5531130e0cf2f2a588ac2dcf.pdf

See discussions, stats, and author profiles for this publication at:http://www.researchgate.net/publication/222430105Waste characterization as an element ofwaste management planning: Lessonslearned from a study in Siem Reap,CambodiaARTICLEinRESOURCES CONSERVATION AND RECYCLING DECEMBER 2006Impact Factor: 2.69 DOI: 10.1016/j.resconrec.2006.03.006CITATIONS24DOWNLOADS67VIEWS1453 AUTHORS, INCLUDING:Kate ParizeauUniversity of Guelph9 PUBLICATIONS 72 CITATIONS SEE PROFILEAvailable from: Kate ParizeauRetrieved on: 08 August 2015Waste characterization as an element of wastemanagement planning: Lessons learned froma study in Siem Reap, CambodiaKate Parizeaua,, Virginia Maclarena, Lay ChanthybaUniversity of Toronto, Department of Geography and Program in Planning,100 St. George Street, Toronto, Ont., Canada M5S 3G3bDepartment of Environmental Science, Royal University of Phnom Penh,Russian Confederation Boulevard, Tuol Kork, Phnom Penh, CambodiaReceived 20 September 2005; received in revised form 1 February 2006; accepted 3 March 2006Available online 18 April 2006AbstractCommunity-based waste management (CBWM) is an alternative waste management strategy forcommunities where municipal governments are not providing waste collection services. In order toassess the feasibility of introducing CBWMto an unserviced community in SiemReap, Cambodia, weundertook a waste characterization study and household survey in the summer of 2004. In the wastecharacterization study, we found that waste generation per capita was low (0.34 kg per capita per day,on average) compared to communities in other developing countries. We did not nd a statisticallysignicant relationship between household waste production and either income or expenditures. Weobservedthat thewastestreaminthestudyareawasmostlyorganicinnature(66%byweight)and contained few recyclable materials (5% by weight). Our results illustrate the importance of awaste characterization study for assessing how many collection vehicles will be needed for a CBWMprogram, whether composting is a feasible option, whether recovery of recyclables will be a signicantincomesourcefortheprogram,andwhethersocialprogrammingisneededtochangehouseholdawareness and waste behaviours. We found that the household survey results on household attitudesand membership were a valuable complement to the waste characterization study, as they provideduswithinformationabouthouseholdsize(andthereforeallowedustocalculatepercapitawastegeneration), the local residents willingness to separate waste streams at the source, and residentsCorresponding author. Tel.: +1 647 828 0468; fax: +1 416 946 3886.E-mail address: [email protected] (K. Parizeau).0921-3449/$ see front matter 2006 Elsevier B.V. All rights reserved.doi:10.1016/j.resconrec.2006.03.006willingness to participate in and pay for CBWMservices. We conclude that the waste characterizationstudy and the household survey together are important tools for planning a CBWM program. 2006 Elsevier B.V. All rights reserved.Keywords: Household waste characterization; Waste composition; Waste generation; Hoarding; Community-basedwaste management; Cambodia1. IntroductionWastecollectioninthedevelopingworldisanissueofgrowingconcern, especiallysince municipal authorities in many areas are either unable or unwilling to provide wastecollection services to all residents in their jurisdiction. On average, up to 50% of residentslack collection services in urban areas of low and middle income countries (Klundert andAnsch utz, 2001). SiemReap, Cambodia (see Fig. 1), where we conducted our study, is quitetypical in this regard, in that the town collects only 50% of the total waste generated (SiemReap Department of Environment, 2003).Fig. 1. Map of Cambodia; town of SiemReap highlighted (adapted fromUnited Nations Development Programme,2004).When municipal governments are unable tocollect all of the waste that is being generated,alternative waste management solutions may be appropriate for the unserviced areas of thetown.Onesuchalternativeiscommunity-basedwastemanagement(CBWM),asystemthat relies on community members to administer and participate in waste-related issues,including the collection, transportation, and diversion of waste. Generally, a CBWMsystemcollects household wastes fromindividual residences (primary collection) and deposits themat a central location for municipal pick-up (secondary collection). The collection systemofteninvolves the employment of hiredwaste collectors, or mayentail householders bringingtheir trashtoa central location(Khulna CityCorporationandSwiss Agencyfor Developmentand Co-operation, 2000).WithfundingassistancefromtheCanadianInternational Development Agency, theauthors collaborated with local and provincial government ofcials in Siem Reap to assessthe feasibility of introducing CBWM to an unserviced area of the town. The selected studyarea consists of approximately 1000 households located along both sides of the Siem ReapRiver to the south of the central part of town. This is a linear study area, and it containsparts of two commune administrative districts. We investigated local attitudes and wastemanagementbehavioursinthestudyareabymeansofahouseholdsurveyandawastecharacterization study. The purpose of the waste characterization study was three-fold: toestimatethequantityofwasterequiringcollection;tobetterunderstandthevariationinwaste production rates within the study area; and to assess the feasibility of including com-posting and recycling as a part of the CBWMprogram. This paper describes the results of thewaste characterization study, selected results from the survey and several lessons learned inconducting a waste characterization study for the purposes of assessing CBWM feasibility.1.1. Waste management issues in Siem ReapSiemReap(hometo85,000residents)isthegatewaytothearchaeologicalruinsofAngkor Wat, which is a UNESCO World Heritage Site attracting one-third of Cambodiastouristdollars(DMonte,2005).TheaestheticappearanceofSiemReapisclearlyveryimportant for its image as a tourism destination. That image is not helped by unsightly litteroating down the Siem Reap River, which runs through the middle of town, or scatteredpiles of waste in areas that have no collection service. In an attempt to reduce the amountof oating litter, the local government has constructed a barrier across the river upstreamof