Trine Lund Hansen (M&R DTU, Danmarks Tekniske Universitet) og Åsa Svärd (Avdelningen för Vattenförsörjnings- och Avloppsteknik, Lunds Tekniska Högskola)

3.1.4.1 Sopsugsavfall

3.1.5.1 Affaldsprøver

Jes la Cour Jansen og Åsa Svärd, Avdelningen för Vattenförsörjnings- och Avloppsteknik, Lunds Tekniska Högskola.

3.3.1.1 Prøveforberedelse

Procedure

	Der udtages parallelt to prøver (ca. 40 ml) i centrifugerør (50 ml)
	Prøverne centrifugeres i 10 min ved hastighed 3500 omdrejninger pr minut i en laboratoriecentrifuge (Sorwall GLC-2)

Ved centrifugeringen skilles biomassen i tre klart adskilte faser.

På overfladen ligger i nogle prøver en tynd olie/fedtfase. Tykkelsen af laget er i visse prøver 1-2 mm; men kan være svær at bestemme præcist

I midten findes en grumset vandfase (klarfase)

I bunden afsættes en fast fase

	Volumen af hver fase aflæses med 5% nøjagtighed (+-2,5 ml).
	Der udtages prøver af hver fase til TS og VS bestemmelse. (Jfr. metodebeskrivelsen herfor)

Vurdering af metoden

Forbehandlingen af prøverne fra rullesigten i Herning og fra neddeling og magnetseparering i Grindsted skønnes ikke at påvirke undersøgelsen udover den tilsatte vandmængde, der kan fratrækkes voluminet af klarfasen i undersøgelsen.

Ved centrifugeringerne var det ofte ikke muligt at identificere noget flydelag og det blev højst målt til at udgøre 3% af det samlede volumen. I praksis er det derfor ikke muligt at give en rimelig sikker kvantificering af flydelagets indhold af tørstof og af tørstof glødetab. Selvom flydelagets indhold af VS er højt er voluminet så lille at det ikke har betydning for vurderingen af fordelingen af det organiske stof i affaldet mellem en flydende og en fast fase. I den efterfølgende datapræsentation vil resultat af målingerne derfor ikke blive medtaget. Det skal bemærkes at prøver af affald fra restauranter og supermarkeder, der indgår i andre aktuelle undersøgelser kan fremvise væsentligt større flydelag.

Den visuelle inspektion af de forskellige lag kunne kun helt undtagelsesvis afsløre lette materialer - plast i flydelaget eller identificerbare tungere komponenter i den faste fase selvom en tydelig lagdeling ved centrifugeringen, med meget finpartikulært materiale øverst og med mange fibre og enkelte plaststykker nederst. Der har således ikke kunnet identificeres fremmedlegemer som f.eks. metalstumper i større omfang.

Forholdet mellem voluminet af den faste og flydende fase kan bestemmes med stor sikkerhed. Fordelingen af organisk stof målt som VS kan derfor også bestemmes med en sikkerhed svarende til usikkerheden på bestemmelsen af VS. Metoden kan derfor benyttes til en almen karakterisering af fordelingen på faste partikler og væske herunder også fordelingen af organisk af f.eks. målt som glødetab i de to faser. Undersøgelsen har kun undtagelsesvis afsløret identificerbare fremmedlegemer i affaldet, således at det er begrænset information herom, der opnås med metoden.

3.4 Fysisk karakterisering: resultater

3.4.1 Sigtning af forbehandlet kildesorteret organiskdagrenovation.

Der er gennemført sigtning som beskrevet i afsnit 3.3 af alle kombinationer af affald og forbehandling, der indgår i projektet. Tabel 3.4.1 giver en oversigt over prøverne og sidst i afsnittet er alle sigtebillederne vist. Det fremgår at forbehandlingen er domineret af prøver fra rullesigten i Herning og skrueseparatoren i Aalborg.

Tabel 3.4.1.
Prøveoversigt

Sidst i dette afsnit er alle billederne af sigterne sorteret efter forbehandlingsteknik og i figur 3.4.1 vist repræsentative billeder af 16 mm sigterne fra de 5 forskellige forbehandlingsanlæg.

Figur 3.4.1:
Udvalgte billeder af 16 mm sigter af affald fra de 5 forbehandlingsteknikker der indgår i undersøgelsen.

Som det fremgår ovenfor og som en gennemgang af billederne sidst i afsnittet viser, spiller forbehandlingsteknikken en væsentlig rolle for mængden af store partikler. Det fremgår tydeligt:

	at der ved forbehandling på rullesigten efterlades betydelige mængder store partikler på 16 mm og 8 mm sigten uanset affaldstype.
	at der ved forbehandling på skrueseparatoren i Aalborg kun undtagelsesvis efterlades partikler med en størrelse på 16 eller derover. Der er også betydeligt mindre affald på 8 mm sigten end ved forbehandling på rullesigten.
	at den meget enkle forbehandling i Grindsted neddeling og magnetseparering (kun 2 prøver) efterlader mange store partikler på 16 mm og 8 mm sigten.
	at affald fra køkkenkværnene i Malmö (kun 1 prøve) indeholder væsentligt færre partikler på alle sigter end de øvrige affaldstyper.
	at affald fra centralsugsystemet i Malmö forbehandlet i stempelseparatoren ligner affaldet fra skrueseparatoren evt. med lidt flere store partikler.

Billeder af affaldstype bagest i afsnittet er samlet således at affald fra samme geografiske område er placeret efter hinanden. Affald indsamlet i områder med individuelle skraldespande er placeret først. Af de i alt 8 par af denne type efter forbehandling på rullesigte og i skrueseparator ses at der ikke er systematisk forskel mellem affald fra de to forskellige boligområder selvom der i nogle par kan ses forskel på enkelte sigtestørrelser.

Især ved forbehandling på rullesigte ses tydeligt at indsamlingsteknikken i husholdningerne afspejler sig i karakteren af de store partikler. I figur 3.4.2 er vist 16 mm og 8 mm sigten fra de 4 geografiske områder, der har leveret affald til behandling i rullesigten.

Figur 3.4.2:
16 mm sigter øverst og 8 mm nederst fra de 4 affaldstyper, der er forbehandlet i rullesigten i Herning.

For affald fra Hovedstadsområdet indsamlet i papirposer kan der stort set ikke identificeres plast på sigterne.

Affaldet fra Aalborg, Kolding og Vejle indsamles i plastposer. Der kan tydeligt identificeres større og mindre stykker plast på sigterne.

På sigterne af affaldet fra skrueseparatoren er det kun i meget begrænset omfang muligt at identificere karakteren af partiklerne; men der kan dog erkendes grønne plaststykker fra poser anvendt til indsamlingen i køkkenet.

