towards modelling of construction, renovation and demolition activities: norway's dwelling...
TRANSCRIPT
This article was downloaded by: [Universitetbiblioteket I Trondheim NTNU]On: 12 November 2013, At: 18:37Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK
Building Research & InformationPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/rbri20
Towards modelling of construction, renovationand demolition activities: Norway's dwelling stock,1900–2100Igor Sartori a , Håvard Bergsdal b , Daniel B. Müller b & Helge Brattebø ba Norwegian University of Science & Technology, Department of Architectural Design,History and Technology , N-7491, Trondheim, Norwayb Norwegian University of Science & Technology, Department of Hydraulic andEnvironmental Engineering/Industrial Ecology Programme , N-7491, Trondheim, NorwayPublished online: 14 Aug 2008.
To cite this article: Igor Sartori , Håvard Bergsdal , Daniel B. Müller & Helge Brattebø (2008) Towards modelling ofconstruction, renovation and demolition activities: Norway's dwelling stock, 1900–2100, Building Research & Information,36:5, 412-425, DOI: 10.1080/09613210802184312
To link to this article: http://dx.doi.org/10.1080/09613210802184312
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of theContent. Any opinions and views expressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon andshould be independently verified with primary sources of information. Taylor and Francis shall not be liable forany losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use ofthe Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions
Towardsmodelling of construction,renovation anddemolition activities:Norway’s dwelling stock,1900^2100
Igor Sartori1,H �avard Bergsdal2,Daniel B.Mu« ller2 andHelgeBratteb�2
NorwegianUniversity of Science & Technology,1Department of Architectural Design,History andTechnology, and 2Department ofHydraulic andEnvironmentalEngineering/Industrial EcologyProgramme,
N-7491Trondheim,NorwayE-mails: [email protected], [email protected], [email protected] and
The activities of construction, renovation and demolition related to the dwelling (housing) stock have a strong impact on
bothmaterial and energy demands. A deeper understanding of the dynamics driving these activities is a precondition for a
more consistent way to address material and energy demands. The method presented herein is based on a dynamic
material flow analysis and is applied to the Norwegian dwelling stock. Input data to the model are population and
socio-economic lifestyle indicators such as the average number of persons per dwelling and the average size of
dwellings; these determine the size of the floor area stock. Parameters such as the lifetime of dwellings and renovation
intervals complete the input set. Outputs of the model are the stock and flows of floor area for the period 1900–
2100. Analysis of the renovation activity is given particular attention. Several scenarios are considered in order to test
the model’s sensitivity to input’s uncertainties. Results are compared with statistical data, where the latter are
available. The main conclusion is that in the coming decades renovation is likely to overtake construction as the
major activity in the Norwegian residential sector.
Keywords: building stock, construction, demographics, demolition, forecasting, housing, material flow analysis,
renovation, residential stock, Norway
Les activites de construction, de renovation et demolition relatives au parc de logements ont un fort impact sur la
demande de materiaux et d’energie. Une meilleure comprehension de la dynamique qui anime ces activites est une
condition prealable a une methode plus coherente de traitement des demandes de materiaux et d’energie. La methode
presentee ici repose sur une analyse dynamique des flux de materiaux telle qu’elle est appliquee au parc de logements
en Norvege. Les donnees fournies au modele sont des indicateurs du mode de vie de la population et des indicateurs
socio-economiques comme le nombre moyen de personnes par logement et la taille moyenne des logements; ces
indicateurs determinent la superficie des logements. Des parametres comme la duree de vie des logements et les
intervalles de renovation completent les donnees d’entree. Les resultats du modele sont le parc de logements et sa
superficie pour la periode 1900–2100. L’analyse de l’activite de renovation recoit une attention particuliere. Plusieurs
scenarios sont envisages afin de mettre a l’epreuve la sensibilite du modele par rapport aux incertitudes des entrees.
Les resultats sont compares aux donnees statistiques lorsque ces dernieres sont disponibles. La conclusion principale
reside dans le fait que dans les prochaines decennies, le secteur de renovation risque de depasser celui de la
construction en tant qu’activite majeure du secteur residentiel norvegien.
Mots cles: parc de batiments, construction, demographie, demolition, previsions, logement, analyse des flux de
materiaux, renovation, parc residentiel, Norvege
BUILDING RESEARCH & INFORMATION (2008) 36(5), 412–425
Building Research & Information ISSN 0961-3218 print ⁄ISSN 1466-4321 online # 2008 Taylor & Francishttp: ⁄ ⁄www.tandf.co.uk ⁄journals
DOI: 10.1080/09613210802184312
Dow
nloa
ded
by [
Uni
vers
itetb
iblio
teke
t I T
rond
heim
NT
NU
] at
18:
37 1
2 N
ovem
ber
2013
IntroductionUp to today in many countries the population’s needfor housing has created a large standing stock of resi-dential buildings, also named the dwelling stock.Associated with a dwelling stock there is considerableactivity of construction, demolition and renovation(C&D activities), and a corresponding impact onmaterial use, energy use and waste generation.Despite the importance of the dwelling stock andrelated C&D activities, the understanding of thedynamics and long-term changes in the dwellingsystem is limited. Projections of waste generation andmaterials demand from building systems are often per-formed on the basis of trend analysis (Kohler andHassler, 2002; Kohler and Yang, 2007). The sameapplies to the analysis of energy demand in buildings(Myhre, 1995, 2000; Johansson et al., 2006, 2007).Applying a simple trend analysis might represent a suf-ficiently good approximation if limited to a shortperiod of time, such as the recent past or the immediatefuture. However, it may fail to grasp the long-termeffects. In order to improve the understanding of thedwelling stock dynamics, several authors call for, ormake use of, dynamic modelling (Johnstone, 2001a,2001b; Kohler and Hassler, 2002; Muller et al.,2004; Muller, 2006; Bergsdal et al., 2007a, 2007b;Bradley and Kohler, 2007). A dynamic analysis takesinto consideration the activity levels of the past andtheir interrelations and attempts to explore how thesewill affect the future activity levels.
