Download Summary of comments received on concept

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Earthworm wikipedia , lookup

Plant nutrition wikipedia , lookup

Soil horizon wikipedia , lookup

Surface runoff wikipedia , lookup

Soil erosion wikipedia , lookup

Canadian system of soil classification wikipedia , lookup

SahysMod wikipedia , lookup

Terra preta wikipedia , lookup

Soil respiration wikipedia , lookup

Crop rotation wikipedia , lookup

Cover crop wikipedia , lookup

Soil salinity control wikipedia , lookup

Soil compaction (agriculture) wikipedia , lookup

No-till farming wikipedia , lookup

Soil food web wikipedia , lookup

Tillage wikipedia , lookup

Sustainable agriculture wikipedia , lookup

Soil microbiology wikipedia , lookup

Pedosphere wikipedia , lookup

Soil contamination wikipedia , lookup

Transcript
Summary of comments received on concept note for global soil reference library
Bob MacMillan, ISRIC, March 1, 2011
From: Macmillan, Bob
Sent: 01 March 2011 17:39
To: Walsh Markus; Keith Shepherd
Cc: Hengl, Tom; Batjes, Niels; Hartemink, Alfred; Bindraban, Prem
Subject: First cut at a concept note for a global soil proeprty reference and conversion library
Hi everyone,
Coming out of the AfSIS review and attendant discussions, two ideas came up for possible concept notes
that we might want to follow up on.
One concept was for a global soil property reference and conversion library. This idea arose from the
reading I did about pedotransfer functions used to convert soil property values reported according to one
method of analysis into equivalent values in another method of analysis. This got me thinking about
whether it might be possible to have a single conversion system based on the similarity approach
proposed by Nemes and Jagtap that we could use to convert all our data using a single consistent
methodology. I discussed this idea with Keith Shepherd and he was surprisingly positive about it. He
expanded the concept to include the idea of having all samples scanned using NIR and MIR spectral
analysis to create a global spectral reference library in the process. The spectral data could end up being
the future reference data and mechanism for converting any value by any method into an equivalent value
in a different method.
This is my first attempt to write up the idea as I believe that we discussed it among Keith, Markus, Achim,
Christian and myself. I am sure I have a lot wrong but it is a start.
If this idea meets with your approval, I suggest we immediately expand the discussion to include Alex
McBratney and Budi as they have a long history of involvement in pedotransfer functions and have
proposed the SINFER framework for developing universal pedotransfer functions.
I think there may well be considerable opportunity for synergy (or duplication) between this idea and the
current ISRIC initiative to collect replacement soil samples for the lost soil samples. I would like to think
that we could combine the two activities to get more out of both. Keith has suggested that we try to get
all analysis done at the same lab. That really boils down to trying to get NRCS to do all analysis at Lincoln.
This is already the idea for the ISRIC soil sample recovery program. We should try to extend this to the
global soil reference library.
We will need to look for some significant funding for this. I will let Keith and Markus estimate the likely
costs. Christian encouraged us to develop this concept note but I am not sure that he envisages Gates
providing funding. Still, once we have something we can agree on, we should try to promote the concept
to potential funding agencies.
The second concept that we discussed in Nairobi was to make use of some web crawling services that the
Gates Foundation has used to develop an application to crawl the web to extract and format data of
interest to the project. The initial idea was just to find and format legacy soil profile data. Markus and
Christian then wanted to expand this concept to crawl the web to extract, synthesize and format
knowledge about soil management practices and best management recommendations by crop and agroecological region in Africa. I will write up the idea of the soil legacy data web crawler starting tomorrow. I
hope that someone else will take on the larger task of the legacy soil management web crawler
application.
I hope this gets people thinking about the concept and that you get back to me with your suggestions for
improvement.
I look forward to hearing back from you on how to best move this idea forward.
Bob
Dr. R. A. (Bob) MacMillan
Science Coordinator
GlobalSoilMap.net
ISRIC - World Soil Information
PO box 353, 6700 AJ, Wageningen, The Netherlands
Visiting address: Droevendaalsesteeg 3 (Building 101)
Phone: +31 (0)317 483756 (direct), +31 (0)317 483735 (secr.)
Skype: dr.bob.macmillan
Email: [email protected] ; [email protected]
Web: www.isric.org, www.globalsoilmap.net, www.africasoils.net
Alfred Hartemink, ISRIC, March 2, 2011
From: Hartemink, Alfred
Sent: woensdag 2 maart 2011 8:03
To: Macmillan, Bob
Cc: Hengl, Tom; Batjes, Niels; Bindraban, Prem
Subject: RE: First cut at a concept note for a global soil proeprty reference and conversion library
Bob,
1. On the reference library there is of course ISIS and I expect much of that to come out of the sentinel sites did Keith allude to that? The conversion library is good but for several attributes that is different for different
soils and locations so there is probably not a single methodology (Keith's favourite) but I agree we should
strive for unified systems. The specs include some of it and let's start from there. I know that NRCS and
Raphael are working on exchange libraries for NIR and eventually MIR. We can discuss this with Jon and Mike
when they visit us later this month
2. I would be curious to see what a web crawling service might do and it is absolutely worth investigating;
Johan has some experience with web searching legacy soil data and with Google scholar some unusual things
can be found. Have they asked for a concept note or pre-proposal?
Alfred
Bob MacMillan, ISRIC, March 1, 2011
1. The idea of the similarity approach is really nice because it can accommodate the need for different
conversions for different soils, geographic locations and environmental settings. You find the 10 most
similar soils. Similarity can be defined in terms of location, environmental setting, soil classification or just
simply the soil property to be converted. Tom and Keith tell me that, in essence, the local similarity