Affaldet fra Grindsted og fra Hovedstadsområdet indsamles i papirposer. Ved sigtning efter den meget begrænsede forbehandling i Grindsted (eller simulering heraf på affaldet fra Hovedstadsområdet) bestående af neddeling og magnetseparering, ses tydeligt store stykker papir på sigterne. Det er ikke umiddelbart muligt at identificere plast i disse prøvetyper.

Affaldet fra køkkenkværnene i Malmö er så fint neddelt at der ikke kan identificeres enkeltbestanddele på sigterne.

Affaldet fra stempelseparatoren benyttet til affaldet fra centralsuget i Malmö er meget fint neddelt; der kan dog identificeres små plaststrimler på sigterne.

3.4.2 Centrifugering af forbehandlet kildesorteret organisk dagrenovation

Der er gennemført centrifugeringer som beskrevet i afsnit 3.3 af alle kombinationer af affald og forbehandling, der indgår i projektet idet de samme prøver, som er sigtet, også er centrifugeret. Tabel 3.4.1 giver en oversigt over de udvalgte prøverne og i tabel 3.4.2 er alle resultaterne samlet.

Tabel 3.4.2
Resultater af centrifugeundersøgelserne

Se her!

Figur 3.4.3 viser prøvernes tørstofindhold opdelt på gløderest og glødetab.

Figur 3.4.3:
Tørstofindhold, gløderest og glødetab i de centrifugerede prøver

Det fremgår at tørstofindhold og glødetab varierer væsentligt afhængigt af hvilken forbehandlingsmetode der er benyttet. Mest skiller biomassen fra køkkenkværnene i Malmö sig ud. Her er tørstofindholdet markant lavere end for de øvrige affaldstyper. Biomassen fra rullesigten i Herning ser generelt ud til at have et højere tørstofindhold end biomassen fra skrueseparatoren i Aalborg; men også en større gløderest. Glødetabet ligger for alle prøver fra 80-95% af tørstoffet. I afsnit 3.6 findes resultater for tørstofbestemmelser fra alle prøver, der indgår i forsøget.

Figur 3.4.4 viser den volumenmæssige fordeling af klarfasen og den faste fase. Flydelaget indgår ikke i figuren, da der kun på få prøver kunne konstateres et sådant lag og i alle tilfælde var det ganske tyndt.

Figur 3.4.4:
Volumenfordeling mellem klarfasen og den faste fase

Det ses af figuren, at den faste fase systematisk udgør en væsentlige større andel af volumenet i prøverne forbehandlet på rullesigten i Herning end behandlet på skrueseparatoren i Aalborg. Affaldet der kun er neddelt (som i Grindsted) ligger imellem. Da der ikke sker nogen frasortering her må fordelingen i dette affald forventes at svare til niveauet for kildesorteret organisk dagrenovation indsamlet i papirposer inden forbehandling.

Affaldet fra køkkenkværnene i Malmö indeholder kun en mindre del fast materiale, naturligvis som følge af det lave tørstofindhold i denne prøve. Den enlige prøve fra stempelseparatoren i Malmö viser et niveau som for rullesigten.

Figuren viser at forbehandling i skrueseparatoren fører til en større andel vand i det forbehandlede materiale og dermed et forventeligt mere tørt rejekt end rullesigten.

Figur 3.4.5. Viser fordeling af VS mellem de to faser efter centrifugering.

Figur 3.4.5:
Fordeling af VS mellem klarfasen og den faste fase efter centrifugering

Det ses at den faste fase indeholder hovedparten af det forbehandlede affalds VS, typisk mere end 80%. Mønstret fra volumenfordelingen er i øvrigt uændret således at det større vandindhold i affaldet fra skrueseparatoren også betyder en større andel VS i klarfasen.

Den store andel VS i den faste fase betyder at der ikke er noget stort tab af VS ved at forbehandlingen resulterer i en tør biomasse. Et højt indhold af vand i biomassen kan dog være gunstigt for at sikre et lavt vandindhold i rejektet, der typisk skal brændes.

I modsætning til de øvrige affaldstyper indeholder klarfasen efter centrifugering af affaldet fra køkkenkværnene i Malmö stort set ingen VS på trods af det meget lave tørstof. Det skyldes formentlig at hovedparten af prøvens vandindhold er skyllevand fra køkkenvasken, og at den del af glødetabet, der ellers måtte forventes i klarfasen er skyllet ud af opsamlingstanken.

3.4.3 Samlet præsentation af alle sigteresultater

På de efterfølgende sider præsenteres alle sigtebillederne ordnet efter forbehandlingen.

3.5 Analytiske metoder

I dette kapitel beskrives procedurerne for de kemiske analyser der er benyttet i dette projekt. Procedurerne er beskrevet ud fra oplysninger fra de laboratorier, som har udført analyserne. Beskrivelserne er ikke helt fyldestgørende for samtlige analyser, da det ikke har været muligt, at skaffe alle relevante informationer.

I beskrivelserne er der lagt vægt på at beskrive det overordnede princip og fremgangsmåden kort og forståeligt, samt angive usikkerheder og måleområde for den enkelte analyse. Rækkefølgen af beskrivelserne er bestemt af de udførende laboratorier, således at analyser udført af DLG Centrallaboratorium er beskrevet først, herefter analyser udført af Miljø & Ressourcer DTU, Danmarks Tekniske Universitet, og til sidst analyser udført af ALcontrol, Sverige.

Beskrivelserne er udført på engelsk for at undgå sproglige misforståelser mellem svensk og dansk.

3.5.1 Crude fat (Fedt)

Analytical approach: The dried sample (5g) is hydrolysed by boiling 1 hour in 125 ml 3M HCl. The fat is collected by filtering through a Whatmann 4 filter and from this filter the fat is extracted with petroleum ether (boiling point 60° C) in a continuous (Soxleth) extraction. The amount of fat in the sample is determined by weighing the extracted fat after evaporation of the petroleum ether and drying the sample at 100° C for 1½ hour. The method is carried out in duplicates.

Manual used for analysis: Legislation 98/64/EC: Determination of Crude Fat. Official Journal of the European Communities, Vol. L257, 19/9-1998

Storage of samples prior to analysis: The dried samples used for this analysis can be stored in closed containers.

Accuracy: DLG Centrallaboratorium is using an internal control sample (standard) which is run every day. Plantedirektoratet has verified the standard. DLG Centrallaboratorium is part of a European ringtest organised by the French company BIPEA.

Repeatability:

0.2% abs if %fat<5%
4.0% relative if 5<%fat < 10%
0.4% abs if %fat >10%

Detection limit: Not determined

Analysis performed at: DLG Centrallaboratorium

3.5.2 Crude fibre (Træstof)

Analytical approach: The sample (1g) is defatted by petroleum ether and hydrolysed with 150 ml 0,13 M sulphuric acid by boiling for 30 min. The sulphuric acid is drained off the sample, it is washed with 3x30 ml boiling water and 150 ml 0,23 M potassium hydroxide is added. This mixture is hydrolysed for 30 minutes followed by filtration and washing as for the acid treatment. The sample is dried (at 130° C), weighed and heated to 475-500° C for at least 30 min.