Siem Reap (see Fig. 2), but this barrier only mitigates the visual surface pollution, notthe pollution that is contaminating the river as a source of drinking water, a habitat for shand a recreational space. A private company hired by the local government uses boats tocollect any oating litter downstream of the barrier.Waste collection services are not provided to the entire town of Siem Reap for a numberof reasons. First, collection services have been contracted to a private waste hauler and someparts of the town are outside of the waste haulers contracted service area. For residentswho are outside the service area, there are substantial costs associated with either obtainingacontractwiththehaulerforthetransportationofwastetothelocaldumpsite,orwithobtaining direct access to the dumpsite. Second, outside of the central town, the poor qualityof local infrastructure limits truck access to houses. Most roads in the peri-urban area aredirt roads and become almost impassable in the rainy season. Finally, many of the residentsFig. 2. Barrier to waste across the Siem Reap River upstream of the town of Siem Reap.living directly adjacent to the river do not have legal tenure on their land, and governmenteviction of these residents presents an occasional threat. It is possible that the municipalityis withholding waste services to deny legitimacy to these settlers. Illegal settlements oftenlack a number of municipal services, including waste collection (Wangombe, 1995).Thelackofwastecollectionservicesinthestudyareahasbecomeamorepressingproblem in recent years because of the changing waste stream. A representative from onelocal authority commented that people used to bury their waste in their gardens as compost,but are unable to do so any more because of the increased plastic content (Om Caat, 2004).Plastic goods and packing are readily available in Siem Reap; we therefore expected to nda high proportion of plastics in the waste streamwhen conducting our waste characterizationstudy, along with the high proportion of organics traditionally found in waste streams in thedeveloping world (as discussed in the results section below).2. Methodology2.1. The household surveyInordertocollectinformationaboutresidentssocio-economiccharacteristics, theirattitudestowardswaste,wastemanagementbehaviours(disposalandwasteseparation),and willingness to pay for collection services, we designed and administered a survey to300 households in the study area in the summer of 2004. The questionnaire contained atotal of 21 questions related to the feasibility of introducing a CBWM program, but onlythe data that were useful for the waste characterization study will be reported on here.Because we suspected that waste behaviours and incomes might vary by location relativeto the river, the household sample was drawn from four strata: households located on theeast side of the river, households located away fromthe river along the east road, householdsFig. 3. Schematic of household locations in the study area.along the west side of the river and households along the west road (see Fig. 3 for a schematicof the study area, and Figs. 4 and 5 for photos). Residents living along the river are illegalsquatters and dwell in substantially lower quality housing than those living on the road side.After a randomstart at each location, every third house within the stratumwas approachedfor inclusion in the sample. If nobody was home at the selected household, the next house-hold was substituted for the missing household. The survey took place over a period of 4 daysduring daylight hours and was directed to the wife or mother of the household wherever pos-sible, since women rather than men usually have responsibility for waste management tasksin Cambodia. This division of responsibilities was veried in the household survey, whichfound that wives were responsible for waste management in 43% of households, femalechildren in 21% of households, and other female residents in 8% of households (n =291,Fig. 4. Houses backing onto the river on the west (left) and east (right) sides.Fig. 5. Houses backing onto the west river (left), and on the west road (right).multiple responses to this question were allowed). We administered a short follow-up sur-vey to the same households approximately 1 year after the rst survey in order to assessthe impact on attitudes towards waste and waste behaviour of an environmental educationprogram that staff from the Royal University of Phnom Penh delivered to local authorities,residents, teachers and monks in the intervening period. The education program providedinformation about howa community-based waste management programwould work, pollu-tion sources in the community, including solid waste, and their impact on the environment.Since the follow-up survey asked far fewer questions than the original survey and focusedmostly on the education intervention, the household survey results discussed below refer tothe rst survey, unless otherwise noted.2.2. The waste characterization studyWe conducted the waste characterization study about 1 month after the household survey,and selected participants from a stratied random subset of the interviewed households (50households were selected, but one did not participate; n =49). The strata in the sample weremonthly household income (nine strata with income ranges of $100 USD each) and houselocation on either the west road, west river, east river, or east road, with roughly similarproportions take from each stratum. The residents of the selected households were asked tocollect their waste (that is, any materials they would normally burn, bury, or throwin the riveror other public spaces) each day for a week in the summer of 2004. Eight plastic collectionbags were provided to each householdone for each day of the study, and one extra bag incase it was required. We recognized that we might be capturing both residential waste andthe commercial waste produced by home-businesses, but this was considered acceptablebecause we were attempting to assess the required capacity of a potential collection system,not the percentage of residential versus commercial waste. However, while this collectionmethod is more likely to produce an accurate estimate of the total amount of waste availablefor collection in the study area, it can also confound analyses of relationships between house-hold income and