procedure is really just a slightly modified form of a local regression equation. So we can have local
regression equations all developed in a similar way and from a single standard reference library. That way
it is a unified system if not exactly a single equation.
I think that this idea perfectly fits my role as Science Coordinator and ISRICs role as a global soil
resource. It gives us a chance to develop a really useful tool both for the project and for soil science in
general. If it works, we will have established a single unified standard for converting between values
produced using different lab methods. This is exactly the sort of coordination that the Science Coordinator
should be doing and is exactly the sort of service that ISRIC should be offering to the world.
2. I agree it is worth a shot to look at. If it works we can maybe add tens of thousands of profiles to the
database that we might never otherwise find. It is supplementary to and complementary to Johan's
manual efforts. Christian explicitly asked for a concept note and he wants to follow up on this with his
contacts. Markus is also keen to see what can be done.
Bob
Tom Hengl, ISRIC, March 2, 2011
On Mar 2, 2011, at 2:13 PM, Tomislav Hengl wrote:
>
> Hi all,
>
> It's good to see that Bob is still live'n'kicking (even with one foot
> in the air :)
>
> Here are contributions to the whole story:
>
> 1. I think it is excellent initiative and I think we have a chance to get that money. I
am willing to make an R package (e.g. such as
[http://casoilresource.lawr.ucdavis.edu/drupal/node/916]) and a data portal to host all
info.
>
> 2. Even if we do not get money, we could start building functions and a framework for a
library. BTW, ISRIC got some money to collect 227 new soil profiles spread around
different countries. We could simply use those soil samples to compare the most frequent
lab analysis techniques.
>
> 3. I would change the name to "Soil Lab data Translation Library (SoilLabTL)".
>
> 4. The translation should have either form (a):
>
> soil.var.met_X <- fun( soil.var.met_A, soil.avar.met_X,
> soil.class.met_X, ..., LAT, LON )
>
> where "soil.var.met.X" is the output result of conversion, "fun" is the mathematical
model (formula and coefficients) used to derive values and uncertainty, "soil.var.met_A"
is the input value of the same variable estimated some arbitrary method,
"soil.avar.met_X" is the auxiliary attribute that helps converting the original values
more accurately, "soil.class.met_X" is the soil type (or any other class) and LAT LON are
geographical coordinates.
>
> or the form (b; pedo-transfer rule):
>
> soil.var.met_X <- fun( soil.avar.met_X, soil.class.met_X, ..., LAT,
> LON )
>
> Hence for each translation function you should contain:
>
> - mathematical model (type of multiple liner/non-linear regression);
> - list of fitted coefficients used in the model;
> - list of required inputs to run the conversion;
>
> The same models could also be used to convert spectral bands (values) to actual soil
properties. Note that each conversion would also give an estimate of propagated
uncertainty (this would be a statistical estimate).
>
> If someone would provide me with a spatially distributed calibration point data, then I
would probably be able to map the model coefficients over the whole globe (so that when
you query some database you would get location-adjusted translation coefficients).
>
> These thing can be put to a live database, which can then be loaded any time someone
loads the SoilLabTL package.
>
> My dream is that, ISRIC, CIAT and similar organizations would work toward crowd
sourcing soil data collection. We could finish with making a toolkit for soil data
collection (it would probably have an auger, a soil description manual, a spectral
scanner, pH meter, EC meter and few other quick-n-dirty instruments) that could be
affordable to anyone. We could even package such toolkit and then ship it to who ever
registers on our data portal. How much would a simple auger + spectral scanner + pH meter
+ EC meter + ... cost?
>
> cheers,
>
> T. Hengl
> Senior research @ ISRIC - World Soil Information
> Mail: ISRIC - World Soil Information P.O. Box 353, 6700 AJ Wageningen,
> The Netherlands
> Location: Room C.002, Droevendaalsesteeg 3 (GAIA building), 6708 WB
> Wageningen, The Netherlands
> Tel: +31(0)317-484199
> Url: http://www.wewur.wur.nl/popups/vcard.aspx?id=HENGL001
Hannes Reuter, ISRIC, March 2, 2011
Additionally scientific literature will be analyzed to identify already investigated conversion parameters thereby
focusing on different continents, countries and regions. By doing so, also uncertainties of the different methods
can be specified.
A preferred way will be the storage of the conversion functions in a versioned database. Initially, for each
method four different conversion levels are foreseen: a global function which is used there as a backup, a
continental function, a country wide function and at the lowest level a regional function. These different rules
will make sure that any given value at any given point can be converted with a specified uncertainty. Each
function can be derived from the analyzed soil samples, or from literature, therefore allowing for sampling and
analysis and development of the IT structure at the same time.
Niels Batjes, ISRIC, March 2, 2011
From: Batjes, Niels
Sent: woensdag 2 maart 2011 11:58
To: Macmillan, Bob
Subject: RE: First cut at a concept note for a global soil proeprty reference and conversion library
Bob,
This proposal will help to address the issue of limited comparability of soil analytical data worldwide. This
pressing issue has been on the soil-agenda since the first ISSS world soil congresses.
-
Soil samples could also be obtained from/through the EU-JRC Lucas soil programme
[http://eusoils.jrc.ec.europa.eu/projects/Lucas/SoilArchive.cfm ] and others.
-
ISRIC’s archived soil samples link to ISIS, not to WISE as currently written (p. 3).
-
Page 3: LL, DUL, SAT in full please
-
Should indeed strive for analyses in a single lab (NRCS Lincoln?), for uniformity. Otherwise, also
need for more elaborate QA/QC procedures including so-called “round-robin” rounds (blind
samples) etc.
-
The proposed similarity approach sounds very promising: consistent yet versatile, as similarity may
be defined in various ways. Ultimately, one should would strive for a larger number of
samples/profiles to underpin the method (say >30).
-
Potential end-users: also include the “broader scientific community”.