The sulphuric acid and potassium hydroxide hydrolysis remove everything except ash and the fraction defined as crude fibre. At the last heating to 475-500° C the fibre is burned off and the ash is the only thing left over.

Manual used for analysis: Legislation 92/89/EC: Determination of Crude Fiber. Official Journal of the European Communities, Vol L344, 26/11-1992

Storage of samples prior to analysis: The dried samples used for this analysis can be stored in closed containers.

Accuracy: DLG Centrallaboratorium is using an internal standard, which is analysed every day. The standard has been verified by Plantedirektoratet. DLG Centrallaboratorium is part of a European ringtest organised by the French company BIPEA.

Repeatability:

0.3% abs if %fiber<10%
3.0% relative if %fiber >10%

Detection limit: 0,4% (is not determined, but is the highest value one accepts of a "blank" Sample.

Analysis performed at: DLG Centrallaboratorium

3.5.3 Crude protein (Protein)

Analytical approach: The method used is the Dumas method, which determinates the total amount of nitrogen after total combustion of the sample. By this method all nitrogen in the sample is measured, that means also the nitrogen in nitrate, nitrite- all the inorganic nitrogen.

0.1-1.0 g of sample is pressed into a pellet in nitrogen free paper. This pellet is combusted at high temperature (850° C) to NO_x, which is led over a column of cupper, where the nitrogen is reduced to N₂ and the cupper oxidised to cupper oxide. The amount of nitrogen formed is measured by thermal conductivity. To convert the nitrogen content to crude protein the N content is multiplied by 6.25.

Manual used for analysis: Legislation 93/28/EC: Determination of Crude Protein. Official Journal of the European Communities, Vol L179, 22/7-1993

Storage of samples prior to analysis: The dried samples used for this analysis can be stored in closed containers.

Accuracy: DLG Centrallaboratorium is using an internal control sample (standard) that is run every day. Plantedirektoratet has verified the standard. Concerning both methods DLG Centrallaboratorium is part of a European ringtest organised by the French company BIPEA.

Repeatability:

0.2% abs if %protein <20%
1.0% relative if 20%< %protein <40%
0.4% abs if %protein >40%

Detection limit: The sensitivity of the measurements in the Dumas method is 0.001%N. The detection limit is thus defined as two times the sensitivity, 0.002% N (or 0.0125% protein).

Analysis performed at: DLG Centrallaboratorium

3.5.4 EFOS Kvæg

Analytical approach: This method is used only in Denmark. The method determines the amount of enzyme-digestible organic matter in the sample. 0.5g of the sample is treated with pepsin-HCl in a water bath at 40° C. Afterwards the sample is heated to 80° C for 45 min. to hydrolyse the starch, this is followed by washing with 2x100ml water. Finally the cell wall materials are broken down by treating the sample with cellobiase, hemicellulase and amyloglycosidase at 40° C in a water bath. The left over from this treatment is the indigestible organic matter and the ash,- that means that by ashing, the indigestible organic matter is lost.

Manual used for analysis: An Energy Evaluation System for Determination of the Energy Content in Feed Mixtures for Ruminants and Horses. PD Announcement FO 10/99 (1999). Ministry of Food, Agriculture and Fisheries, Danish Plant Directorate, Lyngby, Denmark

Storage of samples prior to analysis: The dried samples used for this analysis can be stored in closed containers.

Accuracy: DLG Centrallaboratorium uses an internal control sample (standard) that is run every day. Plantedirektoratet has verified the standard.

Repeatability:

0.5% abs for feedstuff

Detection limit: not determined (a blank gives rise to 0,5%).

Analysis performed at: DLG Centrallaboratorium

3.5.5 Starch (Stivelse)

Analytical approach: The samples are analysed by the enzymatic method "Total Starch" by Megazyme. 0.2 ml ethanol and 3 ml a -amylase are added to 0.1 g sample. This is followed by incubation in boiling water for 6 minutes, now the starch is partly hydrolysed and fully solubilised. The sample is transferred to a 50° C bath and treated with amyloglucosidase for 30 minutes. By this treatment the starch is quantitatively hydrolysed to glucose. The amount of glucose is determined by an enzymatic assay using glucose oxidase and peroxidase. By this assay the glucose is oxidised to gluconate and hydrogenperoxide and the amount of hydrogenperoxide measured by reaction with peroxidase and 4-aminoantipyrine. The amount of glucose (g) is converted to amount of starch in the sample (%).

Manual used for analysis: Megazyme internal reference: Megazyme "Total Starch" assay procedure (amyloglucosidase/a -amylase) AA/AMG 9/97. AOAC Method 996.11 and AACC Method 76.13.

Storage of samples prior to analysis: The dried samples used for this analysis can be stored in closed containers.

Accuracy: DLG Centrallaboratorium uses an internal control sample (standard) that is run every day. For homogene samples Megazyme routinely achieve a standard deviation of 2%.

Precision: Not given

Detection limit:

Analysis performed at: DLG Centrallaboratorium

3.5.6 Sugar (Sukker)

Analytical approach: The amount of total sugar is measured in 2.5 g of sample, which is extracted with 200 ml ethanol. After filtration the ethanol is evaporated and the left over is solubilised in 200 ml of water. 50 ml of this is boiled in 15 ml 0.1 N HCl for 30 min to invert the sugar. The sample is neutralised with sodiumhydroxide and diluted to 100 ml with water. An aliquot (often 25 ml) is used for determination of the amount of reducing sugar by a Luff.Schoorl titration (This means: The reduced sugars react with a surplus of Cu+ + forming Cu + ^. Next the remaining amount of Cu+ + ^reacts with J^- forming J₂. The amount of J2 is determined by titration with sodiumthiosulfate). The amount of glucose in the sample expressed as sucrose is calculated from this titration.

Manual used for analysis: Legislation 71/250/EC: Determination of Sugar. Official Journal of the European Communities, Vol L155 12/7-1971

Storage of samples prior to analysis: The dried samples used for this analysis can be stored in closed containers.

Precision: not given in the method

Detection limit: not given in the method

Analysis performed at: DLG Centrallaboratorium

3.5.7 Dry matter (DM, TS, Tørstof)

Analytical approach: Fresh, blended waste is dried in triplicate aluminium trays (approximately 500 g) at 80° C until constant weight (48 hours).

Manual used for analysis: Internal procedure, not published

Reasoning behind the method: A relatively large amount of sample is used for measurements of DM due to the heterogeneity of the sample. The samples are dried at 80° C to minimise evaporation of ammonium and volatile fatty acids.

