Contribution of Visitors to the Indoor PM in the National Library in Prague , Czech Republic

Temporal and spatial variation of size resolved particulate matter (PM) was measured in the Baroque Library Hall of the National Library in Prague during four intensive campaigns by an Aerodynamic Particle Sizer and three DustTrak instruments. The analysis showed, the indoor air can be considered as well mixed and therefore the simple mass balance model was used to estimate basic parameters (ventilation rate, deposition rates and penetration factors). The results revealed that the movement of visitors during visiting hours is the main source of indoor coarse PM. Mass emission rates, estimated from the comparison of measured and modelled data indicates the visitors should contribute about 50 g of coarse particles yearly, which is about 35% of the total load of indoor PM.

In the period 2009-2010 we performed detailed indoor air quality measurements in an old Baroque Library Hall (BLH), located in Clementinum historical complex in Prague.Founded in 1232, the Clementinum is one of the largest building complexes in Europe.Since 1930 it is currently in use as the National Library of the Czech Republic.The measurements covered monitoring of indoor and outdoor airborne PM, gaseous pollutants and climatic parameters and determination of chemical composition of PM.Moreover, spatial variation of indoor PM was also monitored.The purpose of these measurements was to obtain detailed information on indoor air quality in the library before revitalization of Clementinum.The revitalization, planned for 2010-2018 includes also a construction of new airconditioned depositories located on the same floor as BLH.
Results of gaseous pollutants monitoring and PM composition revealed, that infiltration of ammonium nitrate from the outdoor air and shift to the equilibrium to the gas phase were the reason for the high concentrations of ammonia measured inside BLH (Andělová et al., 2010;López-Aparicio et al., 2011).Temporal variation of size resolved indoor and outdoor PM showed that main source of indoor coarse PM are visitors and fine PM infiltration from the outdoor environment (Chatoutsidou et al., 2015).Similar effect has been already described in several studies related to indoor air quality in museums (e.g., Camuffo et al., 1999;Worobiec et al., 2008;Wang et al., 2015;Xiu et al., 2015).Further, the time behavior of PM 10 concentrations measured in various parts of the hall by 3 DustTrak instrument showed that, for the purpose of modelling, the indoor environment can be considered as well mixed.The assumption of ideal mixing was verified by multi-compartment modelling (Takkunen et al., 2011).The time and size resolved PM number concentration data were also used to estimate particle penetration factor P and deposition rate k.The estimation was based on comparison of measured and modelled data, however no unique solution was found (Chatoutsidou et al., 2015).This is probably due to a nearly linear inverse relationship between particle penetration and deposition that occur simultaneously (Bennet and Koutrakis, 2006).There are several studies that tried to determine P and k separately by increasing the indoor PM concentrations significantly.This was achieved either by opening of windows and doors and measurement of subsequent decay of particles indoors after windows and doors were closed (Vette et al., 2001;Chao et al., 2003;Smolík et al., 2005) or by resuspension activities indoors followed again by period of particle decay (Thatcher and Layton, 1995;Thatcher et al., 2003).
In preliminary tests, performed in BLH in spring 2008 we measured concentrations of coarse PM and CO 2 close to the visitor's route.One week measurements showed strong influence of visitors who periodically increased concentrations of both pollutants at the beginning and during visiting hours followed by decay after.Further two unidentified activities occurred indoors during these tests that increased PM concentrations significantly.However, those data have not been analysed since both episodes have been considered as untypical for standard activities in BLH and only indoor concentrations were measured.The present study examined if those data would help to determine P and k values separately.