The web-crawling concept sounds interesting and worthwhile investigating; interested to see how much
information can be extracted automatically “from within” the disparate sources.
Niels
Keith Shepherd, ICRAF, March 3, 2011
Thanks for your inputs below and Yes, we were all relieved and delighted to see Bob back
in action!
We have been interested in handheld spectral applications for field data collection for
some time (a recommendation of our 2007 review paper - Shepherd KD and Walsh MG (2007)
Infrared spectroscopy—enabling an evidence-based diagnostic surveillance approach to
agricultural and environmental management in developing countries. Journal of Near
Infrared Spectroscopy 15: 1-19).
I have been working on several companies to develop a low cost device but at present
handhelds remain too expensive and importantly lack sufficient stability (over time and
between instruments of the same make) to be viable. It is only be matter of time though.
Perkin Elmer and Bruker both have my design specs and are probably working on it in
secret. In the meantime we can be setting up the calibration and interpretation libraries
and workflows using lab-based instruments and so we can make rapid progress once the
technology is there. In any case we are discussing ways of scaling up data collection
considerably even using lab-based instruments, which we know can give reproducible
results. Incidentally, even pH meters are not straightforward and easily get contaminated
and you get serious lab to lab differences!
I am curious to know what basis you would use for the location-adjusted translation
coefficients for spectral calibrations - are you thinking of kriging approaches or use of
some covariate layers? That sounds very interesting.
With thanks and all the best,
Keith
Keith Shepherd, ICRAF, March 10, 2011
Hi Bob,
It took longer for me to get back on this than I had hoped for. Kindly see some suggestions attached. I think it is
best to get the preparatory phase funded first as there are currently too many uncertainties to write the full
proposal.
I think you are the right person to champion this and then recruit a scientist to lead the implementation and we
can all pitch in.
I will contact NRCS to put out some feelers on the feasibility of lab analyses.
See what you think.
All the best,
Keith
Keith Shepherd, ICRAF, March 11, 2011
Hi Bob,
Thank you very much for your positive feedback. There a lot of emails floating around - this one is on the lab
analyses.
Our lab is well placed to do spectral based measurements (visible-near infrared with ASD spectrometer, FTNIR, and FT-MIR, x-ray diffraction, total x-ray fluorescence, and laser particle size analysis). I believe that we
are best placed to do that part. However we cannot handle the wet chem.
We need a lab specializing in wet chemistry that could run a variety of soil tests and is equipped with ICP-MS
etc. Hopefully this could all be done in one lab but the fewer labs the better and you can't ship foreign soils form
one lab to another in the US. I am also in collaboration with BLGG and am also approaching them for their
capability and interest. Most labs tend to run one soil test method and not another. This all has to be worked out.
I will now get some prior feel from Lincoln and BLGG.
Al the best,
Keith
Bob MacMillan, ISRIC, March 10, 2011
On Mar 10, 2011, at 4:21 PM, Macmillan, Bob wrote:
Keith,
I have not yet looked at your comments and suggestions but will do so today.
I was talking to Alfred earlier today about this, because we were talking about things that we might bring
up for discussion when Jon Hempel and Mike Golden visit ISRIC from NRCS at the end of this month.
I asked Alfred if I could bring this concept up and try floating the idea that the Lincoln lab might do all the
analyses. Alfred wanted to know why you did not feel that the ICRAF lab in Nairobi might not be a good
place to do all the analyses. Perhaps you can provide your reasoning to Alfred.
I intend to champion this and try to find support and funding for it. It would be a major contribution that
ISRIC could make to global soil science as well as to the GlobalSoilMap.net project. It is a big ambitious
project with a clear focus and easily described benefits.
Will look at your comments now.
Bob
Markus Walsh, CIESIN, March 11, 2011
On Mar 11, 2011, at 12:05 PM, Markus Walsh wrote:
Hey guys,
So, in my opinion, Bob, Tom, Keith and others have put some really good thinking behind this! The main
product would be a database for translating and harmonizing various soil measurements + all the associated
pedotransfer functions, web interfaces ... .
I personally think that this sort of thing could / & certainly should be a big "seller" grant-wise.
Imagine just +100 harmonized profiles per country wi the GDSM network analyzed well sampled, analyzed
using the same ref methods (pref. at a ref lab), georeferenced, and with the attendant spectra and other
covariates. This does not seem like a huge unreasonable undertaking and is certainly one that would benefit a
wide range of users. The database could / should also be built up further over time as new measurements and
locations become available.
So I'd say run with it, print up a glossy brochure, run some examples from legacy data to illustrate methods and
algorithms, and ensure that it is a global effort.
I also think that you are aiming much too low by confining this only to legacy data and interpretations thereof.
This effort should really focus on a new data collection, now! 100 properly sampled profiles w e.g. 1 ha block
support per country is not that much to ask for given that NASA just dumped the Glory Satellite into the Pacific
Ocean (see: http://www.nasa.gov/home/hqnews/2011/mar/HQ_11-050_N0_Glory.html).
We could do the ground data collection in Africa for this rapidly, on sites that are actually representative in 1
year. In Europe, US, Brazil, Oz ... this should take much less time, provided that you can handle all the politics.
Just my opinion. Very best, M
Keith Shepherd, ICRAF, March 11, 2011
Thanks Markus,
I agree that new samples would be preferable - then we know they are accurately georeferenced and with
current soil conditions. I do not know if we could get the collaboration from the nodes to do a new sampling
that would represent the widest possible range of conditions but certainly worth probing and would be fantastic
if doable.
Thks,
Keith
Bob MacMillan, ISRIC, March 11, 2011
Thanks Markus,
Appreciate your positive response and support. Like the idea of printing a glossy brochure and also