Storage of samples prior to analysis: It is possible to freeze the samples before analysis.

Accuracy: Not determined

Detection limit: Not determined

Analysis performed at: DTU

3.5.8 Volatile Solids (VS, glødetab, Tørstofglødetab)

Analytical approach: 2-3 g of finely blended, dried, ground organic waste is oxidised in triplicate in porcelain crucibles at 550° C until constant weight (DTU: 2-4 hours). Hereafter the samples are cooled in a desiccator. The VS content is calculated as a percentage of the DM content in each sample.

Manual used for analysis: Internal procedure. Not published.

Storage of samples prior to analysis: The dried, ground sample of organic waste can be stored in closed containers prior to analysis.

Accuracy: Not determined

Detection limit: Not determined.

Analysis performed at: DTU

3.5.9 Nitrogen (Kjeldahl)

Analytical approach:

To measure the Kjeldahl nitrogen content (ammonia + organic nitrogen) the samples (2-6 grams, depending of the nitrogen content) are digested with concentrated H₂SO₄ by raising temperature (200-350° C, during 1½-2 hours). Prior to digestion Kjeldtabs, glass beats, concentrated H₂SO₄ and octanole are added. Distilled water is added to the cooled sample prior to distillation at the Kjeldahl distillation unit. Here the nitrogen is collected in boric acid and titrated with 0.1 M HCl. The samples are run in triplicates.

For ammonia analysis the procedure is similar. However no digestion is performed and the samples are run in duplicates.

Manual used for analysis: Optimisation of Kjeldahl method, Special Course, E &R, DTU, August/September 2001, Massimo Forti og Sune Balle Hansen

Storage of samples prior to analysis: The analysis is carried out on wet samples to prevent any loss of nitrogen during the drying procedure. The wet samples can be frozen prior to analysis.

Accuracy: Standard deviation 5% relative

Detection limit: 1 mg N per sample digested

Analysis performed at: DTU

3.5.10 Volatile fatty acids (VFA)

Analytical approach: The volatile fatty acids are extracted from the wet waste sample (5 ml in Eppendorf tube) by adding 50 m l 17% H₃PO₄ followed by centrifugation (20 minutes, 10000 RPM). The supernatant is transferred to analysis vials (minimum 0.5 ml) and analysed by gas chromatography for acetate, propionate, iso-butyrate, butyrate, iso-valerate and valerate, the sum representing the total VFA concentration. Standards for all the volatile fatty acids are run in 13 concentrations between 0.025 and 60mM.

Manual used for analysis: Internal manual

Storage of samples prior to analysis: The wet samples can be frozen prior to analysis. After extraction the samples should not be frozen (to prevent precipitation in the vials, which can course problems during the GC analysis).

Detection limit: 0,5 mM (= 30 mg/l wet sample)

Analysis performed at: DTU

3.5.11 Chemical oxygen demand (COD)

Analytical approach: 2 ml dissolved, homogenised (no information on the procedure) sample is oxidised with potassium dichromate (K₂Cr₂O₇) in acid solution at 148° C (± 3° C) for 2 hours. Hereby Cr (VI) is reduced to Cr (III) and the colour of the sample changes from yellow into green. This change in colour is measured spectophotometrically at 620 nm and the COD concentration is calculated according to a calibration curve.

Manual used for analysis: Internal procedure, not published. KM Lab AB, Arbetsskrivning Nr 10.

Storage of samples prior to analysis: The dried samples can be stored in closed containers prior to analyses.

Accuracy: The system is calibrated two times a year. For every measurement a control sample (COD=0) is run together with two samples with a known COD concentration (50 and 500 mg/L). The calibration is accepted when triplicate samples lay within +/-5%.

Precision: 25 %. Measurement uncertainty estimated according to Eurachem. This value is corrected by a coverage-factor, k = 2.

Detection limit: 30 mg/l

Analysis performed at: ALcontrol

Literature: SS 028142-2

Analytical Approach: 1 g homogenised sample is extracted by mixing vigorously with HNO₃ and water. Chloride is determined by titration of the diluted sample with silver nitrate.

Manual used for analysis: Internal procedure ALcontrol Laboratories, procedure no. 3022 (14^th edition 2002). Reference: NMKL no. 89, edition 2a, 1986.

Storage of samples prior to analysis: The dried samples can be stored in airtight containers prior to analyses.

Accuracy: The recovery of a known control sample is 100,5 %.

Value derived from latest round robin test.

Precision: 10 %. Measurement uncertainty estimated according to Eurachem. This value is corrected by a coverage-factor, k = 2.

Detection Limit: 10 g/kg

Analysis performed at: ALcontrol

3.5.13 Phosphorous (P)

Analytical Approach: 1 g dried, homogenised sample is digested by adding 7 M HNO₃ and boiled at 120^o C for 60 minutes. The filtered sample is diluted to known volume_. The content of P is determined by ICP-AAS.

Digestion: Swedish Standard SS 02 83 11

Determination: Swedish European Norm SS-EN-11885

Storage of samples prior to analysis: The dried samples can be stored in airtight containers prior to analyses.

Accuracy: The recovery of a known control sample is 97 %. Value derived from a certified reference-material with similar matrix.

Precision: 20 %.Measurement uncertainty estimated according to Eurachem. This value is corrected by a coverage-factor, k = 2.

Detection Limit: 0.1 ppm

Analysis performed at: ALcontrol

3.5.14 Potassium (K)

Analytical Approach: 1 g dried, homogenised sample is digested by adding 7 M HNO₃ and boiled at 120^o C for 60 minutes. The filtered sample is diluted to known volume_. The content of K is determined by ICP-AAS.

Extraction: Swedish Standard SS 02 83 11

Determination: Swedish European Norm SS-EN-11885

Storage of samples prior to analysis: The dried samples can be stored in airtight containers prior to analyses.

Accuracy: The recovery of a known control sample is 101 %.

Value derived from a certified reference-material with similar matrix.

Precision: 20 %. Measurement uncertainty estimated according to Eurachem. This value is corrected by a coverage-factor, k = 2.

Detection Limit: 0.4 ppm

Analysis performed at: ALcontrol

3.5.15 Sulphur (S)

Analytical Approach: 1 g dried, homogenised sample is digested by adding 7 M HNO₃ and boiled at 120^o C for 60 minutes. The filtered sample is diluted to known volume_. The content of A is determined by ICP-AAS.

Extraction: Swedish Standard SS 02 83 11

Determination: Swedish European Norm SS-EN-11885

Storage of samples prior to analysis: The dried samples can be stored in airtight containers prior to analyses.

Accuracy: The recovery of a known control sample is 99 %.

Value derived from a certified reference-material with similar matrix.

Precision: 20 %. Measurement uncertainty estimated according to Eurachem. This value is corrected by a coverage-factor, k = 2.