Construction and demolition are the two activitiesusually receiving the most attention. However,several studies indicate that renovation has becomeor will become the dominating activity related to thebuilt environment; measured either in terms of eco-nomic investments or in floor area. Rønningen (2000)reports renovation activity to 77% of constructionactivity in Norway in 1998, while Kohler andHassler (2002) argue that refurbishment will overtakenew construction as the dominant construction activityin Germany in the years ahead. Similar results arefound for Switzerland by Kytzia (2003), where expen-ditures for usage, maintenance and upgrading of build-ings are reported to exceed expenditures for newconstruction. Caccavelli and Genre (2000) estimatethat more than one-third of total construction’s eco-nomic output in the European Union is related torefurbishment activities, and that this figure is expectedto grow as the housing stock becomes older. Theageing of the housing stock is a fact noted by severalauthors. According to Itard and Klunder (2007), themajority of the housing stock in the European Unionwas constructed after the Second World War. Thesame result is found for the Norwegian housing stockin Bergsdal et al. (2007b). Martinaitis et al. (2007)reports a significant portion of the building stock inCentral and Eastern Europe being constructedbetween the 1960s and the 1990s.
The main purpose of this paper is therefore to proposea methodology for analysing and forecasting the fullrange of C&D activities in a consistent manner, includ-ing renovation.
The modelling approach is based on a dynamicmaterial flow analysis (MFA) initially proposed byMuller (2006) who applied it to the dwelling stock inthe Netherlands. He estimated the floor area stockand flows of construction and demolition, as well asthe corresponding stock and flows of concrete in theperiod 1900–2100. The same model was adoptedand modified in Bergsdal et al. (2007b) and appliedto the Norwegian dwelling stock for the same period.Modifications were made to introduce a simplifiedapproach to the renovation activity, only linked tomaterial’s turnover. The corresponding stock, wastegeneration and material demand for concrete andwood were also estimated. The work presented hereis also based on data for the Norwegian dwellingstock. The renovation flow(s) in this paper is treatedin a more consistent way, linking it directly to thefloor area turnover. Hence, all the C&D activitylevels are measured in physical units as square metres(m2) of floor area per year. The method is intendedto serve as a basis for future analysis of materialsdemand and waste generation, as well as analysis ofenergy demand and opportunities for energy perfor-mance improvement. Both material and energyfigures can, indeed, be expressed as intensities persquare metre. Past and future activity levels and theinfluence and importance of the modelling inputs arediscussed on the basis of scenario evaluations. Theresults are compared against statistical figures wheresuch information is available.
MethodThe conceptual outline of the model is given inFigure 1. Processes and stocks are represented by rec-tangles, flows by ovals, and drivers and parametersby hexagons; solid lines show connections betweenstock and flows, while dashed lines represent influencesof the drivers and parameters. Stocks of populationand floor area, denoted by the letters P and A, aremeasured in persons (pers) and square metres (m2),respectively; the symbol dA/dt denotes the net stockaccumulation. The flows express the amount of floorarea that in a given period of time (nominally a year)enters the stock, leaves the stock or recirculates insidethe stock. These are named new, demolition and reno-vation area flows, respectively, are represented in smallletters by newA, demA and renA, respectively, and aremeasured in square metres per year (m2/year). Inputsto the model are divided in time-series data and para-meter functions. Time-series data are required for:population P, expressed in persons (pers); populationdensity PD, expressed in the average number of
Towardsmodelling of construction, renovation and demolition activities
413
Dow
nloa
ded
by [
Uni
vers
itetb
iblio
teke
t I T
rond
heim
NT
NU
] at
18:
37 1
2 N
ovem
ber
2013
persons per dwelling (pers/dwe); and area density AD,expressed in the average square metres of floor area perdwelling (m2/dwe). Parameter functions affectingthe behaviour of the model are: the estimatedlifetime of buildings, L; and the estimated intervalof renovation for a given material or building’s sub-system, R.
The balance equation is given by:
dA(t)
dt¼ newA(t)� demA(t) (1)
It follows from equation (1) that the new floor area,newA, has to make up for both demolition activity,demA, and additional demand of floor area, dA/dt.
The size of a dwelling stock is driven by the popu-lation’s demand for dwelling services that satisfy theirlifestyle preferences, as the number of persons livingin a dwelling and the dwelling’s size. In otherwords, the stock of floor area is a function of demo-graphic and socio-economic lifestyle parameters.These relations are also acknowledged in Kohler andHassler (2002), and estimation of stock and flows ofdwellings by demand for dwelling services is appliedin both Muller (2006) and Johnstone (2001a). In themodel the two inputs persons per dwelling, PD, andaverage size of dwelling, AD, represent the lifestyle pre-ferences of the population. It is out of the scope of thiswork to analyse the socio-economic context fromwhich such preferences emerge. The input data seriesP, PD and AD are simply acknowledged from historicaldata sources and different scenarios are consideredfor future projection as presented in Figure 2.The data are essentially the same data presented byBergsdal et al. (2007b), and reference is made to thispaper for a more detailed description of the data
sources. Equation (2) gives the stock of floor area asthe product of population and lifestyle indicators:
A(t) ¼ P(t) �1
PD(t)� AD(t) (2)
Figure 1 Conceptual outline of the stock dynamics model
Figure 2 Input data to themodel: (a) population, (b) persons perdwelling and (c) average £oor area per dwelling
Sartori et al.
414
Dow
nloa
ded
by [
Uni
vers
itetb
iblio
teke
t I T
rond
heim
NT
NU
] at
18:
37 1
2 N
ovem
ber
2013
One difference adopted in this paper regards the dataon population. While in Bergsdal et al. (2007b) thedata for population are exactly as given in the statistics,in the present paper they are interpolated by means oflinear regression in order to smooth out the curve in thesame fashion as for the other input data series.This feature is beneficial because the method involvesthe calculation of the stock’s derivative dA/dt(equation 1); derivatives of non-smooth curves couldpresent punctual spikes that are meaningless for thelong-term behaviour while may compromise thereadability of the graphs.
The input data series in Figure 2 are given for theperiod 1800 until present, and with three differentscenarios for the future until year 2100; named thehigh, medium and low scenarios. The high and lowscenarios represent a final value in 2100 of +15%compared with the medium scenario.
Once the stock of floor area is known it is possible tocalculate the stock change as its derivative dA/dt.Specifically, if in a numerical implementation dt isassumed to equal one year, then dA/dt represents theyearly variations in the stock. Hence, assuming forthe moment that the demolition activity demA in oneyear is known, the construction activity newA for thatsame year is easily deduced from the balance equation(equation 1).