waste generation. Further difculties in these analyses are discussed below.Fig. 6. Hand scale used to weigh waste samples.We chose the extended observation period of 1 week (as opposed to 1 day) to minimizewastehoardingbehavioursthatcanskewdatacollection. Additionally, thisobservationperiod allowed us to take account of the daily uctuations in waste generation that may occurwithin a week (Shimura et al., 2001). We weighed the collected waste at each householdusing hand scales (see Fig. 6). We then brought it to a sorting area where it was separatedand weighed again, all on the same day. The sorting area was covered with a tarp to preventthe waste from drying out in the sun, and therefore changing the proportional weights ofthe high moisture organic components of the waste stream.In deciding what categories to use in sorting the waste, we followed a potential usecategorization(see,forexample,Bernache-P erezetal.,2001;Fehretal.,2000;Ojeda-Benitez et al., 2003) rather than the traditional material-based categorization. Since we wereinterested in the feasibility of source separation for composting and recycling, we sortedorganics into high nitrogen organics (such as fruit peels and other kitchen wastes) and highcarbon organics (such as dry leaves). Wood (except for wood shavings) and coconut becamea separate category because they are not easily composted. Of the potentially recyclablematerials, the plastics category had both the greatest diversity and the greatest quantity ofmaterial. Plastic items collected from the study area included grocery bags, netting, tubs,broken toys, bottles, and more. Because of this variety, plastic items were sorted into thosethat were routinely purchased by the local recycling depot and itinerant buyers (such asdrinkingwaterbottles),andthosethatwerenot(suchasplasticbags).Othercategoriesofpotentiallyrecyclablematerials(suchasmetalsandpaper)didnotcontainthesamediversity, and were not present in sufcient quantities to warrant further separation.2.3. Waste hoardingHoarding, in this instance, refers to the practice of saving waste for collection by thestudy team. For example, if residents were informed on Friday that waste collection wasbeginning on Monday, they may have saved their waste over the weekend to present it tothe study team on Monday. Another problem with the same effect as hoarding can occurif residents include neighbours wastes with their own, especially at the last opportunityforcollection. Toexcludeinstancesofwastehoardingontherstandlastdaysofthestudy, the mean daily weight of waste was adjusted by excluding those instances wherethe weight was more than two standard deviations away from the overall mean. Overall,weexcludedvecasesofrst-dayhoardingand2daysoflast-dayhoardingfromtheresults.The occurrence of rst-day hoarding is supported by the rst-day unadjusted waste totalof 124.2 kg (versus the mean of 92.5 kg for the other 6 days of study). The ve outliersremoved because of suspicion of hoarding behaviours accounted for 43.3 kg of the rst-daywaste total. This hoarding could be due to residents saving their waste from the previousdays (as some were approached to participate several days before the study actually began),or it could be due to residents picking up excess waste lying around their house that theyusually leave there. In support of this latter statement, we observed that much of the plasticsand paper collected on the rst day was coated with dirt. Additionally, when we weighedthe rst days collection, we found that there was a disproportionately high amount of dirt:25.6% by weight versus the study average of 14.0%.Last-day hoarding may have occurred because some residents felt that the end of thewaste study was their last chance to have waste collected at their door. However, manypeople did not realize that the study had ended when it did, possibly because eight plasticbags were handed out in case residents needed an extra bag over the 7-day study period. Onthe day after the study ended, we observed that many of the study households had put outbags of waste to be collected. This lack of awareness of the studys end may have helpedto reduce last-day hoarding.Thosesamplepoints morethantwostandarddeviations awayfromthemeanthatoccurredduringtheobservationperiod, but not ontherst orlast day, wereassumedto be uctuations that could be expected to occur regularly and were not excluded fromthemeanweightcalculation.Insomecases,residentcommentsonthemorningcollec-tionroutesubstantiatedthattheseuctuationswerenotduetohoarding, buttonormalvariationsinwastegeneration. For example, aresident onthewest roadwithalargeamount of waste mentioned that she had hosted a gathering the previous night. An alter-nativepossibleexplanationforunusuallyhighwastequantitiesduringtheweekisthatneighboursnotincludedamongthesampledhouseholdsobservedthecollectionactivi-ties and decided to add their waste to that of the sampled households. We had no way ofdeterminingwhetherthiswashappening,butsuspectthat,ifitdid,itwasnotaseriousproblem because the collected waste samples (excluding those from the rst and last days)exceeded two standard deviations fromthe mean only four times (representing only 1%of allsamples).2.4. Participation in the studyOne household on the east river refused to participate in the study fromthe rst day of col-lection. The head of this household repeatedly stated that the family did not have any waste tobe collected since it had no waste. Our research assistants spoke with these householders,andreportedthat theywere not amenable tothe idea of a CBWMsystem. This householdwasnot included in the analysis, and so the effective sample size was n =49. Another householdon the west road only participated on the rst day, and cited illness as the reason for non-participation. The value for its 1 day of participation was included in the study, but no valueswere included for the days of non-participation. Fifteen other households also had 1 or 2 daysof non-participation, either because they forgot, because dogs ate their trash, or because theyhad no waste. Values for these days were not included in the analysis. Of these 15 house-holds, hoarding behaviour was observed in only one case on the day after an incident of non-participation.UnlikeBolaaneandAli(2004),wedidnotndthatparticipantswantedtobecom-pensated for setting out their waste. Bolaane and Ali had asked the participants