Detection Limit: 0.1 ppm

Analysis performed at: ALcontrol

3.5.16 Carbon (C), Hydrogen (H) and Nitrogen (N)

Analytical Approach: Determination by the CN analyser CE Instruments EA 1108. 1-2 mg weighed and dried, homogenised sample is oxidised at 1800^o C in HCl with He as carrier gas. The gasses CO₂, H₂O, NO_x, SO₂, and SO₃ are lead into a reduction tube where NO_x is reduced to N₂, and SO₃ to SO_2. The gasses are separated by gas chromatography and determined by a heat detector. Results are calculated by Eager Windows software.

Internal Procedure Mikro Kemi AB no.2003, 8^th ed. of 2001.

Storage of samples prior to analysis: The dried samples can be are stored in airtight containers prior to analyses.

Accuracy: After standardising a control-sample is analysed, BBOT. The concentrations should be in the following interval:

C = 72,53 +/- 0,30 %, H = 6,09 +/- 0,15 %, N = 6,51 +/- 0,15 %.

Precision: Measurement uncertainty not specified for this matrix.

Detection Limit: C = 3 µg/l, H = 1 µg/l, N = 2 µg/l

Analysis performed at: ALcontrol / Mikrokemi

Literature (References)

CE Instruments, Publication list, June 1995, A65
Manual Fisons EA 1108
Valideringsrapport VL2003 Ed2, Mikro Kemi AB
SNV Rapport 3372
MK8044, Mätosäkerhet för kemilaboratorier med metod MK2003
Miller J.C. & Miller J.N. Statistics for analytical chemistry, 3rd edition.

3.5.17 Calorific Value

Analytical Approach

A fixed amount of a substance is combusted at constant volume and at the reference temperature 25 ° C. The result is the gross calorific value of the sample with all water of the combustion product as liquid water. Formulae for calculation of the net calorific value that is water removed as vapour and burning at constant (atmospheric) pressure or given at constant volume is included in the standard.

Manuals used for analysis:

ISO 1928:1995(E) Solid mineral fuels - Determination of gross calorific value by the bomb calorimetric method, and calculation of net calorific value.

Storage of samples prior to analysis: No requirements is given.

Accuracy: Not given

Precision: Not given

Detection Limit: Not given

Analysis performed by: ALcontrol / Mikrokemi

3.6 Kemisk karakterisering: Resultater

I det følgende vises resultaterne af de kemiske analyser af forbehandlet kildesorteret organisk dagrenovation (biomasse) og rejekt. Resultaterne er opstillet i forskellige tabeller for at belyse forskellige problemstillinger.

Tabel 3.6.1 viser variationen mellem de forskellige indsamlingsområder. Analyseresultater for hvert område er vist for individuel og fælles skraldespande for hver forbehandlingsmetode. I hver af disse kategorier er der beregnet gennemsnit for hver kemiske parameter, ligesom det totale gennemsnit for hver parameter er beregnet for hvert indsamlingsområde.

Tabel 3.6.2 viser de kemiske analyser opdelt efter forbehandlingsmetode. For hver parameter er angivet gennemsnit for de enkelte kombinationer af indsamlingsområde og F og I indenfor hver forbehandling, ligesom de samlede gennemsnit for hver kemiske parameter for den enkelte forbehandlingsmetode er angivet.

I tabel 3.6.3 ses variationen i den kemiske sammensætning over tid. For tre kombinationer af indsamlingssted og forbehandling er der foretaget 5-6 prøvetagninger over en tidsperiode på ca. 1 år. Dette gælder affald fra Hovedstadsområdet (både fælles og individuel) forbehandlet på rullesigte i Herning, affald fra Aalborg (både fælles og individuelt) forbehandlet i skruepresse i Aalborg og affald fra Grindsted (individuelt) forbehandlet med neddeler + magnet i Grindsted. I tabellen vises udvalgte kemiske parametre for de forskellige kombinationer.

Tabel 3.6.4 viser udvalgte kemiske parametre for den kontrolprøve, der er analyseret gentagne gange. Under hver kolonne er angivet middelværdi og relativ standardsafvigelse (standardafvigelse divideret med middelværdien) for den aktuelle parameter.

Tabel 3.6.5 viser resultater fra den statistiske undersøgelse af prøvetagningsmetoden i henholdsvis Herning (rullesigte) og Aalborg (skrueseparator). Prøvetagningsproceduren blev statistisk vurderet ud fra en modificeret prøvetagningsplan. Herved kunne de største usikkerheder i prøvetagningsproceduren identificeres. Der er flere prøver for proceduren for rullesigten i Herning end for skrueseparatoren i Aalborg. Dette skyldes, at der er flere trin i prøvetagningsproceduren fra rullesigten, da denne biomasse er mindre homogen end biomassen fra skrueseparatoren. Undersøgelsen er nærmere beskrevet i afsnit 3.7.

Tabel 3.6.6 og 3.6.7 viser resultater af analyser af den organiske fraktion af rejekterne opgjort på henholdsvis indsamlingsområder og forbehandlingsmetoder.

Sidst i afsnittet findes to oversigtstabeller, tabel 3.6.8og tabel 3.6.9, der indeholder samtlige målte kemiske parametre for alle analyserede biomasse- og rejektprøver. De kemiske analyser af rejektet blev lavet for at undersøge, om der var generelle forskelle mellem biomasse og rejekt eller mellem rejekter fra forskellig forbehandling. De kemiske analyser af rejekterne er foretaget på den organiske fraktion af rejekterne og er dermed udtryk for sammensætningen af denne og ikke hele rejektfraktionen.

I analyseprogrammet indgik oprindeligt måling af COD på samtlige prøver. Målemetoden har dog vist sig at være så usikker, at det er valgt at se bort fra denne analyse.

Tabel 3.6.1.a:
Indsamlingsområde: Kemiske analyser af biomasse fra affald indsamlet i Hovedstadsområdet forbehandlet med forskellige forbehandlingsmetoder.

Se her!
   

Tabel 3.6.1.b:
Indsamlingsområde: Kemiske analyser af biomasse fra affald indsamlet i Aalborg forbehandlet med forskellige forbehandlingsmetoder.

Se her!
   

Tabel 3.6.1.c:
Indsamlingsområde: Kemiske analyser af biomasse fra affald indsamlet i Grindsted forbehandlet i Grindsted (dette affald blev ikke forbehandlet andre steder)

Se her!
   

Tabel 3.6.1.d:
Indsamlingsområde: Kemiske analyser af biomasse fra affald indsamlet i Kolding forbehandlet med forskellige forbehandlingsmetoder

Se her!
    

Tabel 3.6.1.e:
Indsamlingsområde: Kemiske analyser af biomasse fra affald indsamlet i Vejle forbehandlet med forskellige forbehandlingsmetoder

Se her!
     