The demolition activity is a function of the previousconstruction activity and the expected lifetime of abuilding. Then, in order to calculate the flow of demo-lished floor area demA in a specific year, it is necessaryto know the flow of new area newA for all previousyears and the corresponding expected lifetime of build-ings. Literature on the buildings’ lifetime shows thatthis is a quite difficult issue to deal with and the avail-ability of data is limited. Buildings have a long lifetimeand so it is difficult to find data that go back in timeenough to observe the entire history of a buildingstock; often, analyses found in the literature arebased on a relatively small sample of buildings. Onesuch approach is found in the study of Bohne et al.(2006) on the Norwegian dwelling stock. To whatextent the observations made on a restricted group ofbuildings can be generalized to the entire stock isunclear. Lifetime distributions are often approximatedwith different functions, such as normal, log-normal,Weibull, Gompertz (e.g. Bohne et al., 2006; Johnstone,2001b; Muller, 2006). For reasons of simplicity, and inthe absence of better estimations, the authors decidedto adopt a normal distribution function. A normaldistribution function is completely defined by twoparameters: the mean, t, and the standard deviation,s. There is no agreement between different studies onwhat values should be used for t and s. Concerningthe Norwegian dwelling stock, Bohne et al. (2006)suggest an expected lifetime of 126 years. Bergsdal
et al. (2007b) use two scenarios with 75 and 125years, and also attempt an approach where the lifetimevaries between 150 and 95 years for buildings of theoldest and most recent constructions, respectively. Inthis paper the authors decided to use the values oft ¼ 75, 100 and 125, respectively, for the low,medium and high scenarios. Furthermore, in all scen-arios it is assumed that s ¼ 0.25t. The resulting life-time profiles are presented in Figure 3. By applyinglifetime scenarios with a wide range of values, it ispossible to cover a large span of demolition outcomes.It follows naturally that with a longer lifetime thedemolition activity will decrease, and the renovationactivity will increase due to its cyclic behaviour (seebelow). In general, the relatively low demolitionactivity experienced in the past cannot be expected tobe sustained in the future as the current buildingstock is relatively young (post-Second World War).A considerable increase in future demolition activitywill have major implications on material use, energyconsumption, waste generation, etc. Analysis of theeffects of such a development and the possiblemeasures that could be taken is left for future work.
The lifetime profile, L, is defined as:
1 – Normal cumulative distribution function (cdf)
The associated normal probability density function(pdf), which is simply the derivative of the normalcdf, gives the demolition profile. The demolitionprofile is here named D (Figure 4).
In order to understand how the demolition activity inany given time, t, depends from all previous construc-tions, as an intuitive approach the reader can imaginetranslating the demolition profile function, D, ofFigure 4 over the construction flow, newA, from theinitial time of simulation, t0, until the present time ofsimulation, t. To express this concept mathematicallyrequires introducing the concept of convolution. Con-volution is a mathematical operator between two func-tions, f and g, often denoted as f � g, that in a sense
Figure 3 Lifetime pro¢le: the probability of a general squaremetre to be found still standing in the stock as a function of timesince its construction
Towardsmodelling of construction, renovation and demolition activities
415
Dow
nloa
ded
by [
Uni
vers
itetb
iblio
teke
t I T
rond
heim
NT
NU
] at
18:
37 1
2 N
ovem
ber
2013
expresses the amount of overlap of one function as it isshifted over the other. The convolution of f and g overa finite period of time [t0, t] is formally written as:
f � g ;ðtt0
f (t0) � g(t � t0)dt0 (convolution)
Even though the reader might not be familiar with thisconcept, convolution is a standard mathematical oper-ation used in many fields, especially in signal proces-sing. It is out of the scope of this paper to give anexplanation of the mathematical meaning of such anoperation in exact terms; for such a purpose referenceis made to mathematics textbooks and manuals (e.g.Kreyszig, 2006). In addition, as a certain stockalready exists at the initial time of simulation t0 (inyear 1800), the demolition has also to account forthis initial stock A(t0). The composition and the ageof the initial stock are unknown, so it is difficult tomake motivated assumptions. Here it is simplyassumed that it is demolished at a constant rate for100 years; this is equivalent to state that the initialstock has an average lifetime somewhat shorter thannew constructions built in the starting year(Figure 3). This simplified approach is sufficient inthe case of Norway (and possibly most other countries)because the initial stock is very small compared withlater periods. Therefore, possible errors from thisassumption have a very small impact on the demolitionrate in later periods.
It is then possible to calculate the demolition flow,demA:
demA(t) ¼ D0(t)þD � newA
¼ D0(t)þ
ðt
t0
D(t0, t, s) � newA(t � t0)dt0 (3)
whereD0(t) is the demolition of the initial stock, and isequal to A(t0)/100 for the first 100 years of simulation(1800–1900); and zero otherwise. Equation (3)
expresses mathematically the fact that the demolitionin a specific year is given by the sum of all those con-structions from previous years that have now reachedtheir end of life, according to their demolition profile.Therefore, knowing all the values of constructionactivity newA until the generic year i allows one to cal-culate the demolition activity value demA for thecoming year iþ 1, which in turn allows one to calcu-late the construction activity newA in year iþ 1 fromthe balance equation, and so forth in an iterativeprocess.
Calculation of the renovation activity flow completesthe model. The renovation activity is calculated in par-allel to the other activities, so that its value does notaffect the other flows. Data availability on renovationactivity is also poor, and generally the issue is evenmore problematic than for demolition. Renovationactivity can be cyclic, and this feature will be addressedbelow. First of all, the term ‘renovation activity’ has tobe defined. In principle, each material, component,building’s subsystem or energy aspect can be con-sidered independently because they are characterizedby different life cycles, e.g. wood facades have a differ-ent lifetime than roofs, and windows have a differentlifetime than heating boilers. The model does allowconsidering more renovation activities in parallel bydefining several renovation flows, namely renA1,renA2, renA3 and so on. The renovation intervals foreach flow should be found empirically. Nevertheless,as the goal of this paper is to focus attention on themethodology, only one renovation flow is shown asan explanatory case.
In the model the renovation activity has to be re-presented by means of some probability function, thesame way as the demolition activity is treated. Thisfunction is named R and is again chosen to be anormal pdf. In order to show an explanatory casethat is at the same time relevant, the choice of tRvalue is based on the following argumentation. Thefocus on renovation is aimed to serve as a basis forfuture analysis on materials and energy demand of resi-dential buildings, and the relative opportunities toimprove buildings’ energy performance through reno-vation. From a study on the energy demand in the lifecycle of buildings (Sartori and Hestnes, 2007) thatreviewed a number of scientific articles for a total of60 cases from nine countries, it emerges that it islargely accepted as common practice to performenergy analysis over a period of 30–50 years. Thisbecause it is generally assumed that after such aperiod an average building is either demolished orundergoes major renovation works that will consider-ably alter its energy performance. A simplifiedapproach to energy demand analysis could consideran average time after which the overall energydemand itself is ‘renovated’. This is of course anapproximation, but would allow concentrating all the
Figure 4 Demolition pro¢le: the probability associated with thedemolition activity
Sartori et al.
