Technical Definition

37 views
Skip to first unread message

Sumesh K.C.

unread,
Jan 5, 2025, 10:55:03 PMJan 5
to fiboa
Hi Fiboa Team,

Happy New Year!

I would like to congratulate you on all your efforts.

I would like to know if any document was produced/published that provided the technical definition of some of the terminologies: farm, farm boundary, field, field boundary, cropland, cropland boundary, etc.

There is one question I would like to ask:

  1. When delineating the field boundaries on remote sensing images using the DL approach,  are we delineating the field or cropland boundaries?

Thank you.

Best Regards,
Sumesh

Matthias Mohr

unread,
Jan 21, 2025, 12:16:23 PMJan 21
to Sumesh K.C., fiboa
Hi Sumesh,

thanks for reaching out. fiboa is still working on definitions and clarifications.

For fields of the World, the paper says that the predicted classes are boundary, interior, and background - so both.

Best,
Matthias


------ Originalnachricht ------
Von "Sumesh K.C." <kcsume...@gmail.com>
An "fiboa" <fi...@tgengine.org>
Datum 06.01.2025 04:55:03
Betreff Technical Definition

--
You received this message because you are subscribed to the Google Groups "fiboa" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fiboa+un...@tgengine.org.
To view this discussion visit https://groups.google.com/a/tgengine.org/d/msgid/fiboa/f55a4075-2133-4c93-bc6e-42ccce475d46n%40tgengine.org.

Nuno Queiroz de Mesquita Cesar de Sa

unread,
Mar 25, 2025, 4:56:07 AMMar 25
to fiboa, Matthias Mohr, fiboa, Sumesh K.C.
Hi all, 

I just joined the group because im curious to any data on fiboa in Africa and especially we will be doing some fieldwork there. In our case we are adapting the definitions from FAO: https://openknowledge.fao.org/server/api/core/bitstreams/03d673ed-6360-47b6-9ac0-e8ca1eac35e0/content ( SDG indicator 2.4.1 (Farm Survey Module). This figure below kind of summarizes it well:

  

Holding or Farm (FAO Definition): - The holding or farm is all the land being utilized in full or in part for agricultural purposes which is located in a single parish. The holding or farm may consist of one parcel of land or may be in several parcels


Parcel (FAO Definition): Any piece of land, of one land tenure type, entirely surrounded by other lands, water, road, forest or other features not forming part of the holding or forming part of the holding under a different land tenure type.

Field & Plot (FAO Definition, pp 45): A field is a piece of land in a parcel separated from the rest of the parcel by easily recognizable demarcation lines, such as paths, cadastral boundaries, fences, waterways or hedges. A field may consist of one or more plots, where a plot is a part

 or whole of a field on which a specific crop or crop mixture is cultivated, or which is fallow or waiting to be planted


Do you believe these definitions are incomplete/incorrect? They seem quite relevant in our study areas.
Nuno

Sumesh K.C.

unread,
Mar 25, 2025, 10:56:32 PMMar 25
to fiboa, nuno.c...@wwf.de, matt...@mohr.ws, fiboa, Sumesh K.C.
Hi Nuno,

Thanks for the information. I believe these definition are quite applicable.

In the research papers related with field boundary delineation using deep learning, the term "field" is used instead of "plot".  The definition of plot is frequently adopted to define field. A field is defined as a piece of land containing a single crop (or crop mixture) separated by roads, canals, fences or natural barriers. Field boundaries are defined as boundaries where a change in crop type, crop mixture or farm management practice takes place or where two similar cultivations are separated by a natural disruption in the landscape.

Any opinion on this?

Best Regards,
Sumesh

Andy Jenkinson

unread,
Mar 26, 2025, 4:30:13 AMMar 26
to Sumesh K.C., fiboa, nuno.c...@wwf.de, matt...@mohr.ws, fiboa, Sumesh K.C.
This is likely because it is not globally universal, in some countries it is not usual to have multiple separate crops per field and in others it is. However if those papers define a field as being one contiguous crop it is generally an oversimplification - they will tend to have a bias towards defining a field as something visible they can actually detect rather than how it is used​. Algorithms tend to identify separate objects by the fact they have distinct crops in them and/or are separated by boundary features (fences etc). If you asked a farmer you might get a different answer (and indeed this is why many of them think remote sensing is flawed).

Even in countries where a field only has one crop in it, it also does not cater for zones within fields used for field trials. This might not be something that comes immediately to mind if you ask such a farmer what a field is, but you’d still have to design around it if you were making software in the real world.

Then there's the fact that different systems or personas use the word field for different things. For example a government might define it exclusively as something with physical boundaries (paths, fences, hedges) because that is easier for their subsidy system to deal with, whereas a farmer might define it more in relation to each unique semi-permanent cropped area (of which there might be 2 or 3 per physical field).