Tabel 3.6.1.f:
Indsamlingsområde: Kemiske analyser af biomasse fra affald indsamlet i Malmö forbehandlet med forskellige forbehandlingsmetoder

Se her!
   

Tabel 3.6.2.a:
Forbehandling: Kemiske analyser for biomasse behandlet på rullesigte

Se her!
   

Tabel 3.6.2.b:
Forbehandling: Kemiske analyser for biomasse behandlet i skrueseperator

Se her!
   

Tabel 3.6.2.c:
Forbehandling: Kemiske analyser for biomasse behandlet med neddeler + magnet

Se her!
   

Tabel 3.6.2.d:
Forbehandling: Kemiske analyser for biomasse behandlet med centralsug + stempelseparator

Se her!
    

Tabel 3.6.2.e:
Forbehandling: Kemiske analyser for biomasse behandlet i køkkenkværne

Se her!
   

Tabel 3.6.3.a:
Tidsmæssig variation. Resultater af kemiske analyser af biomasse fra affald indsamlet i Hovedstadsområdet forbehandlet på rullesigte i Herning.

Se her!
   

Tabel 3.6.3.b:
Tidsmæssig variation: Resultater af kemiske analyser af biomasse fra affald indsamlet i Hovedstadsområdet forbehandlet i skrueseparator.i Aalborg

Se her!
   

Tabel 3.6.3.c:
Tidsmæssig variation: Resultater af kemiske analyser af biomasse fra affald indsamlet i Grindsted forbehandlet med neddeler + magnet i Grindsted.

Se her!
   

Tabel 3.6.4:
Resultater af gentagne analyser af kontrolprøve

Se her!
    

Tabel 3.6.5:
Resultater fra analysen af prøver fra den statistiske undersøgelse af prøvetagningsproceduren (se afsnit 3.7)

Se her!
   

Tabel 3.6.6.a:
Indsamlingsområde: Hovedstadsområdet. Analyser af den organiske fraktion af rejekterne

Se her!
   

Tabel 3.6.6.b:
Indsamlingsområde: Aalborg. Analyser af den organiske fraktion af rejekterne

Se her!
   

Tabel 3.6.6.c:
Indsamlingsområde: Kolding. Analyser af den organiske fraktion af rejekterne

Se her!
   

Tabel 3.6.6.d:
Indsamlingsområde: Vejle. Analyser af den organiske fraktion af rejekterne

Se her!
   

Tabel 3.6.6.e:
Indsamlingsområde: Malmö. Analyser af den organiske fraktion af rejekterne

Se her!
   

Tabel 3.6.7.a:
Forbehandling: Rullesigte. Analyser af den organiske fraktion af rejektet

Se her!
  

Tabel 3.6.7.b:
Forbehandling: Skrueseparator. Analyser af den organiske fraktion af rejektet

Se her!
  

Tabel 3.6.7.c:
Forbehandling: Centralsug + stempelseparator. Analyser af den organiske fraktion af rejektet

Se her!
   

Tabel 3.6.8:
Sammenstilling af kemiske analyser: TS, VS, pH, organiske delkomponenter, EFOS (enzymfordøjeligt organisk stof) og VFA (volatile fatty acids) for biomasse (A) og rejekt (R ) for samtlige prøvetagninger.

Se her!
    

Tabel 3.6.9:
Sammenstilling af de kemiske analyser (Kemiske enkeltkomponenter og brændværdi) for biomasse (A) og rejekt (R) for samtlige prøvetagninger.

Se her!

3.7 Assessment of sampling and chemical analysis of source-separated organic household waste

Manuscript submitted to Waste Management

Jes la Cour Jansen^1’, Henrik Spliid², Trine Lund Hansen³, Åsa Svärd¹, Thomas H. Christensen³

¹Water and Environmental Engineering, Lund University,
P.O. Box 118, SE-221 00 Lund, Sweden
²Informatics and Mathematical Modelling, Technical University of Denmark, DK-2800 Kongens Lyngby, Denmark
³ Environment & Resources DTU, Technical University of Denmark,
DK-2800 Kongens Lyngby, Denmark

* Corresponding author: email jes.la_cour_jansen@vateknik.lth.se

Abstract

The quality of the waste sampling procedure and chemical analysis was evaluated in a research program on characterization of organic waste obtained after disc screening of source-separated organic household waste. The sampling procedures focused on a truckload of waste and involved several steps of subsampling including shredding, mixing, blending, high-speed-blending, drying and milling prior to analysis of the organic waste with respect to ash content, crude fibers, crude fat, crude protein, sugar, starch, enzyme-digestible organic matter, P, N, C, H, S and calorific value. The statistical evaluation of the procedures involved 10 samples of the same truckload of waste obtained by splitting the sample at each level in the procedure according to a staggered, incomplete nested statistical design. Furthermore one sample was analysed six times over a period of approximately one year. The statistical evaluation showed that no single step in the sampling procedure contributed with excessive variance and that the variance caused by the sampling procedure was approximately the same as the variance in the chemical analysis observed over a year. The variance varied with the analytical parameter but for most parameters the uncertainty was satisfactorily low (of the order of 3-10% expressed as the relative standard deviation, which is considered to be satisfactory for waste characterization).

Introduction

Source separation and biological treatment of organic household waste has been introduced in many Danish cities in recent years as a means of increasing recycling of household waste. Recycling of paper and glass is mandatory in all municipalities and yields 20 - 25% of household waste recycling, but only by including also the organic fraction is it possible to reach recycling goals of at least 50% set by many cities for household waste. The source-separated organic waste is treated by composting or by anaerobic digestion with the intension of using the treated waste as a soil amendment.

The composition of the source-separated organic fraction of household waste that is biologically treated may vary depending on several local factors: The sorting criteria specified by the municipality for use by the households, the efficiency of the citizens in sorting properly, the collection system including the types of collection bags used in the kitchen (paper, plastic) and local storage bins (containers, paper sacks) and finally the pretreatment that is used (disc screen, screw separator, magnetic separator etc.) prior to the biological treatment. The composition of the waste is a key parameter in assessing the potential of nutrient (N, P and K) and carbon (C) recovery as well as the potential for biogas production both on a national/regional level and with respect to identifying the most efficient systems.

One major problem in obtaining reliable data on the composition of source-separated organic household waste is the difficulty in obtaining a representative sample for analysis in the laboratory. This problem is well known with solid waste in general: A collection truck holds maybe 2-5 ton of waste, which is very inhomogeneous both physically and chemically, and the chemical analysis performed in the laboratory may need only 1-2 gram of material. The sampling procedure is costly if good representativity is needed and usually case specific approaches are used in order to obtain as many samples as possible for the resources available. However, often the procedures are not well described and their reliability is not assessed.

The purpose of this paper is to provide a detailed description and statistical assessment of the sampling procedure that we recently applied in a project for the Danish EPA, where we sampled and chemically characterized source-separated organic household waste from five Danish cities representing different approaches to recycling of organic household waste.