416
Dow
nloa
ded
by [
Uni
vers
itetb
iblio
teke
t I T
rond
heim
NT
NU
] at
18:
37 1
2 N
ovem
ber
2013
possible variations in energy consumption in a singleoverall energy performance parameter that can varyover time according to a given renovation flow. There-fore, to the extent that the above approximation can beconsidered valid, the renovation flow presented herecan be regarded as representative of those major reno-vation works that allows altering considerably a build-ing’s energy performance. A value of tR ¼ 40 years ischosen as a central guess. Once again, due to uncer-tainty, instead of a single estimation, three scenariosare given for tR ¼ 30, 40 and 50 years, each assumingsR ¼ 0.25tR (Figure 5). It may be argued that withsmall mean values the probability function wouldneed to be asymmetric in order to avoid renovationactivity in the first years after construction. However,such details are left for further improvements infuture work, and the normal function is adopted here.
While demolition can take place once, renovation cantake place several times during the life cycle of a build-ing. Nevertheless, the probability of a general squaremetre in the stock to be renovated need to be weightedagainst the amount of building mass that is still stand-ing. Moreover, it needs to be weighted against theamount of building mass that is expected to remainstanding long enough to justify renovation works.In other words, the model should not simulate to reno-vate a building today and demolish it tomorrow. Toaccount for this fact, the authors assume that a reason-able period of expected lifetime after renovation is, atleast, equal to the renovation interval itself. This willgive a renovation profile that is damped over thecourse of time. The cyclic renovation profile, RC, isgiven by equation (4) as the result of periodical rep-etitions of the renovation function, R, weightedagainst the lifetime profile, L, shifted tR forward toprevent premature demolition of renovated buildings:
RC(t, t, tR, s, sR) ¼XNk¼1
R(t, tR � k, sR)
� L(t þ tR, t, s) (4)
where k is the renovation round; and N is themaximum number of renovations allowed. The resultis shown in Figure 6 for the medium scenario witht ¼ 100 years and tR ¼ 40 years.
Similarly to what happens for the demolition flow, therenovation flow is given by a convolution of the inputflow with the probability function that describes it, thecyclic renovation profile:1
renA ¼ RC � newA
¼
ðtt0
RC(t0, t, tR, s, sR) � newA(t � t0)dt0 (5)
It is worth noticing that while the area under the demo-lition profile curve is equal to 1 by definition (a buildingcan be demolished only once), the area under the reno-vation profile curve can be higher than 1. Its value canbe calculated and is here named ‘renovation number’,NR; it tells how many times the average square metrein the stock is renovated. For the medium scenarioNR ¼ 1.01, meaning that while some buildings arenever renovated others are renovated twice, and afew even more times, giving an equivalent figure of1.01 renovations per each square metre in the stock,from its construction till its demolition. Scenarioswith high/low renovation intervals or high/low life-time profiles will give renovation numbers that arehigher or lower than 1, depending on the case.
As a final remark, it is worth underlining that the defi-nition of lifetime profile is exogenous to the model; it isknown a priori and it is not affected by renovationactivity. It may be argued that in reality buildingsthat are renovated more often are also likely to livelonger. This sort of cause–effect relation is notincluded in the model, but the same relation can beobserved in the results. In fact, if a longer lifetime isassumed then the renovation activity will increasebecause more cycles of renovation find place withinthe given lifetime.
Figure 5 Renovation function: the probability associated with asingle round of renovation activity
Figure 6 Renovation pro¢le: the probability associated withweighted periodical rounds of renovation activity (left y-axis)compared with the lifetime pro¢le (right y-axis)
Towardsmodelling of construction, renovation and demolition activities
417
Dow
nloa
ded
by [
Uni
vers
itetb
iblio
teke
t I T
rond
heim
NT
NU
] at
18:
37 1
2 N
ovem
ber
2013
Results and discussionThis section presents the results obtained by runningthe model in different scenarios and shows the evol-ution of stock and flows from 1900 to 2100. Themedium scenario is presented first; here all inputs areset at their medium scenario values. Results on stockand flows are compared with available empiricaldata. Subsequently, a sensitivity analysis is performedby varying each input data at a time, from its low toits high scenario values, in order to study the effect ofthe estimates’ uncertainties. Two groups of graphsare presented: one for the input data series P, PD andAD; and the other for the input parameters L and R.Data from all the simulations are presented inTable 3. The main emphasis is on the renovationactivity, and for a more detailed discussion about thestock and flows of construction and demolition, seeBergsdal et al. (2007b).
Medium scenarioThe results for the medium scenario are shown inFigure 7. While population growth in the last centuryhas been nearly linear and is expected to continue tobe so at least until around year 2050, the two lifestyleindicators PD and AD are expected to level off in thefuture (Figure 2). Due to this levelling off the stockwill continue to increase, but at a lover pace thanwhat it has experienced in the past. Thisslowing phase has already started in the last twodecades. Hence, the steep slope that the constructionactivity, newA, has presented since roughly the end ofthe Second World War peaks in the mid-1980s andthen starts to decline until today and up to around2025.
Because of the long lifetime of dwellings, t ¼ 100 yearsin the medium scenario, the peak of construction willproduce a peak of demolition activity, demA, towardsthe end of the century; this peak will be less pro-nounced because of the variance, s, of the lifetime
profile, L. In turn, increased demolition activity willcall for increased construction activity in order to sub-stitute the demolished floor area. Construction activityalso has to meet an increasing demand, but as the stockseems to stabilize towards year 2100, the input andoutput flows will tend to have the same value. Alsothe renovation activity, renA, is expected to increaseas an effect of the initial construction peak, with afirst peak delayed of about 40 years, according to therenovation interval. In the long run, as all the othervariables seem to stabilize towards the end of thecentury, renovation activity will also converge toroughly the same value of construction and demolition.This is the combined effect of the stabilized stock andthe renovation number equal to 1. Nevertheless, it isworth noticing the time evolution for the comingdecades: construction activities are declining whilerenovation activities are increasing. The model showsthat around 2010 renovation is expected to overtakeconstruction and become the principal building’srelated activity for the remaining three decades, untilover 2040.
Comparisonwith empirical dataThe results of the medium scenario can be comparedagainst available data on the Norwegian dwellingstock and activities of construction, demolition andrenovation. The Norwegian bureau of statistics, Sta-tistics Norway, has regularly published censuses onpopulation and housing since 1900 with a ten-yearperiod (Statistics Norway, 2001a) that contain dataon the number of dwellings in stock based on a com-plete census. Unfortunately, data on the size of dwell-ings are only available in the last editions of theCensus for 1980, 1990 and 2001; these records areshown in Table 1 in comparison with the model values.