in theirstudy to separate their refuse into wet and dry wastes, whereas we did not require our studyparticipants to sort their wastes, and so little extra effort was required on their part. Wespeculate that people were happy to participate in our study without compensation becausemany in the study area felt that waste was a problem in the community (75% accordingtothesurveyresults), whileothersprobablysawthestudyasawaytocleanuptheirproperties.2.5. Possible sources of errorSince we used hand scales (see Fig. 6 above) to weigh the wastes collected from eachhousehold and to weigh the components of the sorted waste, measurement error is a factorin this research. For example, the total weight of the sorted component parts of the col-lected waste was compared to the initial sums of the weights of waste collected from eachhousehold. It was found that there was a slight difference in these totals each day, rangingbetween a 0.2% net gain and a 2.7% net loss in weight. On average, we observed a 0.8%loss in weight between the initial collection weights and the separated component weightseach day. This compares with a 6.6% loss in weight observed by Chung and Poon (2001)in the waste characterization process in Guangzhou, China. This source of error could alsobe due to dehydration of the samples during the sorting process, which would explain theoverall net loss observed on most days.ASolid Waste Management ProgramOfcer for the Community Sanitation andRecyclingOrganizationinPhnomPenhcommentedthat hisorganizationhasobservedlesswaste(inweight andinvolume) duringthedryseasoninCambodia(MayJunetoOctoberNovember), whenour studywasconducted, thaninthewet season. Thisobservation suggests that a comprehensive waste characterization study would need to beconductedovermultipleseasons(BoSokhan,2004).Unfortunately,wehadneitherthetimenortheresourcestosampleduringthewetseason.Mohee(2002)andBuenrostroetal.(2001)haveobservedseasonalchangesinthecity-widewastegenerationratesinMauritiusandMexico, respectively, ashaveChungandPooninChina(2001). Otherauthorshaveconductedstudiesatmultipletimesthroughouttheyeartocontrolforthistypeofuctuation(Shimuraetal.,2001).BasedontheirresultsandexpertinputfromCambodian colleagues, we would expect a sample in the wet season to nd larger quantitiesof waste, higher moisture content (Chung and Poon, 2001), and greater amounts of organicmaterials.3. Results3.1. An overview of the demographics of the study areaAccordingtothe householdsurvey, the average familysize inthe studyarea is6.7persons. Theaveragenumber of childrenunder theageof sixis 0.8per house-hold; theaveragenumber of childrenfromagesixto17is1.9. Maleshead76%ofhouseholds, andfemales head24%. Most of the households headedbywomenarethosewherethehouseholdheadis relativelyolder (56%of femaleheads of house-holdsareover 50yearsold, comparedto30%of maleheadsof households), imply-ingthatthesewomenmaybewidows. Theaverageageofthehouseholdheadis45.7years.The most common occupations in the study area include seller, service provider, gov-ernmentstaff,farmer,andanimalraiser.Wefoundthattheaveragemonthlyhouseholdincomeinthestudyareawas$434USD,andaveragemonthlyexpenditureswere$224USD. However, the validity of the former amount is questionable, since this income g-ure is very high for this region. The unreliability of the income data in this case may beduetoareluctanceofrespondentstoanswersurveyquestions(inthecaseofincome),andtoprovideaccuratedata(withrespecttobothincomeandexpenditures).Addition-ally, we discovered that some respondents who run home businesses were confusing theirindividual income with the gross revenue of their business. To give context to the incomevalues reported in the survey, Cambodias gross national income (GNI) per capita in 2003was$300USD,or$25USDpermonth(WorldBank,2005).Withanaverage6.5peo-pleperhousehold,theaveragemonthlypercapitaincomereportedinthestudyareais$68.77 USD, implying that the values reported in the survey are high. The median monthlyhousehold income in the survey area was found to be $225 USD; this value is much lowerthanthemean, supportingtheconclusionthat someofthereportedincomedatawereinated. The implications of these inated income values for our analyses are discussedbelow.3.2. Waste generationThe mean daily weight of the waste collected from all 49 houses in the waste characteri-zation study was 97.0 kg (this was calculated by averaging the daily totals of waste collectedover the 7 day observation period). The mean daily volume was 0.6 m3(similarly calculatedby averaging the daily volume of waste) and the mean waste density was 156 kg/m3(cal-culated by dividing the weight of the waste by its volume for each day, and then averagingthese daily densities).Onaverage,thepercapitawastegenerationwas0.34 kgperday(calculatedbyrstaveraging the daily weight of waste for each household, then dividing this by the numberof residents in each household, as reported in the household survey, and then averaging thedaily per capita waste generation gures across the 49 households). Following is a histogramshowing the frequency of waste per capita data points for individual households. Almosthalf of the households in the study produce between 0.10 and 0.30 kg of waste per capitaper day (see Fig. 7).Fig. 7. Histogram of daily waste per capita results.To give context to these waste generation gures, a study in the centre of the town of SiemReap found that residents produce an average of 0.50 kg of waste each day (ECSPESC andMinistry of Environment, 1997). Waste generation may be lower in the study area becauseof lower average incomes, higher rates of composting or animal rearing (using food scrapsas a source of feed), and less waste from home businesses and tourist establishments. Wastegeneration rates per capita per day vary across the world, and even across the developingworld. The study areas waste generation per capita gures are at the lower end of thosefound in a number of other urban waste generation studies, such as 0.33 kg in Gabarone,Botswana (Bolaane and Ali, 2004); 0.51 kg in Guadalajara, Mexico (Bernache-P erez et al.,2001); 0.63 kg in Morelia, Mexico (Buenrostro et al., 2001); and 1.76 kg in Abu Dhabi City,UAE (Abu Qdais et al., 1997).All of theabovestudiesuseddoor-to-door collectionmethodsfor assessingwastegenerationper capita, andarestudies of residential wastegeneration. Theses studiesdonot discusscommercial