Sampling Site
Sampling site has already been described in previous studies (Andělová et al., 2010;López-Aparicio et al., 2011;Chatoutsidou et al., 2015) and so only details related to this study are mentioned.PM concentrations were measured in an old BLH of the National Library (Clementinum historical complex) located in the Old Town of Prague in the city centre.The interior of the BLH remained intact since the 18th century when it was opened.It houses over 20,000 volumes of mostly foreign theological literature dating from the 16th century until recent times stored in original wooden shelves.The hall is decorated with ceiling frescoes and holds also a collection of geographical and astronomical globes.The BLH has natural ventilation through small openings in the building, windows, and doors.All windows with double glass and covered by curtains are closed while the doors are open only for visiting purposes.Approximate indoor geometrics of the BLH are presented in Table 1.
There are 4 entrances to the BLH, 2 at the north side and 2 at the south side.At the north side the entrances lead from hallway, which serves as a storage room and as an entrance for librarians and restorers.The south side doors lead to the lobby of the BLH and serve as an entrance and exit for visitors.Visitors enter the Hall in groups of max. 25 people with a guide, tours run only along the south side of the Hall.Sightseeing tours took place every day from 10 a.m. and finish differently by the season.A blueprint of the BLH is shown in Fig. 1.

Data Collection
A preliminary test of indoor coarse PM behavior was carried out in spring 2008, followed by three intensive campaigns in spring, summer and winter 2009.In the first test temporal variation of indoor coarse PM was monitored using an Aerodynamic Particle Sizer (APS, model 3321, TSI, USA), located at distance about 5 m from the route for visitors (see Fig. 1), that measured particle number concentrations for 52 specific size bins in effective size range 0.5-20 µm.In addition temperature, relative humidity and CO 2 concentration were measured by an Indoor Air Quality Monitor PS32 (Sensotron, Poland).In 2009 the intensive campaigns measurements were performed using the APS located at distance about 15 m from the route for visitors (see Fig. 1).The instrument sampled from both inside and outside BLH using its own sampling train provided with an electrically actuated three-way ball valve connected to common programmable controller that used an SMPS voltage sent by Condensation Particle Counter (CPC, model 3775, TSI, USA) as a signal for switching.The APS was operated with 5 L min -1 flow rate and used 3 min scan, followed by 2 minutes delay necessary to separate samples and flush sampling train after valve switching.Eventually two five minute sampling cycles for indoor sampling followed by two five-minute cycles for outdoor sampling were used.Data were collected using Aerosol Instrument Manager software (AIM v.8.0, TSI, USA), where particle losses inside sampling trains were incorporated.
In addition spatial distribution of indoor PM was measured with three Dust Trak instruments (DT, model 8520, TSI, USA) during summer and winter 2009.One DT instrument was located just at the visitor's route at height about 0.3 m above floor (DT Visitors), two others were located at the sampling point at the height about 1 m (DT Centre) and 7 m (DT Gallery) above the floor (see Fig. 1).DT monitored PM 10 mass concentrations by detection of scattered light and measured particles from 0.1 to 10 µm based on optical diameter with one minute interval.

Indoor Mass Balance Model
The indoor particle concentration for a well-mixed air volume can be described using a zero-dimensional dynamic mass balance model for a given size of particles I (Nazaroff, 2004;Hussein and Kulamala, 2008): where t is the time [h], C i in (t) is the indoor particle  concentration [# m -3 ], P i is the penetration factor (dimensionless), a is the ventilation rate V is volume of the space under study [m 3 ] and N is a where t is the time [h], C i in (t) is the indoor particle concentration [# m -3 ], C i out (t) is the outdoor particle number of particles size fractions.The Eq. ( 1) assumes that the concentration of indoor particles is a result of particle penetration from outdoors, deposition on indoor surfaces, exfiltration from indoors to outdoors and emissions from indoor sources and ignores coagulation and phase change processes.
In the case indoor concentrations are increased significantly so that decay by deposition and exfiltration dominates and simultaneously emissions from indoor sources and particle penetration from outdoors can be neglected we could reduce the mass balance Eq. ( 1) into (Hussein et al., 2009a, b): For period without indoor sources, without deposition (k i = 0) and with ideal penetration factor P = 1 similar mass balance equation can be written for carbon dioxide: where 2 ( ) Eq. ( 1) can be solved by using discreet time steps: Eq. ( 2) has analytical solution in the following form: where t 0 is the initial time and ,0 Further, assuming the outdoor concentration of carbon dioxide 2 out CO C is constant, integration of Eq. (3) yields: where 2,0 in CO C is indoor concentration of carbon dioxide measured at time t 0 .