Part of the difficulty of this domain is that there are at least four separate concepts loaded into the terminology, which don't necessarily align:
  • Land use rights: land owned by or rented to someone
  • Physical boundaries: permanent features that unambiguously segregate land
  • Cover: commonality of what is physically growing (or absence thereof)
  • Operations: units of common treatment - broken down further from a single crop field, each operation may still have a separate boundary (eg planting vs spraying)
Personally I think the FAOs definitions are fine (probably they omit the last category but that is understandable , they tend to only exist in digital form when generated by machinery), they just won't be the same words used by everyone:
  • a parcel could really be any contiguous area of land, but is most commonly used referring to ownership or usage by different people 
  • a plot again could also be used for any area of land (eg an entire farm when it is being sold) but in a context where farm and parcel exist as separate definitions it makes sense to use it as any area marked out for separate usage from the surrounding  area (ie defined by usage not ownership or physical 
  • field is used to mean many different things in different parts of the world but so long as it's defined distinctly from the other concepts it's fine

Ultimately FIBOA does not attempt to really deal with the fact that there are all these differences between boundaries, each dataset will have a bias towards what it includes and dealing with that remains a user problem. Ultimately that is why we created Global FieldID in the first place and designed it the way we did - every different boundary can have its own unique ID but it doesn't mean they are spatially distinct. For that you need to recognise that a field is not spatially defined, it is conceptual - it's good to have a common unit called a field we can all refer to, but it is not possible to force a single definition of what it is on all farmers. Hence a FieldID's only rule is that it cannot overlap another field.

Kind regards
Andy 

From: Sumesh K.C. <kcsume...@gmail.com>
Sent: Wednesday, March 26, 2025 2:56:32 AM
To: fiboa <fi...@tgengine.org>
Cc: nuno.c...@wwf.de <nuno.c...@wwf.de>; matt...@mohr.ws <matt...@mohr.ws>; fiboa <fi...@tgengine.org>; Sumesh K.C. <kcsume...@gmail.com>
Subject: Re: Technical Definition
 

Nuno Queiroz de Mesquita Cesar de Sa

unread,
Mar 26, 2025, 5:44:09 AMMar 26
to fiboa, andy.je...@varda.ag, Nuno Queiroz de Mesquita Cesar de Sa, matt...@mohr.ws, fiboa, kcsume...@gmail.com
Hi,

Great input and its not an easy discussion at all!

But somehow it might not be so clear in the field (field here is not the agriculture definition but on the ground). For example, in this case it might be visible if the "stars align". 


example.jpg

The separation between field is just a slight increase in distance between the two crops (of the same species but planted at different periods or perhaps just different management) which are owned (perhaps) by two different people. So the "separated by a natural disruption in the landscape" doesn't work in this case. 

Also notice that often it can just happen both crops are managed in the same way by two different people and there is no clear separation in the field - or at least - from a RS perspective. The crops are just slightly separated, few centimetres well below the the GSD of any satellite and probably you cant even distinguish it by UAV. Just by coming on the field and seeing it yourself. 


All that said, real talk, if it s possible to do this:

example_2.jpg

Its a great achievement since these are basic "units" in a sense. This is in Zimbabwe but If we look further closer to the equator it gets messier with smaller fields of course. A great example what this team from Clark university achieved is pretty impressive: https://agroimpacts.github.io/mapping-africa/  

Also by the way, the model described in: https://github.com/fieldsoftheworld/ftw-baselines is excellent and easy to use! Thank you for that effort! I'll be lurking for updates. 

Nuno

Andy Jenkinson

unread,
Mar 26, 2025, 7:29:16 AMMar 26
to Nuno Queiroz de Mesquita Cesar de Sa, fiboa, Nuno Queiroz de Mesquita Cesar de Sa, matt...@mohr.ws, fiboa, kcsume...@gmail.com
Yes this is what I mean by the different orthogonal concepts that the definition needs to capture: ownership is a different delineation than visual distinction, and only sometimes align. Any definition used by an isolated use case is always going to be limited, and earth observation datasets are a good example of that - remote sensing “field boundary” datasets inherently define a field as something that is visible from an image (or series of images). It is impossible for this input alone to be sufficient because you can’t see who owns land from space. Only people “in the field” (or tools with access to their so-called “ground truth” input) can do that.

To do this you must *merge* datasets, and that is what GFID can be used to do - mass generation using remote sensing to give breadth, and human or in-field machinery to infill and correct. Or in other regions/cases (eg agroforestry or very small and fluid “fields” where remote sensing is not very useful) only crowdsourcing will work.

Sent from Outlook for iOS

From: 'Nuno Queiroz de Mesquita Cesar de Sa' via fiboa <fi...@tgengine.org>
Sent: Wednesday, March 26, 2025 9:44:09 AM
To: fiboa <fi...@tgengine.org>
Cc: Andy Jenkinson <andy.je...@varda.ag>; Nuno Queiroz de Mesquita Cesar de Sa <nuno.c...@wwf.de>; matt...@mohr.ws <matt...@mohr.ws>; fiboa <fi...@tgengine.org>; kcsume...@gmail.com <kcsume...@gmail.com>
Subject: Re: Technical Definition
 
Reply all
Reply to author
Forward
0 new messages