wastes(withtheexceptionofBuenrostroet al., 2001, whoalsoconductedaseparatenon-residential wastegenerationstudy), anddonot discusswhethercommercial wastesfromhomebusinesseswerealsopresent intheresidentialwaste stream. We found that many homes in the study area served as a base for a busi-ness,whetheritwasarestaurant,apharmacy,orthesiteforpreparinggoodsthatwerelater sold in another location. The presence of home businesses in our sample no doubtaffected our waste per capita values. One of the primary aims of our research was to assessthecapacitythat wouldberequiredforaCBWMsystemthat wouldcollect all ofthewaste produced in the study area. Therefore, those home businesses that were randomlyselectedforthesub-samplewereconsideredtoberepresentativeoftheconstituentsofthestudyarea, andsotheirinclusioninthestudyenhancedtheaccuracyofourwasteestimates.3.3. Relationship of waste generation to income and locationWe tested for a relationship between waste generation and both household income andaproxyforincome,namelylocationofthehomerelativetotheriver.Wefeltthatthisinformation might prove useful for generalizing the results of the waste generation study toother similar communities in Siem Reap. Additionally, an understanding of the economicnature of waste production in the study area could assist in designing and targeting educationprograms.It seems intuitive that those residents with more income will consume more goods, andtherefore produce more waste. However, most previous research (as discussed below) hasfound that income is not related to waste generation, although several of these studies didnot test for a statistical relationship. It is also difcult to compare these studies because ofdifferences in the way that they measure income. Some use continuous income data, someuse categorical income data, and some use proxy variables for income, such as housing rentalrate and expenditures. However, the use of different measurement approaches is not terriblysurprising given that the difculty of soliciting accurate income data fromhouseholds is wellknown. Many people consider their income to be a private matter while others are reluctantto divulge income data for fear that they might have to pay more taxes. The problem ofincomesolicitationcanbeevenmorechallengingindevelopingcountries,wheremanypeople work in the informal sector with uctuating incomes and have difculty estimatingannual incomes (Adedibu, 1988).Adedibu (1988) used multiple regression analysis to examine the relationship betweenwaste generation and 25 explanatory variables, including income of the head of household,in 324 households in Ilorin, Nigeria. Neither the contribution of income nor the regressionmodel itself was found to be statistically signicant. Based on a sample of 300 householdsin Guadalajara, Mexico, Bernache-P erez et al. (2001) found that there was no relationshipbetween per capita waste generation and family income, although they did not provide anydetails on the type of test used in the analysis or indicate whether income was measured asa continuous or categorical variable.Incontrast toAdedibuandBernache-Perezet al., twootherresearchershavefoundrelationships between income (or income proxies) and waste generation. In a study of 840samples of waste generated by 40 households in Abu Dhabi City, United Arab Emirates, AbuQdais et al. (1997) found a strong positive correlation between household waste generationrates and self-reported annual property rental rates (R2=0.69, signicance not reported).Bolaane and Ali (2004) conducted a waste generation study on 47 households in Gabarone,Botswana and found that waste generation for low and medium income households was thesame, but lower for the high income group. They did not test whether this difference wasstatistically signicant.We tested for the presence of an income relationship in the study area by using bothincome and expenditure data fromthe household survey and comparing themto the adjustedaverage weight of householdwaste andthe per capita weight of householdwaste, usinglinearregression analysis. None of the relationships were found to be statistically signicant (seeTable 1).Wesuspectedthatapossiblereasonforthelackofsignicantrelationshipswastheinaccuracy of the income and expenditure data. One concern was that some householdsTable 1Regression model results for household waste generation by income and by expenditures (n =49)aIndependent variables Dependent variablesAverage waste per householdper dayLn (average waste per capitaper day)R2Probability of F R2Probability of FLn (household expenditures) 0.027 0.268 0.002 0.741Total household income 0.012 0.463 0.026 0.279aHouseholdexpendituresandaveragewastepercapitaweretransformedintotheirnaturallogsinordertoeliminate a problemof heteroscedasticity in the residuals and to ensure that the variable distributions were approx-imately normal. None of the standardized residuals in the regressions exceeded 2.5 and visual inspection of theresidual scatterplots revealed no obvious outliers.withhomebusinessesmighthavereportedgrossincomeratherthannetincome. Otherconcerns included those described in the above noted studies, such as lack of truthfulnessand poor recall for both income and expenditures. In an attempt to reduce the effect of thesepossible inaccuracies, we collapsed the income and expenditure data into three categories(high, medium, low) that each contained about 1/3 of the data points. We then ran an analysisof variance (ANOVA) test for a difference in the mean waste generation levels by incomeand expenditure category, but again found no signicant relationships (see Table 2).Because of the suspected unreliability of the income and household expenditure data,the initial selection of the waste characterization subset was stratied by both householdincome and household location. In the study area, we anticipated that household locationmightberelatedtoincome.Inparticular,weexpectedthatthoseresidentswholiveoneitherbankoftheriver(andwhothereforehavenoofciallandtenure)wouldtendtobe of lower socioeconomic status, and would have distinct patterns of waste production.t-Tests conducted on the adjusted mean household weight of waste and the per capita wastegeneration indicate that there is no statistically signicant difference in waste generationbased on location of the households on the river or on the road (mean household weight ofwaste, p =0.405; per capita waste generation, p =0.605), suggesting that either location wasnot a very effective proxy for income, or that it was a good proxy but that waste generationis not related to income in the study area.Table 2ANOVA results for household waste generation by income and by expendituresSample size Average waste per household per day Average waste per capita per dayMean (kg) Probability of F Mean (kg) Probability of FHousehold incomeLow 13 1.78 0.551 0.275 0.256Medium 20 2.15 0.412High 15 1.90 0.333Household expendituresLow 19 1.88 0.698 0.356 0.941Medium 13 1.93 0.330High 17 2.15 0.3583.4. Relationship of waste generation to household sizeSeveral previous studies have shown that there is a relationship between waste genera-tion per capita and household size. As the number of household members increases, wastegeneration per capita has been found to decrease, probably because of economies of scalein the consumption of goods and packaging. Abu Qdais et al. (1997) found a statisticallysignicant but weak negative relationship between waste generation per capita and house-hold size in Abu Dhabi (R2=0.11), while Bolaane and Ali (2004) found a similar weak,negative relationship in Gabarone, Botswana (R2=0.34). Our results also show that thereis a weak (R2=0.35), but signicant (p =0.000) negative relationship between per capitawaste generation and the number of people in a household.In his waste characterization study in Ilorin, Nigeria, Adedibu (1988) observed that thenumber of people living in a household can vary from week to week as relatives move inand out. If this is the case, our waste per capita estimates may not be completely accu-rate, since the waste characterization study was conducted approximately 1 month after thehousehold survey was completed. We were able to conduct a rough test of the mobility ofhousehold members in the sample area by comparing household size reported in the rstsurvey (conducted before the waste characterization study), to household size reported inthe follow-up survey 1 year later. Of the 273 households that were observed in both surveys,162 households (59.3%) reported different household sizes in the two time periods. Themean number of household members in the initial survey was 6.56, and the maximum was17;inthefollow-upsurvey,themeannumberofresidentswas6.42,andthemaximumwas 14. We ran a paired t-test analysis of these means, and found that there was no sta-tistically signicant difference between them (p =0.319). Although the average householdsize has not changed signicantly, the presence of so many households with differences inhousehold sizes over a 1-year period is still a concern. It suggests that mobility is fairlyhigh in the study area and that even 1 month after the initial survey, at least some house-holds may have lost or gained residents and thus affected our estimates of per capita wastegeneration.3.5. Waste compositionThe waste composition results (as reported in Table 3) are based on the aggregate weightof all waste collected from the study households once it had been sorted into its componentparts.The composition of waste in the study area is largely organic. Kitchen wastes, yard waste,wood, and coconut shells collectively account for 66% of waste by weight. This amountissimilartothatfoundinresidentialwastecharacterizationstudiesinotherdevelopingcountries.Forexample,68%ofwastebyweightisputrescibleinGabarone,Botswana,(Bolaane and Ali, 2004), 53% is putrescible in Guadalajara, Mexico, (Bernache-P erez etal., 2001), and 58%is putrescible by weight in Guangzhou, China, (Chung and Poon, 2001).However, we note that the amount of organic matter observed in the waste stream in ourstudy area represents the amount of organic waste available for collection, not the amountgenerated, since 35% of the households surveyed reported that they currently compost atleast some of their waste or feed it to animals.Table 3Waste composition, by weightWaste composition Percentage (by weight)High nitrogen organics (mostly kitchen wastes) 31High carbon organics (mostly yard wastes) 22Stones and dirt 14Non-recyclable plastic 13Wood and coconut 13Paper 3Metal 1Textiles 1Recyclable plastic 1Glass 1Shells and bones 0.3Medical waste (both hazardous and non-hazardous) 0.3There is a substantial amount of plastic in the waste stream, although it is probable thatthe weights for plastics are slightly exaggerated as this total often included dirt and moisturefrom organics that could not be separated from the plastics. The study found that there arevery fewrecyclables in the waste stream(3%of the waste by weight is made up of paper, 1%is recyclable plastic, and 1% is metal). According to the household survey, 89% of sampledresidents already separate recyclable materials from the waste stream to sell (to itinerantbuyers or local recycling depots) or to give away (to friends or to the less advantaged, forexample). Thesaleofrecyclablesisaregularpractice, andonaverage, ahouseholdinthe study area earns $1.14 USD per month from selling items reclaimed from the wastestream.The amounts of paper and metal observed in the study area are generally lower than thosereported in other residential waste characterization studies (Bernache-P erez et al., 2001;Bolaane and Ali, 2004). Chung and Poon (2001) observe similarly low amounts of paper(6%) and metal (1%) in the waste stream in Guangzhou, China, and note that householdersfrequently set aside paper and metal waste for redemption at private recycling depots. Fewstudies differentiate between recyclable and non-recyclable plastics in the manner that wehave, limiting a comparison of recyclable plastics in the waste stream. Chung and Poon(2001) found that recyclable plastic beverage containers only constituted 0.1% of the wastestreamby weight, in comparison to the category of all recyclable plastics in our study, whichconstituted 1% of the waste stream by weight.Several households in the study area have home businesses and some of the waste fromthese businesses was found to pose special problems for a CBWM project, both in termsof quality and quantity. For example, a household on the west road that runs a pharmacyfrom its residence consistently set out mixed medical waste. This type of waste could posehealth hazards to waste collectors. We observed that in many instances, home businessesadded substantial amounts of waste to the total amount of