Repository navigation
Späder en mellanhand ut skillnader i uppströmsdata? (MF-002) #49
Replies: 4 comments
|
Here is a proposal, written up as MF-002. It is a draft and has not been decided. The cause is in the specification. A node applies one share, Shareᵢⱼ, to its whole emission mass (§3, §4), so every recipient of a node gets the same mix per unit of the key. Dilution can only be avoided through how nodes are delimited. The proposal adds four things:
In the example above, one node allocating by mass gives D = ½ · (|0.50 − 0.17| + |0.50 − 0.83|) = 0.33. A third of the upstream emissions end up with the wrong customer. With a node per supplier or per customer, D = 0. A node per site makes the node's own Scope 1 and 2 more accurate but leaves D unchanged, as long as goods from both suppliers pass through the same site. Coverage is diluted the same way. If the low-emission supplier reports under MASSIV+, so that its 40 t are actual data (A), and the other supplier's 200 t are unknown (U), one node gives both customers Coverage 0.18, that is (5 + 40) / 245. With a node per supplier, X gets Coverage 1 and Y 0.01. D also depends on the key. Allocating by revenue can give D above 0 even with a single supplier, if two customers pay different prices for the same goods. Open questions:
The text change to the specification has not been written yet. Comments are welcome here. |
|
Here is a draft reply, as far as I had time today:
1. Återinför A2-meningen: gods är allokeringsinformation och evidens inom noden, inte nodgrund – annars kan ingen uppfylla “noden vet”.
2. Återinför “uppströms” i avgränsningsregeln, eller argumentera uttryckligen för kundnoder; D kan inte vara argumentet.
3. Nyckelinkonsekvensen: ρ fysiskt, Share valfritt – rekommendera fysisk nyckel för uppströmsutsläpp, och skriv ut att monetär nyckel straffar grönt stål.
4. Avgränsa MF-002 till icke-blandande noder, eller definiera ρ för blandande (stycklista) och “ej tillämpligt” vid fysisk blandning.
5. Flödesnoder som §1-noder är instabila mellan perioder – medvetet val?
6. Nedströmsregel: partitionsändring åtföljs av föregående period omräknad under nya partitionen.
7. D är lokalt: skriv att det inte fångar ärvd utspädning, och avgör vad §5 “per kvalitet” betyder när samma kund köper båda kvaliteterna.
From: Birger Löfgren ***@***.***>
Sent: Monday, 19 October 2026 16:25
To: ICONSOF/massivplus-public-draft ***@***.***>
Cc: Christoffer Krewer ***@***.***>; Mention ***@***.***>
Subject: Re: [ICONSOF/massivplus-public-draft] Does a pass-through node dilute low upstream emissions? (Discussion #49)
Here is a proposal, written up as MF-002<https://docs.massivplus.org/forslag/mf-002-utspadning/>. It is a draft and has not been decided.
The cause is in the specification. A node applies one share, Shareᵢⱼ, to its whole emission mass (§3, §4), so every recipient of a node gets the same mix per unit of the key. Dilution can only be avoided through how nodes are delimited.
The proposal adds four things:
1. A recommendation. If a node knows which inflow went to which recipient, it should be delimited so that inflows from different suppliers are not allocated together. How finely to split remains the user's choice, as in §1.
2. What delimits a node. A counterparty, or a physical unit whose emissions can be measured on their own, never the goods. A node per product would give a number that follows the product, which §5 allows only at a sink.
3. A measure, Dilution (D). The share of the node's upstream emissions, A + U, that ends up with the wrong recipient:
4. D = ½ · Σⱼ |Shareᵢⱼ − ρᵢⱼ|
ρᵢⱼ is the recipient's share of upstream emissions if each supplier's emissions follow that supplier's goods. Both terms are shares, so D works with every allocation method in §3. This is the separate measure §8 anticipates for the relevance axis, for the upstream part.
1. Disclosure. The node reports D alongside Coverage, together with what set its limit: missing data or its own choice.
In the example above, one node allocating by mass gives D = ½ · (|0.50 − 0.17| + |0.50 − 0.83|) = 0.33. A third of the upstream emissions end up with the wrong customer. With a node per supplier or per customer, D = 0. A node per site makes the node's own Scope 1 and 2 more accurate but leaves D unchanged, as long as goods from both suppliers pass through the same site.
Coverage is diluted the same way. If the low-emission supplier reports under MASSIV+, so that its 40 t are actual data (A), and the other supplier's 200 t are unknown (U), one node gives both customers Coverage 0.18, that is (5 + 40) / 245. With a node per supplier, X gets Coverage 1 and Y 0.01.
D also depends on the key. Allocating by revenue can give D above 0 even with a single supplier, if two customers pay different prices for the same goods.
Open questions:
* D measures the node as a whole. With 20 tonnes from the low-emission supplier and 180 from the other, D is only 0.08, but X gets 36.8 t instead of 8. Is a measure per recipient needed as well?
* D is computed on A + U together. Two suppliers with equal emissions, one A and one U, give D = 0, yet one node still gives both customers Coverage 0.5. Should D also be computed for A and U separately?
* In an electricity or district heating grid the goods cannot be followed, so ρ is not defined.
* Without knowing which inflow went to which recipient, D can only be estimated from the inflows. That works for mass and energy content, while revenue needs an approximation.