including some actual worked examples so that potential supporters and contributors could see what were
aiming at.
Not to be picky but, while we were suggesting that we work initially with currently available archived soil
samples, the concept note did certainly include a suggestion that it also plan for the inclusion of data from
new, purposely sampled, reference sites. Your suggestion just expands this idea.
Let this reply hanging all day so forgot where I was going next with it. Will just send it now.
Bob
Alfred Hartemink, ISRIC, March 11, 2011
On Fri, Mar 11, 2011 at 10:45 AM, Hartemink, Alfred <[email protected]> wrote:
Prem,
The interpretation and soil functions (including what you mention below) is clearly envisioned as being part of
GlobalSoilMap.net and AfSIS. In the digital soil mapping literature that is labeled as "digital soil assessment"
and a major topic of discussion for the Ispra meeting in June.
Before we can make interpretative maps of any kind but we need to have the base soil maps and that is currently
the priority. Given the fact that we have vast amount of data (yes, it is mixed and some is old) for most parts of
the world there are no funds to collect new data. Once we have the first maps there is no limit to its use and
interpretation. As these maps will also include uncertainties (high in many areas) these first maps will open up
possibilities for collecting new data. We all agree that collecting new data is essential but there is hardly any
donor willing to pay for data collection - they all want maps and information systems that tell us where to put
fertilisers or where crop production can be enhanced - ideas are needed who will pay for collating and updating
the underlying soil information. Of course I hope that funds will become available for this proposal just like I
hope that our projects will revitalise and modernise soil survey and data collection. Perhaps that crop modelling
community is not only willing to criticise the available soil information but also to help finding the ways to
compile the existing information and to collect new data.
At last, your notion that "for many users of soil data there is little value in much information that comes out of
soil science" is not helpful in all this.
Alfred
Prem Bindraban, ISRIC, March 11, 2011
________________________________________
From: Bindraban, Prem
Sent: 11 March 2011 06:18
To: Macmillan, Bob; 'Shepherd Keith'; Hengl, Tom
Cc: Walsh Markus; Batjes, Niels; Hartemink, Alfred
Subject: RE: First cut at a concept note for a global soil proeprty reference and conversion library
Dear Bob and others,
This is a great initiative. Indeed for many users of soil data there is little value in much information that comes
out of soil science, e.g. in relation to modelling nutrient and water uptake by plants (own experience also).
Often data analysed with different methods are simply thrown on one heap and considered to represent the same
thing (pH is pH), simple “transfers” used, “guestimates” (something between a guess and an estimate), etc.
I suggest to take this idea one level further. Include crop-response curves for instance. Users of soil information
ask for more than a “soil type” or “property value”. E.g. I got this back from Shamie Zingore (of Int Plant
Nutrition Institute) upon my question what soil information he would need: “Now you provide maps with soil
types, but add information to show the probability where we can expect to get certain nutrient problems
(deficiencies etc) and response. These maps should be available at the district level.” You already indicate that
pedo-transfer functions need to be location specific – that could be made more generic (Nemes & Jagtap). Crop
responses are location specific as well; process based crop/soil models could be used to try to develop more
generic methods, e.g. by removing impact of weather and crop varieties etc.
Thinking a little ahead; the link with plant growth captures likely the most important users – not just plants, but
through process crop-soil models also to other themes like SOM and climate change, soil water use and food
production, fertilization and soil productivity (just think of the impossible measures on phosphorus!),
biodiversity as related to standing biomass based on soil properties, etc. I see very many linkages. It would be
good to include the crop users from the initial stages as a separate, but closely interacting, work package of the
project.
Yes – let us join our efforts and resources, linking our small sampling program to larger ones like Afsis. In India
they have been taking many samples (of the upper 20-30 cm only though) to look into micro nutrients (A.K.
Singh will present this at our seminar of the 21st March). Try to get hold of him also.
I recommend to include Johan Bouma in this discussion. I had a talk with him yesterday about some strategic
IRIC issues (not this in particular) and in-between the lines is sensed (maybe incorrectly) that he seems not so
pleased on the pedotransfer approach and feels that we lack hard data (the real thing). His (former) group was
much involved in the development of pedotransfer functions etc.
If crop responses would come in, then Hein ten Berge (colleague from Agro-systemsresearch could come in).
He is really good in linking soil properties to crop responses (and is a soil scientist by the way; both process
hydrological model (Darcy flow etc) and nutrient). Also – Frits van Evert from the same department is looking
into quantitative approaches to link soils to crops (and even production systems) as well.
Would Peter Finke (Belgium) not be a good biometrician to include – but you guys know better.
Great initiative – keep me informed and lets get any institutional arrangement required to get this right in place
also.
Best
Prem
Prem Bindraban
Director ISRIC - World Soil Information
www.isric.org
Leader International Research
AgrosystemsResearch
Wageningen UR, The Netherlands
Tel: +31-317-480881 / +31-6-22990783
Alfred Hartemink, ISRIC, March 11, 2011
Prem,
Jon has offered to fully analyze the soil samples of the monolith recovery project. In September he mentioned that if it
were a few hundred samples he could silently squeeze them in their routine stream of samples. We should discuss that
with them and Vincent later this month.
This newly proposed project concerns very large number of samples and it wouldn't be wise if we would over-ask from
NRCS at the expense of our own project.
BLGG is not an option as they are fully commercial and do not use all internationally standardised methods. I am puzzled