waste requiring collection froma household. However, one home business produced waste that could be very benecial forCBWM. A carpenter in the study area generates large quantities of wood shavings and, ifcomposting is part of the CBWMprogram, these shavings are an ideal high-carbon additivefor composting piles.4. DiscussionWhile the results of the waste characterization study are valuable as a reference point forcomparison with other communities in Southeast Asia and the developing world in general,the primary importance of these results is their usefulness for waste management planning.Following is a discussion of how the results can be of use in designing a CBWM system inthe study area.4.1. Design of the collection systemAn extrapolation of the waste generation results from the 49 observed households tothe entire study area indicates that the total daily generation for the study area would be1980 kg (calculated by dividing the average daily weight of waste collected from all of thehouseholds by 49, and multiplying this number by 1000), and the total volume for 1000households would be 12.2 m3per day (calculated by dividing the average daily volume ofwaste collected from all of the households by 49, and multiplying this number by 1000).We used these extrapolations to estimate the capacity and number of collection vehicles aswell as the frequency of collection that would be required for the CBWMcollection system.Because conventional garbage trucks are too cumbersome for the study area, alternativetransportation for waste collectors will be required for CBWM. The successful use of a cartattached to a motorcycle for collecting waste in the waste characterization study illustratedfor the local community that this method of collection could be appropriate for the CBWMproject as well.4.2. Source separationBecause the composition of waste in the study area is largely organic, source separationand composting of organics might be a useful strategy for reducing the amount of wasterequiringdisposal.Thirtyvepercentofrespondentstothehouseholdsurveycurrentlymakecompostfromorganicwastes;thepredominantreasongivenforcompostingwasimproving soil quality. Additionally, 75% raise or feed animals, supposedly using some oftheir organic household wastes for this purpose. These ndings suggest that there is already aculture of separating organic waste and composting in this area. However, although the wastecharacterization study may point to composting as an appropriate option for managing thestudy areas waste, results fromthe household survey raised questions about the feasibility ofsource separation. When asked directly about their willingness to separate organic materialfrom the waste stream, only 49% responded positively. This less than enthusiastic responseseems rather strange, given that many households already separate organics for backyardcomposting or for feeding animals. A2analysis indicated that those who already makecompostaremorewillingtoseparatetheirfoodwastes,andthosewhodonotcompostare more opposed to separating their food wastes (p =0.001). We did not nd a signicantrelationship between animal raising and willingness to separate food wastes (p =0.612).Of those willing to separate their wastes, 32% said they are willing to separate all organicwastes, and38%arewillingtoseparatesome. Therest saidtheywouldonlyseparatespecic materials Therefore, despite the high percentage of organic material available forcomposting, it is uncertain whether a source-separation project can proceed with such lowrates of willingness to participate. In addition, wood and coconut (13% by weight) do notbiodegrade quickly, and so may not be as suitable for composting as other organic materials.Alternatively, the community might want to consider mixed waste composting, although theproduction of a high quality compost product can be difcult with mixed-waste composting(Hoornweg et al., 1999).As noted above, only 5% of the waste (by weight) is composed of potentially recyclablematerials. The implication of the low recyclable waste content is that few revenues can beexpected from recovering recyclables. This is unfortunate for the economics of a CBWMscheme evenif it does not include source separation. ManyCBWMprojects expect collectorsto be able to supplement their salaries, which are usually very low, by picking out recyclablesfrom the waste that they collect (Ansch utz, 1995; Richardson, 2003).4.3. Community educationThe toxicity of some commercial waste materials observed in the waste streampresents adanger for collectors (such as some of the pharmaceutical wastes described above). Identi-cationof toxic wastes not suitable for collectionwill needtobe includedinawareness-raisingprograms (for both residents and collectors) prior to project implementation.Ourobservationsduringthestudyindicatethat conceptionsofwastevariedfromhousehold to household. Although we asked for everything that people usually burn, bury,throw in the river, or discard on the ground, it was clear that we received different typesof waste from different people. For example, some people cleared the leaves from theiryard each day and considered this matter to be waste, while others did not give us their yardwaste. Afewhouseholds repeatedly claimed that they had no waste for us to collect, againproblematizing the consistent denition of waste. These observations reinforce the needto engage with local residents about their denitions of waste in order to effectively designan appropriate waste management system. We have not come across any other studies thatinvestigate local meanings of waste.5. ConclusionsThis waste characterization study has proven useful for the design and social program-ming of a CBWM project in Siem Reap. Results from the study have helped determinethe number and capacities of collection vehicles that will be needed for the project. Ourwaste composition results show that there is very little potential for recovery of recyclablesfrom the waste stream. While there is abundant organic waste available for a compostingprogram, about half of the surveyed households are not willing to separate their organicwaste at source. This attitude might change if there is an education program to help house-holders understand the benets of source separation. The waste composition studys ndingof toxics in the waste streamsuggests that an education programmay also be