* A tool could suggest a node partition by trying splits and showing how D falls with the number of nodes. Whether that is practical needs to be explored.
The text change to the specification has not been written yet. Comments are welcome here.
—
Reply to this email directly, view it on GitHub<#49?email_source=notifications&email_token=CL6EX5ALN3FR5XFAOUYY52L5UYP3VA5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCOJQGQZDGOJZUZZGKYLTN5XKO3LFNZ2GS33OUVSXMZLOOSWGM33PORSXEX3DNRUWG2Y#discussioncomment-19042399>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/CL6EX5CQVAVZ66SJQIAFZG35UYP3VAVCNFSNUABJKJSXA33TNF2G64TZHMYTENJRGU4DSOBTGQ5UI2LTMN2XG43JN5XDWMJRGA3TGMJXGWQXMAQ>.
Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS<https://github.com/notifications/mobile/ios/CL6EX5AEL74BJBOSQOXJ6Q35UYP3VA5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCOJQGQZDGOJZUZZGKYLTN5XKO3LFNZ2GS33OUVSXMZLOOSVGM33PORSXEX3JN5ZQ> and Android<https://github.com/notifications/mobile/android/CL6EX5HCADPMGOZ2NFBLWU35UYP3VA5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCOJQGQZDGOJZUZZGKYLTN5XKO3LFNZ2GS33OUVSXMZLOOSXGM33PORSXEX3BNZSHE33JMQ>. Download it today!
You are receiving this because you were mentioned.Message ID: ***@***.***>
|
|
MF-002 har nu omarbetats till ett konkret förslag om D vid sidan av Coverage. Nodindelningen och informationsutbytet har fått var sitt förslag, så att de kan granskas och beslutas separat. Trådens inledning är uppdaterad på svenska; tidigare kommentarer finns kvar. Läs omarbetat MF-002. Texten ligger på publiceringsgrenen och inväntar granskning före webbpublicering. Christoffers tidigare invändningar är centrala för omarbetningen:
Kundperspektivet är också preciserat. Ett lågt samlat D kan dölja en stor avvikelse för en liten kund. Förslaget innehåller därför både områdets D och ett relationvärde samt avvikelse i ton. Faktisk och okänd del hålls isär så att oförändrade totalsummor inte döljer utjämning av Coverage. Räknefallen är reproducerbara och kontrollerar bland annat att enbart omindelning inte förbättrar D. Formlerna är genomräknade; val av referens och reglernas praktiska användbarhet behöver bedömas i sak. Inget av förslagen är antaget. Denna tråd kan nu fokusera på D:s konstruktion, referensval och redovisning till kunden. Nodgränsen och innehållet i datautbytet diskuteras i MF-004, tråd #62 och MF-005, tråd #63. |
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Hur kan MASSIV+ visa att en mellanhand bevarar skillnader i sina inkommande utsläppsuppgifter när dessa fördelas vidare till kunderna?
En grossist köper 100 ton material från vart och ett av två leverantörsflöden. Det första bär 40 ton CO2e i uppströmsbelopp och det andra 200 ton. Grossistens egna utsläpp är 5 ton. Kund X får materialet från det första flödet och kund Y från det andra. Talen är påhittade.
Med en gemensam nod och massa som allokeringsnyckel får båda kunderna 122,5 ton. Hålls de inkommande skillnaderna isär får X 42,5 och Y 202,5. Totalen är densamma. Skillnaden är vilken information om inköpen som bevaras i fördelningen till kunderna.
Omarbetat MF-002, 2026-10-30
MF-002 behåller avsikten att införa D vid sidan av Coverage. Coverage visar den faktiska andelen; D föreslås visa avvikelsen mot en dokumenterad referensfördelning av samma inkommande underlag.
Det nya förslaget beräknar D över samma verksamhet, period, underlag, referens och mottagarrelationer. Därmed ska enbart ändrad intern nodindelning inte förbättra måttet. Faktisk och okänd del jämförs separat. Ett samlat värde kompletteras med ett relationvärde för kundens del, tilldelat belopp, referensbelopp och skillnad i ton.
Referensen väljs efter det redovisade syftet och den relevanta information som finns eller rimligen kan tas fram. Den ska motiveras, fastställas före jämförelsen och versionshanteras. Ett högt D mot en fysisk referens gör inte i sig en annan tillåten allokeringsnyckel felaktig. Saknas en motiverad referens anges ej beräknat.
Läs det omarbetade MF-002 med konkret standardtext, räknefall och beräkningsbilaga. Omarbetningen ligger på publiceringsgrenen och inväntar granskning före webbpublicering. Webbsidan uppdateras när ändringen har granskats och införts.
Den tidigare nodregeln om motpart, fysisk enhet och gods behandlas separat i MF-004, tråd #62. Informationsinnehållet mellan organisationer behandlas i MF-005, tråd #63. Samtliga är utkast och ingår inte i standarden.
Ger det föreslagna D en användbar komplettering till Coverage, både för verksamheten och kunden? Är reglerna för referensval och mottagarnivå tillräckliga för en begriplig jämförelse?
Trådens inledning har skrivits på svenska och uppdaterats för det omarbetade förslaget. Tidigare kommentarer finns kvar som diskussionshistorik och avser delvis den tidigare konstruktionen.
All reactions