Chris Jenkins (Email)
INSTAAR, University of Colorado
17 June 2004
|
Entering Seabed Sample Descriptions (Word-based Data) |
Contents
One of the special functions of dbSEABED is to handle word-based descriptive data - linguistic data - and from them, produce output attributes that are compatible with the outputs from numeric analytical data.
A requirement for successful processing in this way is that the input linguistic data is reasonably well organized, although allowance is made for a variety of descriptive linguistic structures that are usual and familar to geologists, ecologists, navy divers, etc.
dbSEABED is flexible. It recognizes that different kinds of study use a different voculabaries, and a variety of linguistic structures and ways of attaching quantities. To be useful in dbSEABED, sediment descriptions and analysis data do not need to be absolutely perfect or straightjacketed: it is a data mining system and will make the best of what is supplied, while rejecting (and reporting for future attention) incomplete or erroneous structures.
At the same time, the language of the descriptions must be able to be computed on, ie. it should have an basically arithmetic character. The processing is therefore not quite "natural language" processing and stream of consciousness" prose, such as found in most DSDP and ODP core summaries, is not able to be handled without extensive editing at the time of data entry. An arithmetic character to descriptions is probably a good goal anyway in the direction of scientific rigor and observational discipline.
This web page sets out many commonly encountered descriptive structures, and describes how they should be entered to promote successful handled by dbSEABED.
Visual (deck) descriptions should be organized so more use can be made of the digital data. By "organized" we mean that the descriptions should not be prose , but itemized lists of observations that clearly divide the sample into its separate geological components.
An example
"Dredge load with rocks and sediment plus some liquid slush, rocks
red slush green and sediment coarse. Oops, color chart lost overboard.
Slush appears to be organic in part, rocks possibly Triassic red beds
and
the sediment carbonate-rich. Forams, pteropods, occasional bryozoans
and
bits of rock. "
This description closely follows one that was entered into dbSEABED
recently. It might be better organized in the log as follows:
Dredge load - 3 componentsThis data would be organized along these lines in dbSEABED;
1. rock, possibly Triassic red bed
also small bits of rock
2. sediment, coarse carbonate sand
rich in forams, pteropods, occasional bryozoans
3. green liquid slush - appears to be organic in part
| Sample metadata | = | Dredge load - 3 components | ||
| Rest of dataset metadata | = | Colour chart fell overboard | ||
| Component 1 | = | rock, also small bits of rock | = | rock = possibly Triassic red bed |
| Component 2 | = | sediment | = | coarse carbonate sand, rich in { forams, pteropods, occasional bryozoans } |
| Component 3 | = | green liquid slush | = | appears to be organic in part |
Notice that the order of components of the description is maintained, making data entry very easy in secretarial types of software like EXCEL and WORD.
Realize that descriptions consist of quantifiers, modifiers and objects (QMO) and treat descriptions like arithmetic expressions:
Usually (and hopefully) 'TheSedimentFraction' is the whole sediment. An example QMO description is: "abundant green shark_teeth + 30% fine sand + 10% nobbly phosphatic_nodules".(m* Ax + n*By + o*Cz ... = TheSedimentFraction).
You can use a shorthand to remove ambiguities from descriptions. The
"/" and "-" are used in dbSEABED to denote quantifiers (m,n,o) and
modifiers
(A,B,C) respectively, and point to the object that they refer to.
So, "green- shark_teeth /abundant + 30%/ fine- sand + 10%
phosphatic_nodules
-nobbly" is computationally equivalent to the example above.
Decide whether your syntax is front- (ODP/Scientific) or rear (NIMA/Navy) significant in abundances, and then place the descriptions in the correct corresponding field. Perhaps there is no front- rear- order of significance.
Most sediment descriptions are composed of a sequence of constructions like:
[quantity] (modifier) objectthat is, combinations of terms expressing abundance, some modification of a property and the object of interest itself. (A fourth, operators, can be considered as quantifiers.) In the example, “slightly muddy coarse bryozoan sand with trace quartz” we distinguish 3 objects and accompanying modifiers:
[slight] ( ) mud + [ ] (coarse) bryozoan_sand + [trace] ( ) quartz.a. Objects are terms which describe a property that may be texture, carbonate, lithology, strength, colour, etc.