Data forP andPD are taken from the censuses (Bergsdalet al., 2007b), with the only difference being thatsmooth curves rather than the original data points areused (Figure 2a and b). The empirical data used toderive the AD curve are instead collected from surveysof housing conditions made on relatively smallsamples of dwellings (some thousands) and not onthe entire stock (Bergsdal et al., 2007b). These
Figure 7 Stock and £ows evolution for themedium scenario
Table 1 Total £oor area in the stock, census versusmodel
Year Censuses1,approximate total(millionsm2)
Model, approximatetotal(millionsm2)
Difference(%)
1980 135 148 9.61990 184 194 5.42001 238 239 0.4
Note: 1Statistics Norway (2001a).
Sartori et al.
418
Dow
nloa
ded
by [
Uni
vers
itetb
iblio
teke
t I T
rond
heim
NT
NU
] at
18:
37 1
2 N
ovem
ber
2013
considerations can explain the difference between thecensuses and model values for total floor area in thestock.
For construction activity, yearly records since 1946 areavailable from Statistics Norway (2007a, 2007b). Asshown in Figure 8, the model seems to be able to repro-duce quite faithfully the behaviour of the constructionactivity when looking at the long-term trends whileignoring short-term fluctuations; this is a positivefeature of the model. Nevertheless, data on floor areafrom the model look generally higher than thoserecorded in the statistics for both stock and newconstruction.
Three factors can justify the model’s overestimation ofnew floor area. First, as mentioned, the assumedaverage size of dwellings in the model AD (Figure 2c)is derived from surveys conducted on relatively smallsamples of dwellings and so the input AD may be anoverestimate itself. Second, the shape and magnitudeof the newA curve is affected by the assumed functionexpressing the lifetime of buildings. Differencesbetween the actual lifetime of buildings and theassumed lifetime profile, L, may then also be respon-sible for the observed overestimation. Third, statisticson floor area include only new constructions. In themodel, the flow newA is instead meant to account forall new floor area that comes into the stock in order
to satisfy a higher demand. Part of the additionalnew floor area may come from the extension of existingbuildings or simply by converting into living spacesparts of the building that were originally not countedas living area (e.g. basement or loft).
Concerning demolition and renovation activities,unfortunately no comprehensive data sets were foundfor comparison. The data collected from the literatureare summarized in Table 2 and compared with themodel’s results.
Regarding demolition, the results of the model appearto be in line with those reported by others. In hisreport, Myhre (2000) estimated that in the periodbetween 1983 and 1995 the number of dwellingsdemolished oscillated between 3000 and 5000 peryear, corresponding to 300 000–500 000 m2. Heused a value of 530 000 m2/year for projections until2030. He also reported that Rødseth et al. (1997) esti-mated dwellings departure to be 7000 units towards2015. Results from the model lie in the same range,but marking an increasing trend from 450 000 to1 200 000 m2/year in the period between 1998 and2030.
Regarding renovation, the results of the model appearto be an underestimate compared with other sources;this is remarkable, considering that renovation flowalready turns out to be the most significant one in themodel for the years ahead. Rønningen (2000) baseshis study on economical evaluations of cost persquare metre in new buildings and in renovation pro-jects, and concludes that in 1998 renovation activityin square metres corresponded to about 77% of newconstructions in the same year. He also emphasizesthat this covers only officially reported renovation pro-jects, and the numbers are therefore uncertain and thereal renovated surface area might be higher. The equiv-alent figure calculated in the model gives 52% for thesame year. It might be argued that the value reportedby Rønningen could be affected by yearly fluctuations.However, it is also possible to compare the cumulativerenovation activity with other sources. StatisticsNorway censuses usually do not report data on either
Table 2 Summary of data comparison for demolition and renovation activities
Source Period Demolition (m2/year) Renovation
Myhre (2000) 1998^2030 530 000 ^Model 1998^2030 450 000^1200 000 ^R�nningen (2000) 1998 ^ 77%of new constructions (m2/year)Model 1998 ^ 52%of new constructions (m2/year)Statistics Norway1 Total,1971^2001 ^ 50%of 2001stock (dwellings)Model Total,1971^2001 ^ 19%of 2001stock (m2)
Note: 1Personal communication with B.E. Br �aten,Statistics Norway, 2006.
Figure 8 Construction activity in the model and in statistics(Statistics Norway, 2007a, 2007b).
Towardsmodelling of construction, renovation and demolition activities
419
Dow
nloa
ded
by [
Uni
vers
itetb
iblio
teke
t I T
rond
heim
NT
NU
] at
18:
37 1
2 N
ovem
ber
2013
demolition or renovation; data on renovation werecollected for the first time only in the Census of2001. The respondents were asked to specifywhen the last round of renovation was performed,specifying that:
Extensive improvements and renovation aredefined as major work that has been carried outto raise the standard of the dwelling.
(Statistics Norway, 2001b)
The figures collected were not complete, so they werenot officially published in the Census 2001 report;nonetheless, it was possible to access that informationfrom Statistics Norway (B. E. Braten at StatisticsNorway, personal communication, 2006). From Stat-istics Norway it emerges that in the period between1971 and 2001 a total of at least 983 323 dwellingwas renovated, plus a number of 263 886 ‘don’tknow’ respondents. Considering that the Norwegiandwelling stock in 2001 consisted of a total of 1 961548 dwellings (as recorded in the same census), theresulting figure is quite outstanding: 50% of the build-ing stock has been renovated in the course of the past30 years. On the one hand, this might suggest thatthe pdf for renovation (Figure 5) should have itsmean at 30 years rather than 40 years, then causing ahigher renovation flow in the model. On the otherhand, caution has to be used in reading the resultsfrom the census. Indeed, it is not clear how the respon-dents might have interpreted the ‘major work that hasbeen carried out to raise the standard of the dwelling’.Nevertheless, it remains a fact that the model’s esti-mate, even though not directly comparable becauseexpressed in square metres rather than in number ofdwellings, is clearly lower: 19% of building stock reno-vated in the last 30 years.
Scenarioswith input data seriesThe driving force for the dwelling stock is the popu-lation’s demand for floor area of living. Strong nationaleconomic growth in the last 50 years is reflected inhigher living standards and subsequently a higherdemand for floor area. This growth is demonstratedin the general development in P, PD and AD for thelast 50 years, and the resulting increase in constructionactivity and stock size for the same period. However,the trends in AD and PD have shown signs of levellingoff in the last years, and the scenarios assume asimilar trend for the future; implying decreased futuregrowth levels for the dwelling stock compared withthe last 50 years.