by Keith's comments about the ICRAF lab - it would be an obvious choice,
Bob MacMillan, ISRIC, March 11, 2011
Prem,
I had a quick chat about your response with Tom at breakfast.
His immediate thought was that ISRIC could host a number of inter-related databases that
collectively offered a number of useful functionalities. We started here with a concept
for a database that offers 2 closely inter-related functionalities, namely:
A) conversion of values between different analytical methods vis a soil methods
conversion library
B) relation of soil property values by many methods to soil spectral signatures to create
a global soil spectral reference library
Your suggestion represents one more area in which this initial functionality could be
extended and improved. In this case, the objective is to add another functionality of
C) global library of soil response to fertilizer application by soil type, location and
soil properties
Tom suggested that we could envisage an overall framework that could support a number of
different targeted applications. This allows us to keep our work focused on a clear and
specific target for any given application (e.g. conversion of methods) while being able
to expand and extend the framework through related parallel activities.
So, with this in mind, I would suggest that your idea for fertilizer response data is a
good one but that it be done separately, and in parallel, by a separate group with the
appropriate expertise and mandate. If we design the framework correctly, it ought to be
able to accommodate multiple applications, including ones we have not yet anticipated.
I agree with Alfred that response to fertilizers by different soils and soil properties
is a key interest of AfSIS and that AfSIS will definitely be wanting to prepare materials
that relate soil properties to response to fertilizers. So, AfSIS, in particular, may
have a strong interest in following up on your suggestion.
My personal position is that the concept note I prepared with Keith is already broad
enough and difficult enough to achieve that expanding it to include fertilizer response
data has the potential to make it too unweildy to manage properly and succeed. I suggest
that we think about how we can make the fertility response data a linked extension to the
proposed project but have it devised, funded and run separately.
I would like to add that activities like coordinating the creation of a global library
for converting soil property values and supporting the widespread application of soil
spectral methods are perfect examples of the kinds of activities that ISRIC is especially
well positioned to do to support global soil science. This kind of service to the global
soil science community ought to be what ISRIC is all about.
You have asked us in the past to think of how we can obtain funding for ISRIC. From my
point of view the best way to obtain funding for ISRIC is to come up with really good
ideas that deserve to be funded and then to go out and look for funding for them. This
makes lots more sense than identifying funding bodies and then looking to see what they
are prepared to support. This approach may get ISRIC funding but it is not as likely to
support ISRIC to do the things it should most be doing.
One final point, I think that Johan Bouma is an asset to any ISRIC activity and would be
particularly useful in the context of this current concept note proposal. In fact, if he
is willing to lead the effort, I would be really happy to have him do so. He might even
be the lead scientist that Keith has suggested be hired to lead the project
implementation.
Regards, Bob
Alfred Hartemink, ISRIC, March 11, 2011
Malawi would be the perfect case for the map and its functionalities. Once that is
published it would strengthen the case for the conversion library. Secondly, I think data
collection for the conversion library also has to serve for cross correlation of the
maps. That may be add to the selling points.
Indeed, ISRIC was setup as a soil reference centre and should lead this.
Alfred
Keith Shepherd, ICRAF, March 11, 2011
From: Keith Shepherd [mailto:[email protected]] On Behalf Of Shepherd Keith
Sent: Friday, March 11, 2011 6:55 AM
To: Ferguson, Rich - Lincoln, NE
Cc: West, Larry - Lincoln, NE; Hempel, Jon - Lincoln, NE; Monteith, Steve - Lincoln, NE; Benham, Ellis - Lincoln, NE;
Kruger, Linda - Lincoln, NE; Bouc, Linda - Lincoln, NE; Jones, Patty - Lincoln, NE
Subject: Proposal for global soil characterization
Dear Rich,
We are in discussion with ISRIC and GlobalSoilMap.net on a new proposal to run a variety of different soil test
methods for a representative set of soil profiles from all regions of the world. This would allow cross-calibration
of methods, conversion of existing data to standard method for global soil mapping, and at the same time
provide a high quality global spectral library. We are talking about perhaps up to several hundred profiles from
each of 6 regions - an ambitious project.
Currently we are conducting a feasibility study. I think our lab could do the spectral measurements on a variety
of instruments, but not the conventional wet chemistry. We want to do all the non-spectral measurements in one
lab if possible to achieve consistency. These would include the conventional suite of soul survey analyses plus
different agronomic soil test methods in common use, as well as different organic matter/carbon methods. Of
course the Lincoln lab comes first to mind in terms of having the capability, capacity and quality for such an
undertaking. We are thinking that this project would have a one-year preparation phase, followed by a 3-year
implementation during which the analyses would be done.
Step 1 is to assess your reaction in terms of Lincoln' potential interest and capacity. If completely out of the
question then alternative suggestions will be very welcome!
Step 2 would be to start defining methods and costs.
While ambitious the global value of such an undertaking is immense.
I very much look forward to hearing from you.
All the best,
Keith
Alfred Hartemink, ISRIC, March 11, 2011
Keith, I will visit BLGG with Jon Hempel and Mike Golden later this month - who is your contact there? Mine is Peter van
Erp.
Alfred
Rich Ferguson, NRCS, March 11, 2011
From: "Ferguson, Rich - Lincoln, NE" <[email protected]>
Date: March 11, 2011 8:45:28 PM GMT+03:00
To: Shepherd Keith <[email protected]>
Cc: "West, Larry - Lincoln, NE" <[email protected]>, "Hempel, Jon - Lincoln, NE"
<[email protected]>, "Monteith, Steve - Lincoln, NE"