In effect sediment descriptions are linear expressions of the form:
m * a * x + n *b * y + …. = total sediment .(Terms like 'with', 'or' or 'to' can be regarded as quantifiers or neutral).
Some modifiers are marked with a "-". All quantifiers
are marked with a "/". These point in the direction of the term they
modify
or quantify:
e.g., "fne- snd", "altrd- baslt"; ab/ bryz" and "shrk_teeth /rre" for
"fine sand", "altered basalt"; "abundant bryozoans" and "shark teeth
rare".
Changing words to abbreviations is best done in WORD using the find/replace functions on a big dataset. Its good to take the care to set 'Match Case' and/or 'Whole words' in the Replace setup, and to view each case as you are going along (ie. not to use 'replace all').
Also, be careful to do the abreviations per item of meaning. For example "SAND DOLLARS" to "snd_dllrs", "TRACES OF" to "trcs_of/" and "FINE TO MEDIUM" to "fne- to med-". Do these combined meanings before the individual ones like "SAND" to "snd".
Mostly, the choice of vocabulary doesn't worry dbSEABED, because its dictionary (actually thesaurus) is adaptable. The dictionary already includes French words and terms such as "beautiful" (bryozoans), "sardine tin" (dredged up !), "coffee grounds" (slush) and "fettucine" (seaweed). Such terms are treated either as neutral or given geological meaning.
Geological terms however, should be used in standard ways. For example, conforming to a geological dictionary such as that by AGU. If a term is non-standard, say so in a note (metadata) and notify the maintainer so proper arrangements can be made in the dictionary.
dbSEABED does not work through usual English word classes of verb/adjective/adverb/noun. Instead, it divides on what descriptive meaning is carried.
To do this it is best to do a search in WORD on the file
"db8_dct.rtf".
It is not advisable to search for the plural because plurals may be
marked as (for instance) "band;s" or "band/s", signifying optional
plural.
Alternatively, the web dictionary listing can be accessed db9_dictionary.htm .
Underscores are meant to link terms that complement each other. "limt_cemt" is a good example: limt is compositional and cemt is structural/textural. If they were separate then a lot of meaning would be lost. If you do what you (as a geologist) think is best to maximize the carrying of the geological meaning, you will be doing the right thing. That is the key test.
All underscore-linked terms need to be added to the thesaurus.
The main objection to joining a greater number of terms is that the thesaurus would really blow out in size. With "porphyritic rhyolite" ("prphrytc- rhyl"), "porphyritic" is really a standard (expectable) modifier for "rhyolite" and they could be kept separate.
Top of page
Using
standard stems and endings
When suggesting or making new abbreviations, it is best to draw on already defined abbreviations and just add standard variations like endings. This way, adjectives, adverbs and participles can be easily generated.
Examples of stems include:
FINE: fne-, fnely-, fner-, fneng_upwrdsExamples of endings:
SORTING: srtng, -srtd, wl_srtd-, prly_srtd-
QUARTZ: qtz, qtzse, qtz_gn;s
STONE: stn, limstn, wckestn, sndstn.
This has not always been adhered to (there are homonyms and synonyms in the thesaurus, English isn't perfect and neither are we).
...ized ...zd ...ed ...d ...ing ...ng ...s & ...es ...s ...ose ...se
Some stem conflicts include:
| GRAIN and GREEN | gn & grn |
| BAND and BOUND | band & bnd |
| CAST (colour) and CAST (vs mold) | cast(colr) & cast |
| SOFT (biological) and SOFT (geotechnic) | sft & so |
| HIGH DENSITY (abundance) and HIGH DENSITY (geotechnical) | hi_dnsty & hi_dens |
Do you encounter others ? Email me !
There is a set of semi-standardized terms that convey particle size: "0.8mmszd-", "-<4cmhi", "~8mwvl-", "4:9phiszd-", "0.061mmlng" are examples.
They are standardized as follows.