Historical results should be equal for all scenario vari-ations. However, there are slight deviations in theregression curves for the input data series, althoughthey are too small to be identified in Figure 2. Asa result, minor deviations can also be found in the
result graphs shown in Figure 9. The effects onthe results are however negligible.
Population scenario results,PThe upper part of Figure 9 shows simulation results forhigh and low scenarios for population, P. Populationgrowth is a fundamental driver in the dwelling system,and the results demonstrate this fact. While the stockstabilizes towards 2100 in the low scenario, the stock con-tinues to grow rapidly in the high scenario. Relative to themedium scenario, the 2100 stock values will be 115%and 85% for the high and low scenarios, respectively.The same figures are 124% and 88% for construction.
Demolition activity is the result of previous stockdemand and construction activity, and due to thelong lifetime of dwellings, the effects on demolitionactivity are not very strong.
Regarding renovation activity, both scenarios predict thatthis will surpass construction activity within the next fewyears. In the low scenario itwill remain so for nearly half acentury. The first renovation peak is mainly a conse-quence of the high construction activity in the lastdecades, and the peak values are therefore almost thesame for the two scenarios. Population growth thereafterplays a more important role with respect to renovationactivity. The high scenario experiences continuedand strong growth for the last four decades, endingat 115% of the medium scenario, whereas the corres-ponding value is 81% in the low scenario.
Personsper dwelling scenario results,PD
Applying high and low scenarios for persons per dwell-ing, PD, gives similar main trends as for population.However, the development in construction activity inthe low scenario shows considerable differences fromthe high and medium scenarios. The decrease in con-struction activity in the next 15–20 years is muchless pronounced for the low scenario due to less rapidchanges in future PD values.
Regarding renovation, two features distinguish the lowscenario for PD from both the high scenario and thescenarios for the other input data series. First,whereas the timing of the first renovation peakremains the same, the low scenario does not experiencea subsequent decline. Renovation activity remains at astable level before increasing and then stabilizing againtowards the end of the century. Second, renovationnever exceeds construction activity.
Floor area per dwelling scenario results,AD
Assumptions about the average size of dwellings proveto be very important. Whereas the stock increasesstrongly in the high scenario, it is actually decreasing
Sartori et al.
420
Dow
nloa
ded
by [
Uni
vers
itetb
iblio
teke
t I T
rond
heim
NT
NU
] at
18:
37 1
2 N
ovem
ber
2013
towards 2100 in the low scenario, implying moredemolition than construction. Since demolition is theleast affected activity, the main influence is to befound in construction activity; ending at 120% and75% of the medium scenario for the high and thelow scenarios, respectively.
The renovation curve for the high AD scenarioresembles the one for the high P scenario. In bothcases there is a strong and continued growth in thedwelling stock, and the resulting increase in construc-tion activity from around 2020 is reflected in increasing
renovation activity one renovation interval later. Thedecrease in stock size in the low scenario takes placetoo late for its influence to be seen on renovationactivity; the effect will show after 2100.
Scenarioswith input parametersScenarios for the input parameters include varying theexpected average lifetime, t, of residential buildings andthe average renovation interval, tR. Simulation resultsare presented in Figure 10, with lifetime scenarios inthe upper part and renovation scenarios in the lowerpart.
Figure 9 Simulation results I: sensitivity analysis on input data series
Towardsmodelling of construction, renovation and demolition activities
421
Dow
nloa
ded
by [
Uni
vers
itetb
iblio
teke
t I T
rond
heim
NT
NU
] at
18:
37 1
2 N
ovem
ber
2013
Lifetime scenario results,LAlteration of the lifetime assumptions affects both thetiming and the magnitude of the construction anddemolition peaks, and of the renovation rounds.Stock size is however not affected as the population’sdemand for dwelling services (in m2) is independent ofbuilding lifetime in the model. The fluctuations aremore rapid and more pronounced in the low scenario.Demolition activity increases as buildings are takendown earlier in the low scenario, and as stock demandis unchanged construction activity will also increase inresponse to the higher demolition rate. The activitylevels are considerably different for a longer buildinglifetime, demonstrating the much slower turnover offloor area accompanied with a long lifetime. Table 3shows that the largest deviations are not found in2100, as for the input data series, but rather in 2050.It is the demolition activity that is affected the most,with 2050 values of 52% and 183% of the mediumscenario in the high and the low scenario, respectively.Whereas the next peaks in construction and demolitionoccur around 2060 in the low scenario, the peaks aredelayed until the next century in the high scenario. Ass is set to 0.25t, the activity distributions are alsomore dispersed in the high scenario.
Renovation activity levels are strongly affected by life-time scenarios. Renovation is by far the dominating
activity for the entire forecasting period in the highscenario, while even demolition will be considerablyhigher than renovation in the low scenario fromabout 2020. This simply reflects the fact that onlywhen buildings live for a long time renovation activityis consistent. Another way to look at it is with the reno-vation number,NR. WhereasNR is 1.01 in the mediumscenario, the same figures are 1.63 and 0.42 for thehigh and low scenarios, respectively. Relative to themedium scenario, the activity levels are 120% and62% for the high and low scenarios, respectively. In2050, the effect is even more pronounced with thesame values being 146% and 48%. The 40-year reno-vation intervals are clearly seen in both scenarios.
Renovation scenario results,RStock size and levels of construction and demolitionactivity are independent of the renovation intervalsand are therefore not affected in the renovation scen-arios, i.e. their graphs are the same as in the mediumscenario in Figure 7.
Variations in the renovation interval, tR, are, asexpected, of considerable importance for both thetiming and magnitude of renovation activity. The lowscenario with tR ¼ 30 years has a considerably higherrenovation activity throughout the entire time
Figure 10 Simulation results II: sensitivity analysis on input parameters
Sartori et al.
422
Dow
nloa
ded
by [
Uni
vers
itetb
iblio
teke
t I T
rond
heim
NT
NU
] at
18:
37 1
2 N
ovem
ber
2013
horizon of the simulations, and especially for the futureprojections. A lower tR value in the low scenarioimplies a more rapid replacement of materials andbuilding components. The peaks in renovation activitytherefore appear sooner than in the high scenario, andas lifetime is unchanged, the renovation number, NR,and total renovation activity increase. WhereasNR ¼ 1.01 in the medium scenario, the same figuresare 0.54 and 1.84 for the high and low scenarios,respectively. Relative to the medium scenario’s valuerenovation activity level in 2100 is 56% for the highscenario and 177% for the low scenario, respectively.In the high scenario renovation levels are so low thatfrom around 2050 demolition becomes higher.