RESULTS AND DISCUSSION
For analysis of coarse particle behavior we use data from preliminary test performed in 2008 and measurements of spatial variability of indoor PM 10 and temporal variability of indoor/outdoor coarse PM measured in 2009 (see Fig. 1).Results of spatial variability of PM 10 , measured by three DT are shown in Fig. 2. It should be mentioned that one instrument was located just in the area of visiting tours and thus should monitor predominantly PM generated by visitors.Two others were located at a distance approximately 15 m from the visitors.As can be seen temporal variability of PM 10 measured by all instruments exhibits the same trend indicating that for purpose of modelling the indoor air can be considered as well-mixed.This was confirmed also by the multi-compartment modelling (Takkunen et al., 2011).
Temporal variability of coarse PM and CO 2 measured in preliminary tests performed in 2008 is shown in Fig. 3.As can be seen both PM and CO 2 concentrations periodically increase at the beginning of visiting tours and reach maximum value at the end of visits, followed by decay.From the Fig. 3 it can be also seen two episodes with resuspension activities that raised the indoor PM concentrations substantially.The peak concentration achieved after the first event was about an order of magnitude higher than peak concentrations reached after the visits and about 4 times higher than average outdoor concentrations measured during three campaigns in 2009.

Estimation of Ventilation Rate
Since value of 2 out CO C has not been measured, it was gradually chosen from the interval 〈400, 600〉 ppm and subsequently, the coefficient of ventilation rate was obtained by the least square method applied on the Eq. ( 6).The final values of parameters (a, 2 out CO C ) were chosen on the base of minimal cumulative standard deviation of the used interpolation.The average value a = 0.13 h -1 was stable with the relative standard deviation (RSD) approximately 14%.The same average value was found in spring campaign in 2009 (Chatoutsidou et al., 2015).

Estimation of Deposition Rate
In the next step deposition rates k i were estimated using data for exponential decay observed on 20 May 2008, when indoor concentrations C i in were significantly elevated.For this purpose data of the upper part of decay curve were fitted for different size fractions using the least square method by Eq. ( 5).However, there were very low concentrations for particles larger than about 2.5 µm resulting in insufficient counts to provide adequate count statistic and therefore these data were excluded.The values of k i as a function of particle size found from the fit are compared with results for four sizes fractions found from the best fit between the measured indoor concentrations and the obtained modelled indoor concentrations (Chatoutsidou et al., 2015) in Fig. 4. Shown are also deposition rates for fine particles estimated from SMPS measurements performed in spring 2009.As  can be seen both methods give practically the same results in the size range where both measurements overlap.Note also that measurements with significantly elevated indoor concentrations were performed in spring 2008 while measurements evaluated by best fit method were performed in spring 2009.It indicates quite stable conditions in BLH.
For particle sizes larger than about 1 µm the deposition rate k i relates to settling velocity V S (Lai and Nazaroff, 2000): where A u is the area of all upward-facing horizontal surfaces.This should differ from floor area since there are shelves with books on all walls of ground-floor and gallery and furniture including collection of geographical and astronomical globes but should be constant for all coarse particles.Thus to estimated it we calculated parameter A u /V using experimentally determined deposition rates and gravitational settling velocity assuming that all particles are spheres with density of 2 g cm -3 .The results showed that the parameter A u /V is constant with the average value 0.42 m -1 and the RSD is approximately 3%.This confirms that settling was the primary deposition mechanism for coarse particles.Deposition rates, modelled with this parameter using three-layer model (Hussein et al., 2012), are shown in Fig. 4.