The purpose of this standardization is to permit automatic parsing at a later stage. Whencurrently, the units of measurement can only be mm, cm, m, um, in, ft and phi the symbols "<", ">" and "~" symbols must lie at the front the codes "szd", "hi", "lng", "sze", "rnd" or "wvl" end the term; (sized, high, long, size, wavelength, round) a range of size can be rendered using a colon between 2 numbers with the smaller number at front leading decimal points should be preceded with a "0" as with all modifiers, a hyphen ("-") is used at the front or rear of the terms to point to its associated object.
Non-standard arrangements will be reported to the "*.DGN" diagnostics file.
There is a set of semi-standardized terms that convey percentage abundance. Examples are "/<0.5%", "30:50%/", "<10%/".
They are standardized as follows.
The purpose of this standardization is to permit automatic parsing at a later stage. WhenA "/" must be placed at the front / back depending on the direction of the object being quantified. A % must come after the percentage numeric. The numeric can be integer or decimal. A range of percent can be conveyed using ":", as in "/30:50%" meaning 30 to 50% range.
The smaller numeric comes first.Less and greater than or approximation can be put in front of the numeric: "<33%/"
Non-standard arrangements will be reported to the "*.DGN" diagnostics file.
This syntax is only required in fields that expect word-based data.
dbSEABED recognizes that different types of studies may use a variety of word and linguistic structures. Structures are introduced that (i) convey the meanings unambiguously and artithmetically, (ii) keeps the terms in their original order in decriptions (this reduces re-formatting work during data entry).
Example 1 - Quantifier and modifier pointers. quant1/
object1 object2 /quant2 modif1-
object1
object2 -modif2
In dbSEABED the "/" quantifier-pointer and modifier-pointer structures
resolves cases where an object can be quantified or modified by terms
to
left or right. Some descriptions employ "forams 25%", others "25%
forams".
The insertion of pointers is easy during the EXCEL stage of data entry.
Example 2 - Grouped List. quant/ {
object1
+ object2 + ... }
This type of 'grouped component' or "{}" structure is usually
found in petrology (grain or clast count) data. It can be recognized at
the time of data entry (and "{" and "}" properly inserted, or dbSEABED
will recognize the structure itself (currently only in a "PET" line).
Thus, the pet (grain / clast) count:
">50% pblszd- crl_shls + pblszd- baslt + pblszd- andest"will code (manually or automatically) in dbSEABED format to:
"10-50% cly_trrg"
"10% pbls_volc + shls"
"3% macrfna_shls"and ...
"pblszd- crl_shls + pblszd- baslt + pblszd- andest, >50%"
"cly_trrg, 10-50%"
"pbls_volc + shls, 10%"
"macrfna_shls, 3%"
">50%/ { pblszd- crl_shls + pblszd- baslt + pblszd- andest }"Grouping is currently only available for Quantifiers, but will be extended eventually to Modifiers.
"10-50%/ {slt + cly_trrg }"
"<10%/ { pbls_volc + shls }"
"3%/ { anml_dbr },,,,,,,,PETend ".
Example 3 - Proportion of Special Fraction. object1
// modif2- object2, quant1
In many grain counts (PET sets) percentages of a grain type are given
as a proportion of some other fraction, such as of the coarse fraction
(gvl + snd). The "//" structure allows this to be expressed in terms of
a PRODUCT object (forward of "//") and QUOTIENT (behind "//"). An
example
is:
"1,PET, ,0,,hvy_min // csefrct,1.18"
"1,PET, ,0,,hvy_min // snd,2.22"
"1,PET, ,0,,hvy_min // non_carb- snd,1.33"
"1,PET, ,0,,opq // non_carb- snd,0.665"
"1,PET, ,0,,opq // hvy_min,/>50"
"1,PET, ,0,,leucx // hvy_min,12-15%/"
In some cases the absolute grain proportion of the whole sediment can be worked out. This occurs when the proportion of the quotient is known from before, such as where G:S:M are given in a preceding TXR, GRZ, CMP, SFT or LTH line. For this reason, it is important in dbSEABED datasets to have these lines preceding PET lines; the effectiveness of the data mining is increased.
A shortage of data in a dataset doesn't concern dbSEABED - it simply
mines what is available and what is complete enough. Obviously though,
time at sea is better spent if the survey datasets are more complete
and
richer in attributes.