General considerationsRenovation activity has become increasingly importantand follows the growth of stock and constructionactivity. For all input data series scenarios, exceptlow PD, renovation will take over as the dominantactivity in the next decades. Varying t and tR instead
has a dual effect. Short lifetime or long renovationintervals result in low renovation activity, while theopposite is the case for long lifetime or short renova-tion intervals that generate high renovation activity.However, regardless of the scenario, renovationactivity is expected to increase in absolute figures inthe nearest decades because of the peak in constructionactivity in the 1980s.
With the exception of the low scenario for AD, stock isincreasing for the entire projection period. This alsoimplies that demolition activity is always lower thanconstruction activity. However, as the input dataseries level off towards 2100, stocks will be growingless; at the same time demolition increases and con-struction decreases or flattens out, so that the twocurves will show signs of converging.
Table 3 presents a comparison between magnitudes(%) of stocks and flows for the different scenarios rela-tive to the medium scenario in 2050 and 2100. Theupper part shows the numerical values for themedium scenario in the same years.
Table 3 Scenario results relative to themedium scenario
Stocks/£ows Medium
2050 2100
Floor area A (m2) 3.45Eþ 08 3.96Eþ 08newA (m
2/year) 4.05Eþ 06 4.26Eþ 06demA (m
2/year) 2.19Eþ 06 3.78Eþ 06renA (m
2/year) 3.50Eþ 06 4.21Eþ 06
Input variation Low High
2050 2100 2050 2100
Population,P A (%) 92 85 106 115newA (%) 73 88 114 124demA (%) 100 95 100 105renA (%) 96 81 107 115
Persons/dwelling,PD A (%) 116 118 93 87newA (%) 121 115 82 87demA (%) 101 113 100 95renA (%) 121 117 94 85
m2/dwelling, AD A (%) 95 85 107 115newA (%) 84 75 117 120demA (%) 100 97 100 106renA (%) 97 84 108 116
Lifetime,L(t, s) A (%) 100 100 100 100newA (%) 145 118 74 83demA (%) 183 120 52 81renA (%) 48 62 146 120
Renovation,R(tR, sR) A (%) 100 100 100 100newA (%) 100 100 100 100demA (%) 100 100 100 100renA (%) 185 177 57 56
Towardsmodelling of construction, renovation and demolition activities
423
Dow
nloa
ded
by [
Uni
vers
itetb
iblio
teke
t I T
rond
heim
NT
NU
] at
18:
37 1
2 N
ovem
ber
2013
Scenarios for the input data series show that deviationsfrom the medium scenario values generally becomelarger with time, both for the high and the low scen-arios, although with a few exceptions. The delayrelated to demolition is clearly seen, with changesbeing close to zero in 2050, but increasing towards2100. Scenario results for the input data series give adeviation range of 85–118% for stock size and a cor-responding range of 73–124% for constructionactivity. The same figures are 95–113% and 81–121% for demolition and renovation, respectively.
The effects of changing the input parameters t and tRare more dramatic. Lifetime scenarios have a deviationrange of 74–145% for construction and 52–183% fordemolition. The same range is 48–146% for renova-tion. As opposed to the data input scenarios, the life-time scenarios have their largest deviations from themedium scenario in 2050. Changing the renovationintervals directly affects the renovation activity themost. The deviation range is 56–185% for the renova-tion scenarios. Stock size, construction and demolitionactivity are not affected in these scenarios.
ConclusionsForecasting of construction, demolition and renovationactivities in the built environment has significant impli-cations for policies on materials demand, energy conser-vation and climate change. The built environment is adynamic system where past activity levels strongly influ-ence future development. This dynamic is not suffi-ciently understood, and this paper presents anexploration of the methodology for assessing bothstock and flows related to residential floor area in acoherent way. The stock is assessed by social indicatorssuch as population, the average number of persons perdwelling and the average size of dwellings. The flowsare assessed based on assumptions on parameters suchas the lifetime of dwellings and renovation frequencies.
Modelling results are compared with other sourceswhen available. The comparison shows that construc-tion activity might be somewhat overestimated.However, construction activity appears to be quitewell represented by the model with respect to long-term behaviour. Renovation and demolition activityare harder to asses because less information is avail-able. The generally poor knowledge about these acti-vities is exactly one of the main motives for using adynamic model.
The dynamic behaviour of the model is clearly seen fromthe modelling results. The expected stabilization of life-style indicators will cause the stock of floor area to growslower than in the past, even though population growthis expected to continue nearly linearly until about 2050.As a consequence, the construction activity is expectedto slow down in the coming decades. Nevertheless, the
high construction activity recorded in the post-warperiod is expected to bring a substantial increase indemolition activity in the second half of the currentcentury. In turn, this will cause an increase in construc-tion activity in order to substitute the departed units.Construction activity then is expected to increaseagain and remain at higher values in the second half ofthe century, despite the expected slowing down of popu-lation and stock growth.
The dynamic effect is also visible in the results for therenovation flow with renovation peaks delayed fromthe construction peaks according to the given renova-tion intervals. With only one exception, all the scen-arios related to the input data sensitivity analysispredict that renovation will overtake construction asthe dominating C&D activity within the next fewyears. However, construction activity will dominateagain in the second part of the century. Therefore,the dominance of the renovation activity may be a tem-poral phenomenon because construction activity isexpected to re-raise as the post-Second World Warbuilding stock is due to be replaced. Results from theparameters sensitivity analysis are more dramatic andless uniform.
The suggested approach for renovation modelling isexemplified for major renovations of dwellings.However, the presented method allows for manydifferent types of renovation to be modelled. Reno-vation related to different materials, building com-ponents or energy uses can all be modelled separatelyby applying different renovation intervals.
Dynamic modelling of C&D activities supply valuableinformation about the possible future developments inthe residential sector and demonstrates the importanceof considering history when making assumptions andforecasts about the future. The method also allowsone to investigate how variations in different inputdata and parameters will affect the model forecast ofstock and flows of residential floor area. The requiredinput data are relatively easy to obtain from censusesor surveys on population and housing conditions. Atthe same time parameters as buildings’ lifetime andrenovation intervals are difficult to estimate. Thisdrawback is overcome in this method by defining pro-bability functions and analysing different scenarios,partially to compensate for the uncertainty inherentin the definitions of those probability functions.These features are believed to make the method pre-sented here suitable for application also to othernational residential stocks than the Norwegian one.However, data availability will vary and have to beconsidered in each individual case.
For future development of this work, the non-residential part of the building stock could also bemodelled dynamically following a similar approach.