<[email protected]>, "Benham, Ellis - Lincoln, NE"
<[email protected]>, "Kruger, Linda - Lincoln, NE"
<[email protected]>, "Bouc, Linda - Lincoln, NE" <[email protected]>,
"Jones, Patty - Lincoln, NE" <[email protected]>
Subject: RE: Proposal for global soil characterization
Keith,
The project sounds both ambitious and very worthwhile. If the timing, funding (including funding source and
mechanism), stakeholders, and overall project management were appropriate / suitable, I believe this facility would be
very interested to participate as you described.
Assuming that “several hundred” were to mean ~500, and with 6 regions, and ~5 layers per pedon, the Soil Survey
Laboratory (SSL) would provide full characterization on ~15,000 samples over a 3 year period. Depending on scope of
analyses requested, and based on a similar project underway with the Environmental Protection Agency, I would
guesstimate SSL funding at $2 million or less for the 3 year period.
The SSL is currently engaged with two large projects (a nationwide carbon assessment and a national wetlands condition
assessment) in additional to its regular soil survey workload, and is completely obligated until May, 2012, and could not
consider significant additional sample volume until then. The 1-year preparation phase you referred to is therefore in
step with current obligations.
I discussed your idea with my work partner Steve Monteith (Dr. Monteith is also supervisory scientist at the SSL), and he
is also very interested in the potential of this program. So that makes two so far.
1)
2)
3)
4)
5)
6)
7)
Before presenting your idea to upper management, Steve and I are requesting elaboration on parameters for this
proposed project. Would you please provide a very brief executive summary of this proposal, including:
Purpose
Proposed stakeholders and their respective roles
General idea of where samples would be collected, including a very brief overview of sampling strategy
General list of analyses, conventional and spectral
Deliverables and outcomes assessment
Proposed source(s) of funding
General timeline
Rich
PS How is the spectral analysis of samples we sent coming along? Hopefully no surprises!
Dr. Richard Ferguson
Chemist, Supervisor
Soil Survey Research and Laboratory
National Soil Survey Center
Natural Resources Conservation Service
Federal Bldg., MS-41
100 Centennial Mall North
Lincoln, NE 68508
(402) 437-5332
Keith Shepherd, ICRAF, March 14, 2011
Dear Bob,
The response below from Lincoln looks very promising. BLGG have also responded and are compiling a list of
their methods.
Would you like to share the proposal concept with Rich at this stage? We may not be able to cover all the
questions but could start drawing up a list of desired analyses. We may also update the proposal with answers to
some of his other questions if we can.
See what you think.
All the best,
Keith
Bob MacMillan, ISRIC, March 14, 2011
Keith,
This is a fantastic and most encouraging response. You are thanked for taking the initiative to contact SSL
and BLGG to assess whether there was any interest and capacity in these labs to take on the task.
I think we do two things to respond.
1. Share the current proposal with them as is. At the same time send the current proposal and the
questions posed by SSL around to a targeted distribution list and invite contributions.
2. Revise the initial proposal to respond to the questions from SSL and whatever feedback we get from
our distribution and produce a more concrete second version of the proposal that is more specific about
soil properties to be included, methods to be applied, purpose, stakeholders, sampling design,
deliverables/outcomes and so on.
With this kind of scientific support and technical commitment, the idea has legs and it should be possible
to find funding for it. It would be a real contribution to the soil science world and a good thing for ISRIC to
host.
Thanks again,
Bob
Bob MacMillan, ISRIC, March 14, 2011
Rich and others,
As per the discussion below, here is the original copy of the concept note Keith referred to.
You will see that our plan now is to produce a new version that is more concrete and responds to the
questions you posed.
We will distribute the original concept note and your questions to a targeted group and invite contributions
for improvement. Please feel free to add you own. It really is just an open idea right now and can grow
with the input and expertise of others.
The potential for involvement and support from Lincoln makes this a feasible project to promote with a
reasonable chance of being supported and realized. I would be very proud to see something like this get
organized and implemented.
Bob MacMillan
Dr. R. A. (Bob) MacMillan
Rich Ferguson, NRCS, March 16, 2011
From: Ferguson, Rich - Lincoln, NE [mailto:[email protected]]
Sent: woensdag 16 maart 2011 23:15
To: Macmillan, Bob
Cc: Keith Shepherd; West, Larry - Lincoln, NE; Hempel, Jon - Lincoln, NE; Monteith, Steve - Lincoln, NE; Benham, Ellis Lincoln, NE; Kruger, Linda - Lincoln, NE; Bouc, Linda - Lincoln, NE; Jones, Patty - Lincoln, NE; Bindraban, Prem;
Hartemink, Alfred; Markus Walsh; Batjes, Niels; Hengl, Tom
Subject: RE: Proposal for global soil characterization
Bob,
The feedback I have received is favorable and we are eager examine refinements that address our questions.
Question: What are the current concepts on funding?
Rich
Bob MacMillan, ISRIC, March 16, 2011
Rich,
I will take on responsibility for trying to find funding. Any help or suggestions would, however, be
appreciated.
We discussed this idea with Christian Witt from the Gates Foundation in Kenya. He did not make any
commitment to fund it but was willing to help us to identify potential sponsors. This is also something that
I will try to push for funding from sources that might typically fund ISRIC including Dutch government and
European EU funding. This could and should be a core service offered to the international soils community
by ISRIC.
I have your questions to guide me in preparing the next version of the note. I will make use of people
here to make the note far more specific with respect to soil properties to include and methods for each
property. We will prepare a table or database schema to illustrate the proposed contents of the dataabse.
I will need to look to lab experts such as youiself and Keith to guide me in identification of soil variables
and methods of analysis. So, feel free to offer suggestons even before I prepare my response. Give me
your lists of preferred properties and methods to include. It will help me.
