of the meteorological mea- und theoretische Untersuchung der Beatmung von Fest-
surements are needed for the calculation. Never- stofftanks, Technische ifberzvachung 24, 1, 28-33, Berlin,
theless, the meteorologists will generally not supply 1983.
these summary data so they have to be calculated 10 R.C. Reid, J.M. Prausnitz and B.E. Poling, The Properties
of Gases and Liquids, McGraw-Hill, New York, 1987, 4th
from historical data. A possible disadvantage of the
edn.
delta method is that for tanks in climatic environ- 11 H. SchmidIi, Evaporation loss from solvent tanks, C&a-
ments very different from central Europe the dis- Geigy, Mathematical Applications, December 1992 (Ciba-
tributional assumptions might be violated, and the Geigy, Basel).