Here is a checklist of what might be sought - as a minimum.
| Lithology | general geological description component by component |
| Texture | grainsize and sorting, if possible with estimates of rock:gravel:sand:mud (or r:g:s:silt:clay) ratios |
| Colour | either as Munsell Code (GSA Rock Colour code) or visual description (by a female preferably - not colour blind) |
| Grain composition | grain types with estimated percentages |
| Special structures | bioturbation, pellets, voids, lamination, imbrication, rootlets, form of top surface, ripples |
| Organics | smell, organics present ?, voids or bubbles |
| Consolidation | describe whether liquid/soft/stiff/hard and whether loose/friable, etc |
| Biota | especially shelly remains/livings |
| Outsize objects | clasts, nodules, shells and other shelly biota; very important for seabed roughness (for acoustics and hydrodynamics) |
A system of initial (top-level) description then second pass (more detailed) description works well and dbSEABED treats both as valid observations. Deck descriptions are more reliable in some ways, for outsize objects, colour and structures.
On board, observers might have a pocket penetrometer, GSA color chart, microscope, etc; these results can be associated with the descriptions in spreadsheet columns. Later laboratory analyses can be kept separately and digitally merged later with the spreadsheet data.
Before any data on lithology, tecture, strength, etc can be carried through dbSEABED, information has to be provided on the origin of the data. For this reason, "1,SRC" and "1,SFS" lines are mandatory. If they are not present, processing will not produce a result.
A typical stream of lines is:
| Divider just for format style:
0,FMT,********************************************************** Mandatory SRC line: 1,SRC,AGSORigSeis102,AGSO+USYD,DaveFeary,copyright,notto3rdparty,typein,typein,cruisereport,GAB,19-Jun-91 to 28-Jun-91,May2000 Metadata: 0,SRC,AGSO Rig Seismic Cruise Great Australian Bight to prepare for ODP 0,SRC,Dave Feary Noel James & Gavin Birch Describe which fields contain data: 0,LTH,Top|Bot|FieldLith|FieldConsol|FieldTexr|Layering|GeoAge|SedEnvir|Colour|Sorting||GrnChar|FreeFormDescn|Altrn/Wthrng|Munsell|C14Date||TrcFoss 0,PET,Top|Bot|GrainTyp|%GrainTyp|GrainTypSize|||GrnShape|GrnSortng|GRnCharDescn Mandatory SFS line: 1,SFS,102 DR 03,-35.279,130.752,4180,,4180,3660,,,4.00 kg (~10kg in rept),,24-Jun-91,17:20,W-central Ceduna Terrace Metadata noting author's rock type in dredge: 0,LTH,Lith A Actual attribute data describing the seafloor: 1,LTH,,,mudstn wi/ vthn/ mn_ctng,,,outr_surfc bord,CAMPAN to MAASTR,?/ estuarn,vdk- gry to blk,,,,,,10YR3/1 to 10YR2/1 1,PET,,,mud // mudstn,almst_whly/ New type of rock in same dredge 0,LTH,Lith B 1,LTH,,,sltstn_cbl wi/ :5mmthk fe_ri_minlyr,wl_cnsldtd,,unfrm- txtr & colr_varns,,?/ estuarn,grysh brn,,,,,,10YR5/3,,,astersma + chondrts + planolts 1,PET,,,pyrt // sltstn,mnr/ New site: 1,SFS,102 DR 03 (PD 2),-35.279,130.752,4180,,4180,3660,,,2.00 kg,,24-Jun-91,17:20,W-central Ceduna Terrace Metadata noting author's additional informal dredge name 0,LTH,Pipe dredge 2 This sites' attributes: 1,LTH,,,mud wi/ orng fe_strks,,,,CAMPAN to MAASTR,,vdk- grysh brn,,,,,,10YR3/2 1,PET,,,nan // mudstn,brrn_of/ |
Use "1,XXX," when the sample is from the seabed surface.
Use "2,XXX," when the sample comes from the subsurface.
Use "3,XXX," etc when the sample is from within some segregation
such as a mottle or clast. That information can be in the
layer/structure
field of "3,LTH" or as "0,LTH" metadata.
Chris Jenkins (Email)
INSTAAR, University of Colorado
17 June 2004