Where we will get funding is still unclear to me. If the idea is supported by Kieth, USDA/NRCS lab,
ISRIC and others (BLGG) then we at least have the crediobility to approach funders with an idea that has
support and has the demonstrated capacity to implement the idea. It is an idea that is marketable but it
still needs to be marketed.
Bob
Rich Ferguson, NRCS, March 17, 2011
From: Ferguson, Rich - Lincoln, NE [mailto:[email protected]]
Sent: donderdag 17 maart 2011 17:18
To: Macmillan, Bob; Keith Shepherd
Cc: West, Larry - Lincoln, NE; Hempel, Jon - Lincoln, NE; Monteith, Steve - Lincoln, NE; Benham, Ellis - Lincoln, NE;
Kruger, Linda - Lincoln, NE; Bouc, Linda - Lincoln, NE; Jones, Patty - Lincoln, NE; Bindraban, Prem; Hartemink, Alfred;
Markus Walsh; Batjes, Niels; Hengl, Tom
Subject: RE: Proposal for global soil characterization
Bob, Keith,
We will assemble a list of properties and methods.
But before we do, a question for you:
In addition to proposing the measurement of properties for which a viable spectral model is anticipated, should it also
be proposed to measure properties for which a viable spectral model is not anticipated either because of too low a
concentration and/or because of lack of connection to spectrally significant, more dominant property; e.g. trace
elements such as Silver and Selenium? Trace elements/minor properties might be of interest in their own right under
rubrics such as “background levels” or “soil survey”, even if no spectral dividend is anticipated.
Please share your thoughts.
Meantime, the preliminary proposal has earned the attention of upper management, and Larry West will be in touch
with comments and further discussion topics.
Rich
Bob MacMillan, ISRIC, March 17, 2011
Hi Rich,
I am not a lab expert and will have to seek the advice of experts and defer to it when it comes to issues of
content such as which properties to include and which methods to run for each property.
Philosophically, the decision about what to include or not include ought to be based on known or assumed
importance of a property and frequency of use or interest in it. We want to include at least a core list of
key soil properties and all methods of analyzing for them that have had significant use anywhere in the
world. We can afford to leave out some esoteric methods or properties that are not widely considered. It is
up to others to help us to arrive at this list.
We have a few people here at ISRIC who can contribute to this debate, but I am not really one of them. I
am more of a facilitator here looking for guidance from more knowledgeable scientists.
Hopefully Keith will be able to give you a more informed response to your main question.
I had to work on some other things the last few days but will get back to this proposal tomorrow.
We are fortunate to have our input and your willingness to take a lead role in planning and executing this
potentially important project.
Bob MacMillan
Rich Ferguson, NRCS, March 17, 2011
Thanks Bob,
We will look forward to hearing from Keith, too.
I agree that esoteric, heavily management dependent, and/or time-sensitive properties (e.g. enzyme assays, microbial
biomass, and the like) are/might be of limited appeal/viability, but that said, and where so much effort will be expended
collecting and shipping samples, one might argue that list of properties should be otherwise comprehensive for maximal
informational and interpretive value.
Another paradigm would be to measure only those properties that experience suggests are expected to return viable
pedotransfer functions of MIR or VNIR spectra. That list of properties is already quite sizeable.
The SSL would be supportive either way.
Regardless of philosophy or constraints, the SSL routinely archives samples for any future analyses that might be
requested. Incidentally, the SSL curates the largest archive of soils in North America – Alfred inspected it on his last visit
to Lincoln. It may well be the largest archive in the world.
Rich
Tom Hengl, ISRIC, March 18, 2011
Hi all,
I have been following this discussion because I work closely with Bob. I just wanted to
say that I can help you build R package (SRL) that can be maintained collaboratively. In
principle, I only need a database with soil lab data and spectral signatures and I would
produce scripts and functions to fit and serve translation modules (I think you only need
to store the fitted model parameters and then load them every time you load the package).
I do not think that we can use the existing profiles for model building because I think
that, for consistency purposes, we should scan every profile (at least this is something
I would do) so we get a 2D image of the soil (as a continuum).
The key things to decide are:
1. The (statistically) minimum number of profile to have a complete Soil Reference
Library (I would assume that it is about 250, but this could be even lower number);
2. Location of profiles (climatic zones, biomes, land use types, landform positions);
3. Selection of soil properties (primary/secondary) for the database;
4. Selection of translation functions (regression models, fuzzy matching, neural nets
etc);
5. Soil scanning procedure (e.g. 1 m width -> 2 m depth);
For the selection of soil properties, we do have a list of priorities e.g. the target
variables for the GlobalSoilMap.net project (http://GlobalSoilmap.org/specifications),
but this certainly needs to be extended.
In summary I think that this is a great initiative and that we should form a consortium
and start applying for funds. Who should take the lead?
PS: Personally I think that we should almost combine this project with the ISIS recovery
project who have already ensured some budget (they plan to collect 100+ complete profiles
in the coming 12-18 months in different parts of the world). There is no need to walk to
distant parts of the world twice ;) For more info please contact Vincent van Engelen
and/or Prem Bindraban.
T. Hengl
Senior research @ ISRIC - World Soil Information
Mail: ISRIC - World Soil Information P.O. Box 353, 6700 AJ Wageningen, The Netherlands
Location: Room C.002, Droevendaalsesteeg 3 (GAIA building), 6708 WB Wageningen, The
Netherlands
Tel: +31(0)317-484199
Url: http://www.wewur.wur.nl/popups/vcard.aspx?id=HENGL001
Bob MacMillan, ISRIC, March 18, 2011
Tom does work closely with me and has a lot of skills and ideas to offer.
Please consider these thoughts of his as contributions to the discussion and not as
recommmendations or obligations.
He has raised a few points that will require discussion and debate before being