needed to helpresidents understand what materials should and should not be set out for waste collection.Our researchhasalsoshownthat therecommendationsfor CBWMdesigncanbeenhancedifthewastecharacterizationstudyisconductedinconjunctionwithahouse-hold survey. For example, a survey is useful for calculating waste generation per capitabecause there are unlikely to be better sources for providing up-to-date data on the num-berofhouseholdmembers,especiallyinthedevelopingworldwheredemographicandcensus data are often unavailable or unreliable. A survey can also collect information onwaste management attitudes and behaviours, such as whether residents are willing to sourceseparate their waste. It can provide information about culturally contingent perceptions ofwaste that may be important in the design of an education campaign prior to launchinga CBWM project. Of course, a survey is also essential for answering questions about theoverall feasibility of a CBWMsystem, such as household desire for, and willingness-to-payfor, collection services.The researchresults presentedinthis paper represent onlypart of the informationrequiredtoassessthepotential forimplementingaCBWMproject. Otheraspectsofourstudy,reported elsewhere (Parizeau, 2005), have focused on the political and nancial feasibilityof this type of project, such as the cost of labour and equipment, appropriate managementarrangements, the size of the collection fee, and the willingness of various stakeholders toparticipate in the project. Based on these other results, and linking them to the results of thewaste characterization study, we have concluded that the project is nancially feasible, butwill require considerable political will to move forward. We found that local governmentauthorities were very supportive of a CBWM project but that the local waste contractorsees CBWM as a threat because it represents lost future customers. The contractor is there-fore reluctant to provide secondary waste collection services at a reasonable price. At themoment, negotiations between the contractor and the local community are on-going. Weconclude by noting that, although conducting a waste characterization study, complementedby a household survey, is an important element of planning for CBWM, it is just one ofmany needed to set the stage for a successful CBWM project.AcknowledgementsThe authors would like to acknowledge funding assistance from the Canadian Interna-tional Development Agency and fromthe David Chu Scholarship programat the Universityof Toronto. Mr. Phourng Lina of the Department of the Environments Ofce for Pollu-tion Control in Siem Reap provided both translation services and indispensable researchassistance for this work. We are also grateful for the helpful comments provided by twoanonymous reviewers.ReferencesAbu Qdais HA, Hamoda MF, Newham J. Analysis of residential solid waste at generation sites. Waste ManageRes 1997;15:395406.Adedibu AA. Measuring waste generation in third world cities: a case study of Ilorin, Nigeria. Environ MonitAssess 1988;10:89103.Ansch utz, J. Community-basedsolidwastemanagement andwatersupplyprojectsproblemsandsolutionscompared: a survey of the literature. Urban Waste Expertise Programme Working Document 2; 1995.Bernache-P erez G, S anchez-Col on S, Garmendia A, D avila-Villarreal A, S anchez-Salazar ME. Solid waste char-acterisation study in the Guadalajara Metropolitan Zone, Mexico. Waste Manage Res 2001;19:41324.Bo Sokhan. CSARO solid waste management program ofcer. Personal interview; 2004.Bolaane B, Ali M. Samplinghouseholdwaste at source: lessons learnt inGaborone. Waste Manage Res2004;22:1428.Buenrostro O, Bocco G, Bernache G. Urban solid waste generation and disposal in Mexico: a case study. WasteManage Res 2001;19:16976.Chung SS, Poon CS. Characterisation of municipal solid waste and its recyclable contents of Guangzhou. WasteManage Res 2001;19:47385.DMonte, D. How much difference does UNESCO make? The Art Newspaper. Available online at the GlobalHeritage Fund homepage (http://www.globalheritagefund.org); February 11 2005.European Commission Support Programme to the Environmental Sector in Cambodia (ECSPESC) and Ministry ofEnvironment, Department C, Cambodia. Solid waste management questionnaire survey: town of Siem Reap.Questionnaire Survey and Provincial Waste Generation Report, Siem Reap, Cambodia; 1997.Fehr M, de Castro M, Calcado MDR. A practical solution to the problem of household waste management inBrazil. Resources. Conserv Recycl 2000;30:24557.Hoornweg D, Thomas L, Otten L. Composting and its applicability in developing countries. Washington, US:World Bank; 1999.Klundert van de A, Ansch utz J. Integrated sustainable waste managementthe concept. Gouda, The Netherlands:WASTE; 2001.Khulna City Corporation and Swiss Agency for Development and Co-operation. Community-based pilot projecton solid waste management in Khulna City: General Project Description; 2000.Mohee R. Assessing the recovery potential of solid waste in Mauritius. Resour Conserv Recycl 2002;36:3343.Ojeda-Benitez S, Armijo de Vega C, Ramirez-Barreto ME. Characterization and quantication of household solidwastes in a Mexican city. Resour Conserv Recycl 2003;39:21122.Om Caat, Siem Reap Commune Leader. Personal interview; 2004.Parizeau, K. Community-based waste management: a feasibility study for a project in Siem Reap, Cambodia.UnpublishedCurrent IssuesPapersubmittedtotheDepartment ofGeographyandPrograminPlanning,University of Toronto, Canada. Available online at the Waste Econ homepage (http://www.utoronto.ca/waste-econ); 2005.Richardson, D. Community-based solid waste management systems in Hanoi, Vietnam. Unpublished MastersResearch Paper submitted to the Faculty of Forestry, University of Toronto, Canada; 2003.ShimuraS,YokotaI,NittaY.ResearchforMSWowanalysisindevelopingnations.JMaterCyclesWasteManage 2001;3:4859.SiemReapDepartmentofEnvironment.ReportonsolidwastemanagementinSiemReap,SiemReaptown,Cambodia; 2003.UnitedNationsDevelopment Programme. Cambodiaannual report 2003. UnitedNationsDevelopment Pro-gramme, Cambodia; 2004.Wangombe J. Public health crises of cities in developing countries. Soc Sci Med 1995;41:85762.WorldBankTheInternational Development AssociationandtheInternational FinanceCorporation. Coun-tryassistance strategyfor the Kingdomof Cambodia. Available online at the WorldBankhomepage(http://www.worldbank.org); 2005.