Routine sampling procedure

The organic waste studied originated from full-scale urban source-separation schemes or experimental schemes that had been in operation for more than six months. The organic waste was separated by the citizens in their kitchen in special bags, brought to local bins that were emptied weekly or every second week by ordinary waste collection crews. The waste was collected in compacting trucks holding 2-4 tons per load and transported to the pretreatment facility, in some cases after transfer to long-haul vehicles.

The sampling procedure described below and statistically evaluated is the procedure used for waste subject to disc screening only. This yields the most inhomogeneous waste and, thus, this procedure is believed to be the one most vulnerable to uncertainty. The procedure has been applied to the source-separated organic waste after disc screening and focuses on the amount of waste contained in a single truckload.

Chemical characterization

Standard environmental and feedstuff analytical procedures were used to characterize the organic waste.

The wet sample taken after blending of the waste (before water was added) was characterized in terms of volatile components (volatile fatty acids, ammonia), water content and pH.

The dried sample obtained after the full procedure described above was used for characterization in terms of:

	Direct measure of organic matter: Volatile Solids, calorific value and EDOM (Enzyme Degradable Organic Matter)
	Specific analyses of organic matter: Fat, protein, crude fiber, starch and sugar
	Basic elements involved in biological conversion: Carbon, hydrogen, nitrogen and sulphur
	Basic elements being important nutrients: Kjeldahl–nitrogen, phosphorous (P) and potassium (K)
	Other elements, potentially inhibiting: Chloride

Testing of sampling procedure

In order to assess the routine sampling procedure, a special effort was made on a single occasion involving sampling of multiple samples according to a staggered statistical design. The basic idea was to find the overall uncertainty of the full procedure from sampling in the field, sample preparation in the laboratory and chemical analysis and to identify any step in the procedure significantly contributing to the overall uncertainty.

The chemical parameters involved in the statistical analysis are those made on the dried sample since only these represent the full sampling procedure and chemical analysis.

Statistical sampling design

The design of the experiment is a so-called staggered design and is an incomplete nested design as shown in Figure 3 (Bainbridge, 1965; ISO 5725-3). The complete nested design appearing in Figure 3 requires 32 samples, as each sample is split into two subsamples at every step in the procedure not including the full amount of the waste sample. However, a special design involving only 10 measurements was actually performed. The advantages of this design is that all the levels in the hierarchy are assessed with the same relative precision, thus being a much more economical design compared to conventional designs. The samples actually taken are shown in the right side of Figure 3. The number to the left of the parentheses is the high-speed blender (h-s blender) number while the indices inside the parenthesis are part of the pile (i), shredder (j), barrel (k) and blender (l) number, respectively, giving m(ijkl), the high-speed blender number within part, shredder, barrel and blender. The indexing is further explained below.

The sampling design took the starting point in the pile of organic waste on the concrete slab that was divided in two parts (P) indexed by i being either 1 or 2. The second sampling level is indicated by the collection in the wheelbarrow and shredding of the waste (S) with index j. From each part two shredded samples were taken, and j is 1 or 2 within each part.

For two of the shredded wastes only one barrel (B) was formed, while from each of the remaining two shredded wastes two barrels were formed. Thus the third index k indicates the barrel within shredder, showing the value 1 at two of the shredded wastes and 1 or 2 at the other shredded wastes. Thus a total of six barrels were formed form the four shredded wastes.

At the fourth level two blended (I) samples, indexed l, were formed from two of the barrels and only one sample from the remaining four barrels. Thus the index l takes the value 1 or 2 for two blended samples, and 1 if only one blended sample was formed from the barrel.

Finally at the fifth level two high-speed blended (D) samples, indexed m, were formed for two of the blended samples and only one for the remaining six blended waste samples.

Table 1 shows the described indexing (from bottom and upwards in Figure 3) and also the data for the measurement of parameter EDOM (enzyme-digestible organic matter).

The mathematical model for this design, for example, Y being the response variable EDOM, can be written as

Y_ijklmn = µ + P_i + S(P)_j(i) + B(PS)_k(ij) + I(PSB)_l(ijk) + D(PSBI)_m(ijkl) + e_n(ijklm)

where the random terms P, S(P), B(PS), I(PSB), D(PSBI) (properly indexed) are components of variance describing variation between parts, shredder samples within parts, between barrels, samples within shredders, between blenders within barrels and between high-speed blenders within blenders, respectively. Finally e_n(ijklm) is the measurement error experienced at the laboratory where the samples are analysed. The parenthesis notation indicates the hierarchy of the variation. In the design two measurements were taken at the laboratory for all 10 samples, such that the index n has the value 1 or 2.

In the study two different laboratories performed the chemical analyses. Half of the measurands were analysed by Laboratory 1 and the remaining measurands were analysed by Laboratory 2. Laboratory 1 made double determinations, whereas Laboratory 2 made only one determination.

Analysis of variance

In a complete nested design the data are analysed using a standard nested analysis of variance technique, but because of the incompleteness of the design, the analysis must be based on a more general approach. The details of this technique are considered beyond the scope of the present discussion. However, the analysis of variance of one measurand, EDOM is presented in detail in Table 2.

Table 2 shows that the coefficients for a component of variance, for example  (the variance between blenders within parts, shredders and barrels), are not very different in the mean squares where the component appears. Thus, for example when testing the B(PS)-term, it is reasonable to use the I(PSB)-term directly, giving an approximate F(2,2)-value = 2.467/2.477 = 1.00 and corresponding p-value = 0.50. The table indicates, that the variation between high-speed blender samples (within blenders, parts, shredders and barrels) is of the same order of magnitude as the variation between the two measurements taken on the same final sample. The measured variation between blenders, I(PSB), barrels, B(PS), and shredders, S(P), are all of the same order of magnitude and not significantly different. Finally, the variation between parts is significant in this table.

This methodology of analysis leads to two apparently significant sources of variation, namely the variation between parts and variation between blender samples. All the measurands considered were analysed along the same lines expecting that other measurands would show similar significant difference between samples at the same levels in the sample hierarchy. However, no systematic pattern was found by the described technique for analysis of variance. This suggests that no single step in the procedure, or in statistical terms a specific level in the hierarchy, in general is particularly the cause of variance in the data.

Nonparametric analysis of levels in the hierarchy

Instead of analysing the data each measurand at a time an attempt has been made to analyse each division level across all the measurands. As an example the level 'part' is chosen. For all measurands the test of 'part' in the analysis of variance is performed as shown in Table 2. For all levels in the hierarchy cases an F-test quantity and the corresponding p-value is computed.

Assuming that there is no variation induced from the source of variation 'part' the p-values will be uniformly distributed over the interval [0, 1]. Table 4 shows in the first row all p-values for 'part' organised according to measurands. The other rows contain the p-values for the other levels of variation.