Estimation of Penetration Factor
Now that we know both a and k i values for all measured particle sizes, reduced indoor mass balance model Eq. ( 2) was used with data from night-time non-source periods to obtain the size dependence of the penetration factor.Optimizing modelled values we obtained the dependence of the penetration factor on the particle size, shown in Fig. 5.As can be seen results for particles larger than about 2 µm exhibited larger scatter probably due to very low concentrations of these particles during night.Thus for the next analysis the data were extrapolated using parabolic type of expression suggested by Fuchs (1964) for penetration through a crack resulting from gravitational settling.The penetration factors are compared with results of analysis based on the best fit method (Chatoutsidou et al., 2015) in Fig. 5, where results for fine particles based on SMPS measurements are also shown.As can be seen there is continuity with results for fine particles but large difference for coarse particles.This can be caused again by low concentrations of larger particles during night and also by broad size bin (3-20 µm) considered for particles > 3 µm in previous analysis.Comparison with data on penetration factor published by other authors has been presented in the previous paper (Chatoutsidou et al., 2015).If we compare results of both analyses, the data from elevated indoor concentration resemble results obtained from experiments in laboratories, while data from the best fit resemble results obtained from experiments in real buildings (Chen and Zhao, 2011).

Estimation of Visitors Influence
In order to estimate impact of visitors inside the library ventilation rate a, deposition rates k i and penetration factors P i determined in the previous part were used to model particulate emissions resulting from the movement of visitors during visiting hours.Since k i was determined only for size range 1-2.5 µm it was extrapolated to larger sizes using Eq. ( 7).For this purpose the full set of data, including visiting hours was used and deviation between modelled values   mod, in i C t calculated by Eq. ( 4) and measured values   in i C t was minimized for S i = 0 out of visiting hours and S i = const in visiting hours.The results showed that visitors emit on average 5 × 10 6 # min -1 in the size interval 0.7-10 µm.The estimated particle number emission rates were converted to mass emission rates assuming that all particles are spherical with density 2 g cm -3 .The average emission rates calculated for four size intervals PM 0.7-10 as the sum of mass emission rates for differentiated size bins are given in Table 2.
As can be seen the most of the particle mass generated during visiting hours was larger than 1 µm in diameter.This agree well with observation of Thatcher and Layton (1995), Abt et al. (2000) and Hussein et al. (2015a) who found that movement in the indoor environment significantly    et al., 2010).Figs.6(a)-6(c) provides an example of comparison between indoor concentrations measured in spring 2009 and indoor concentrations calculated using values of k i , P i , and S i estimated in this analysis.The dash-dot line indicates the measured indoor concentrations, the dash line indicates the concentration of particles counted by the model that does not include the emission source term, and the solid line indicates the concentration obtained using the full model, which also includes visitors as a source of particles.Fig. 6(a) shows comparison for the 1-2 µm particle fraction and Fig. 6(b) comparison for the 2-3 µm particle fraction, where k = 0.2 h -1 , P = 0.4 and S = 1.5 × 10 6 # min -1 and k = 0.6 h -1 , P = 0.1 and S = 3.0 × 10 5 # min -1 , respectively.As can be seen both models reproduce well the observed indoor concentrations that are influenced mainly by penetration from the outdoor air and deposition on indoor surfaces.Fig. 6(c) shows comparison for the 3-5 µm particle fraction, where k = 1.5 h -1 , P = 0.0 and S = 1.5 × 10 5 # min -1 i.e., for particles that should not penetrate from the outdoor air but should be generated by the visitors only.As can be seen model without emission source predicts zero indoor concentrations while the full model predicts well periodical increase caused by visitors followed by decay after visiting hours due to deposition on indoor surfaces.

CONCLUSIONS
Temporal and spatial variation of size resolved PM monitoring, performed in the Baroque Library Hall of the National Library in Prague during several intensive campaigns, revealed that the visitors are the main source of indoor coarse PM.As the previous analysis showed the indoor air can be considered as well mixed, the simple mass balance model has been used to estimate basic parameters of the BLH -ventilation rate a, deposition rates k i and penetration factors P i .The parameters were used to model particulate emissions resulting from the movement of visitors during visiting hours.Mass emission rates, estimated from the comparison of measured and modelled data indicates the visitors should contribute about 50 g of coarse PM yearly, which is about 35% of the total load of indoor PM.

Fig. 3 .
Fig. 3. Temporal variation of indoor coarse PM and CO 2 concentrations measured in spring 2008.

Fig. 4 .
Fig. 4. Deposition rate as a function of particle size.

Table 1 .
Approximate indoor geometrics of the BLH.

Table 2 .
Estimated values of mass emission rates.