finalized. Some, such as only using new sites, selected according to objective criteria,
that can be accurately geolocated, fully described and characterized by spectral scanning
are worthy of discussion but we are not far enough along yet to make these decisions. We
need to talk these topics through. Please consider Tom's comments as representing his
contributions to the discussion but not necessarily any final decisions.
As to the suite of properties to include, that also is a topic of discussion. We
obviously have our immediate priorities for the GlobalSoilMap.net project but I recognize
the validity of arguments presented by Rich about including as wiide a range of
properties and methods as possible, given the large proportion of total costs that go
into just physically obtaining a sample. Also, we acknowledge that the library will serve
many conversion needs beyond just GlobalSoillMap.net and so will clearly need to include
some soil properties not yet of interest to the GlobalSoilMap.net project.
Bob
Larry West NRCS, March 21, 2011
Bob,
Thought I would add my perspective to the laboratory data harmonization discussion.
First, as far as we can stretch time, resources, and expertise, the NSSC Soil Survey
Laboratory will support the effort. If we are to be a global soil science community,
being able to reliability and rigorously compare data collected at different times for
different projects at different institutions is critical. Thus, the effort is badly
needed. I hope, however, that we do not just focus on IR spectra as a tool for deriving
data - new tools will continue to be developed. Also, we should make every effort to use
archived samples in various laboratories across the globe. Vast amounts of time and
resources have been invested in collecting and analyzing these samples and interpreting
the data. We need to use these samples to the extent possible.
Please let me know if I or others at the NSSC can do anything to help move this effort
forward.
Larry
Bob MacMillan, ISRIC, March 21, 2011
Larry,
We all very much appreciate receiving your input and advice.
I am only one person and am operating outside my comfort zone on parts of this
discussion. So I really benefit from input and advice.
I started setting up a web page where we can have a recorded discussion on this and other
issues. Hope to get it out in the next few days. Will be able to keep track of all
comments and come to an agreed consensus.
The good news is that you have added your voice to supporting the basic idea of the
reference and conversion library. Also, clearly having agreement from NRCS to provide
laboratory services (subject to funding and capacity limitations) will be hugely
important to selling the project. We are credible if we can demonstrate we have the
capacity to do the work if we get the funding.
I believe we have hit on a very specific and useful activity that will be globally
beneficial.
Will follow up on your comments.
Bob
Keith Shepherd, ICRAF, March 21, 2011
Dear Bob,
I have some project deadlines for end of month and so may be a bit quiet during this
period.
We should not be limited to properties that predict well by IR as new methods and
modifications are constantly evolving, as Larry also said.
We will need to prioritize methods as the quantity of soil needed will soon become a
limiting factor, and there is need to archive a good quantity for cross-calibrating new
methods that evolve in future. It would be good if we could start with a list of those
commonly used in international soil survey (does ISRIC have a standard set?) and then
supplement with those of high agronomic importance (i.e. widely used soil nutrient
tests). These are the type of decisions that need to be made during the preparatory
phase, and perhaps efforts at this stage should be put on getting a good communication
for fund raising purposes and communicating with node leaders on the scope and
feasibility of new sampling campaigns.
We should also not underestimate the number of profiles that may be needed to adequately
sample the soil data space and make robust inter-calibrations. Depths within profiles
will be highly correlated and will not contribute much to sampling the property space.
This is an area (Tom, Markus) where some good stats using existing data sets could come
in handy sooner than later. David Brown tried something simple in our earlier paper on
the NRCS archives but I am sure something more comprehensive could be done. I get send
the data sets.
Many thanks for all the inputs.
Keith Shepherd, ICRAF, March 24, 2011
Dear Bob,
Kindly see the BLGG response below.
I am not sure that BLGG will be able to offer the same scope in methods as Lincoln.
We (ICRAF) are developing an MOA with BLGG on other projects but I would not want to include anything in
that beyond exploring possibilities with them on the globalsoil.net project, and any concretization would
anyway presumably be through ISRIC on behalf of the globalsoi.net consortium.
If Alfred is visiting BLGG he may be able to explore further.
With thanks and best regards,
Keith
Begin forwarded message:
From: Peter van Erp <[email protected]>
Date: March 24, 2011 9:28:10 AM GMT+03:00
To: [email protected]
Cc: Thomas Terhoeven-Urselmans <[email protected]>
Subject: reference lab
Dear Keith,
I am glad to give you positive feedback to your question raised concerning BLGG being a reference lab for the
globalsoil.net initiative. We would see this as a first sub-project within the general MOU between ICRAF and
BLGG AgroXpertus.
I brought the issue in the last Management Team meeting and the members were generally interested in this
collaboration and see the potential positive long term effects of it. It fits into our medium and long term goals to
approach foreign markets to a bigger extent.
Nevertheless, the important question now is which methods need to be implemented. We have a certain set of
methods, which proofed to be reliable under Dutch conditions. For example, our main multi nutrient soil
extraction medium is 0,01 M CaCl2 and not Mehlich III. Thus, in case you need Mehlich III we need to invest
into the installation of this method. In the end it is for us as well a question of needed investments and projected
turnover. But we can't answer this question before we have discussed the methods. At this point the question of
competitiveness with other labs (eg for Mehlich III analysis) comes into the picture as well.
What is your opinion an what do you propose howwm tot continu?
Kind regards,
Peter