Sartori et al.
424
Dow
nloa
ded
by [
Uni
vers
itetb
iblio
teke
t I T
rond
heim
NT
NU
] at
18:
37 1
2 N
ovem
ber
2013
The lifetime of these buildings is generally lower andthe renovation intervals are more frequent, with theimplications this has for the material and energy useand waste generation. However, modelling the non-residential building stock would require careful con-sideration of relevant parameters to represent thedrivers in the demand for non-residential floor area.Furthermore, the non-residential building stock is com-prised of a variety of building types that are less hom-ogenous than the residential stock, and wouldprobably need to be disaggregated further. Thedrivers and the parameters might be very differentthan those included in the analysis of the residentialbuilding stock.
AcknowledgementsThe authors would like to thank Dr Glen Peters at theIndustrial Ecology Programme, NTNU, for valuablehelp and discussions about the algorithms, data struc-turing and programming. Furthermore, they thankProfessor Anne Grete Hestnes at the Department ofArchitectural Design, History and Technology,NTNU, and Dr Bjørn J. Wachenfeldt at the Depart-ment of Building and Infrastructure, SINTEF, for valu-able discussions and feedback during this work.
ReferencesBergsdal, H., Bohne, R.A. and Brattebø, H. (2007a) Projection of
waste amounts from the AEC-industry in Norway. Journalof Industrial Ecology, 11(3), 27–39.
Bergsdal, H., Brattebø, H., Bohne, R.A. andMuller, D.B. (2007b)Dynamic material flow analysis for Norway’s dwellingstock. Building Research & Information, 35(5), 557–570.
Bohne, R.A., Brattebø, H., Bergsdal, H. and Hovde, P.J. (2006)Estimation of the service life of residential buildings, andbuilding components, in Norway, in Proceedings: The CitySurface of Tomorrow, 8–9 June 2006, Vienna, Austria.
Bradley, P.E. and Kohler, N. (2007)Methodology for the survivalanalysis of urban building stocks. Building Research &Information, 35(5), 529–542.
Caccavelli, D. and Genre, J.L. (2000) Diagnosis of the degra-dation state of building and cost evaluation of inducedrefurbishment works. Energy and Buildings, 31(2),159–165.
Itard, L. and Klunder, G. (2007) Comparing environmentalimpacts of renovated housing stock with new construction.Building Research & Information, 35(3), 252–267.
Johansson, P., Nylander, A. and Johnsson, F. (2006) Electricitydependency and CO2 emissions from heating in theSwedish building sector – current trends in conflict with gov-ernmental policy? Energy Policy, 34(17), 3049–3064.
Johansson, P., Nylander, A. and Johnsson, F. (2007) Primaryenergy use for heating in the Swedish building sector –current trends and proposed target. Energy Policy, 35(2),1386–1404.
Johnstone, I.M. (2001a) Energy and mass flows of housing: amodel and example. Building and Environment, 36(1),27–41.
Johnstone, I.M. (2001b) Energy and mass flows of housing: esti-mating mortality. Building and Environment, 36(1), 43–51.
Kohler, N. and Hassler, U. (2002) The building stock as aresearch object. Building Research & Information, 30(4),226–236.
Kohler, N. and Yang, W. (2007) Long-term management ofbuilding stocks. Building Research & Information, 35(4),351–362.
Kreyszig, E. (2006) Advanced Engineering Mathematics, 9th edn,Wiley, Hoboken, NJ.
Kytzia, S. (2003) Material flow analysis as a tool forsustainable management of the built environment, in M.Koll-Schretzenmayr, M. Keiner and G. Nussbaumer (eds):The Real and the Virtual World of Spatial Planning,Springer, Berlin, pp. 281–295.
Martinaitis, V., Kazakevicius, E. and Vitkauskas, A. (2007) Atwo-factor method for appraising building renovation andenergy efficiency improvement projects. Energy Policy,35(1), 192–201.
Muller, D., Bader, H.P. and Baccin, P. (2004) Long-term coordi-nation of timber production and consumption using adynamic material and energy flow analysis. Journal of Indus-trial Ecology, 8(3), 65–87.
Muller, D.B. (2006) Stock dynamics for forecasting materialflows – case study for housing in the Netherlands. EcologicalEconomics, 59(1), 142–156.
Myhre, L. (1995) Some Environmental and Economic Aspects ofEnergy Saving Measures in Houses: An Estimation Modelfor Total Energy Consumption and Emissions to Air fromthe Norwegian Dwelling Stock, and a Life Cycle AssessmentMethod for Energy Saving Measures in Houses, NorwegianUniversity of Science & Technology, Trondheim.
Myhre, L. (2000) Towards Sustainability in the ResidentialSector. Note No. 41. Byggforsk – Norwegian BuildingResearch Institute (NBI), Oslo.
Rødseth, A., Barlindhaug, R. and Østervold, J. (1997) A Simu-lation Model of the Norwegian Housing Market. Draft,Department of Economics, University of Oslo, Oslo.
Rønningen, O. (2000) Bygg- og anleggsavfall Avfall fra nybyg-ging, rehabilitering og riving. Resultater og metoder, Sta-tistics Norway, Oslo.
Sartori, I. and Hestnes, A.G. (2007) Energy use in the life cycle ofconventional and low-energy buildings: a review article.Energy and Buildings, 39(3), 249–257.
Statistics Norway (2001a) Series of Population and HousingCensus (folke- og boligtellingen): 1900, 1910, 1920, 1930,1946, 1950, 1960, 1970, 1980, 1990, 2001, StatisticsNorway, Oslo.
Statistics Norway (2001b) The Population and Housing CensusHandbook 2001, Department of Social Statistics, StatisticsNorway, Oslo.
Statistics Norway (2007a) Statistics on construction, inHistoricalStatistics (available at: http://www.ssb.no/histstat/tabeller/17-17-4t.txt).
Statistics Norway (2007b) Statistics on construction, in StatisticalDatabase (available at: http://www.ssb.no/english/subjects/10/09/).
Endnote1The possible renovation of the initial stock is omitted intention-ally. Considering it would have changed the results only for thefirst century of simulation (1800–1900), as the initial stock isextinguished in such time. Changes in this period are of no inter-est for forecasting purposes and anyhow have no effect on thedevelopment of flows because in the model construction anddemolition flows are independent from the renovation flow,which is also independent from its past values.
Towardsmodelling of construction, renovation and demolition activities
425
Dow
nloa
ded
by [
Uni
vers
itetb
iblio
teke
t I T
rond
heim
NT
NU
] at
18:
37 1
2 N
ovem
ber
2013