The test for uniform distribution of the p-values can be carried out as a Kolmogorov-Smirnov test or an Anderson-Darling test (Stuart et al., 1999). The details of these tests are left out here, but they can be illustrated as shown in Figures 4 and 5.

Figure 4 plots the empirical cumulative distribution of the p-values together with the uniform distribution being the straight line from (0,0) to (1,1). The two other lines are 5% significance limits for the empirical distribution. If the empirical distribution stays within the significance limits the hypothesis of uniformly distributed p-values cannot be rejected at a 5% level of significance. The Anderson-Darling test tests the same hypothesis, but its power is somewhat better and a little more emphasis is put on the upper and lower parts of the distribution. Both tests indicate that the variation related to the high-speed blender level is significant when compared to the residual variation. In this test the residual variation is assumed to be measured in two independent samples taken from the same high-speed blender. However, in the present investigation the two measurements were made as double determinations on the same sample and hence the variation between these double determinations presumably underestimates the variation between independent samples from one high-speed blender level. Thus, in our opinion, the significant variation between high-speed blender samples is probably due to the analytical procedure and hardly an expression of real variation induced at the high-speed blender level.

No general signs of significant variation are seen at the other levels. In the lower right figure all p-values for all levels except the high-speed blender level are tested and generally all these p-values can be assumed to be uniformly distributed in the interval [0, 1].

For the remaining measurands only one determination was made, due to shortage of sample material, making the high-speed blender level the final level of variation. The design, of course, is the same as shown above. The analysis of variance method is similar, and again only few significant variations were found. The resulting p-values for these measurands for the different levels of variation are shown in Table 5.

The test for uniform distribution of the p-values is again carried out as a Kolmogorov- Smirnov test or an Anderson-Darling test. The results are very similar to the above results as seen in Figure 5. Generally no signs of significant variation are seen at the other levels. Again in the lower right part of Figure 5 all p-values for all levels except the high-speed blender level are tested and generally all these p-values can very well be assumed to be uniformly distributed in the interval [0, 1].

Analysis of variation from repeated analysis of samples

In order to assess the analytical quality of the participating laboratories one sample was prepared and subsamples from it were sent to the analytical laboratories along with the samples from the routinely sampled organic source-sorted household waste. The sample was analysed six times during the experimental period of about a year.

The results reported by the two laboratories on principally identical samples are shown in Table 6. In that same table the standard deviation assessed by the variation between high-speed blender samples in the experiment is given in the last line. No trends or outliers are seen.

One important observation is that the standard deviations judged by the variation between high-speed blender samples generally do not deviate significantly from the standard deviations estimated from the repeated analysis. When judging the figures it should be born in mind that the high-speed blender standard deviations are based on only two degrees of freedom (and therefore somewhat imprecise) but the ratios between these two sets of standard deviations are neither systematic nor large.

This result demonstrates that the high-speed blender variation observed in the waste sampling study is representative for the measurement uncertainty in the experiment. Since the uncertainties observed in the sampling study and in the repeated analysis of the samples over time are comparable there is no indication that the sampling procedure is inadequate adding significant additional uncertainty at any specific level of the procedure.

Conclusions

With EDOM as an example it was possible to find a few statistically significant sources of variation related to the levels in the waste sampling procedure for some of the measurands. However none of these significant sources of variation are general in the sense that they are common to several measurands. In fact, the number of single significant results does not exceed what under all circumstances would be expected. It may therefore be concluded that no additional variance or inhomogeneities seem to be introduced at any of the levels in the sampling procedure for the source-separated organic household waste. The six repeated independent analysis of a single sample showed that the analysis of the individual measurands is stable and exhibit a reasonable level of variation comparable to what was found in the analysis of the sampling procedure.

The variance varied with the analytical parameter but for most parameters the uncertainty was satisfactorily low (of the order of 3-10% expressed as the relative standard deviation, which is considered to be satisfactory for waste characterization).

The overall conclusion is that the waste sampling procedure and analysis is reasonably representative of the waste contained in a single truck and the uncertainty of the data is reasonable low.

The statistical design applied indicates that valuable information about the quality of the waste sampling procedure and chemical analysis can be obtained without taking an excessive number of samples. In the case presented 10 samples were needed, which is nine additional samples since one sample should have been taken anyway. The overall research programme contained a total of 68 field samples, and only about 13% of the samples had to be analysed in order to document the quality of the waste sampling procedure.

References

Bainbridge, T.R., 1965, Staggered nested designs for estimating variance components, Industrial Quality Control 22 12-20.

International Organisation for Standardisation, 1994, Accuracy (Trueness and Precision) of Measurement Methods and Results, ISO 5725-3, Geneva.

Stuart, A., Ord. J.K., Arnold, S., 1999, Kendall’s Advanced Theory of Statistics, 6th ed., Oxford University Press Inc., New York.

Legends

Figure 1: Field sampling procedure.

Figure 2: Laboratory sampling procedure.

Figure 3: Completed nested design with indication of the actual samples.

Figure 4: Illustration of Kolmogorov-Smirnov test (Laboratory 1).

Figure 5: Illustration of Kolmogorov-Smirnov test (Laboratory 2).

Table 1. The indexing and analytical results of EDOM (% of VS).

Table 2. Results from the analysis of variance of EDOM (% of VS).

Table 3. Results of the analysis of variance for "parts" (EDOM, % of VS)

Table 4. p-values for the measurands for different levels related to Laboratory 1.

Table 5. p-values for the measurands for different levels related to Laboratory 2.

Table 6. Results of repeated analysis of the same sample and the variance observed in the analysis of the sampling procedure represented by the variance at the level of the high-speed blender (right column).

Figure 1:
Field sampling procedure
   

Figure 2:
Laboratory sampling procedure
    

Figure 3:
Nested design with indication of the staggered design showing the actual samples.
   

Figure 4:
Illustration of Kolmogorov-Smirnov test (Laboratory 1). The abscissa axis shows observed p-values and the ordinate axis is the cumulative probability.
    

Figure 5:
Illustration of Kolmogorov-Smirnov test (Laboratory 2). Empirical distribution of p-values and 95% confidence limits. The abscissa axis shows observed p-values and the ordinate axis is the cumulative probability.
    

Table 1.
The indexing and analytical results of EDOM (% of VS).

Table 2.
Results from the analysis of variance of EDOM (% of VS).

Table 3.
Results of the analysis of variance for "parts" (EDOM, % of VS), reduced.

Table 4.
p-values for the measurands for different levels related to Laboratory 1.

Table 5.
p-values for the measurands for different levels related to Laboratory 2.

Table 6.
Results of repeated analysis of the same sample and the standard deviations observed in the analysis of the sampling procedure represented by the standard deviation at the level of the high-speed blender (right column).