JSON Well Log Format
Setup
Capture the JSON Well Log Format GitHub content to local disk by:
$ git clone https://github.com/JSONWellLogFormat/JSONWellLogFormat.git
Background
Most well log and drilling data in the oil and gas industry is trapped within tapes and disk files of ancient and hard to access data formats like DLIS, LAS, LIS, BIT, XTF, WITS, ASC and SPWLA.
These formats represents orphaned technologies and are outdated in all possible ways. Their syntax is overly complex, convoluted and awkward, available support software is limited, software tools are rare and documentation is poor or nonexistent.
But still: These are the main storage and communication media for well logging information in the 2018. The amount of data is immense and growing, as is the aggregate cost of maintaining and utilizing this information.
The JSON Well Log Format
The JSON Well Log Format is a modern well log format designed for the future requirements of simplicity, compatibility, speed, massive storage, massive transmission, cloud computing and big data analytics. It overcome many of the deficiencies of existing well log formats.
- Based on the JavaScript Object Notation (JSON) open standard (RFC 8259 and RFC 7493)
- Non-proprietary
- Text-based, lightweight and human readable
- Full UTF-8 support according to the JSON standard
- Built-in no-value support
- Simple syntax consisting of collections of name/value pairs (objects) and ordered lists of values (arrays)
- Compact type system
- Quantity and unit support based on the Unit of Measure Standard from Energistics
- Date and time support through the ISO 8601 standard
- Well log semantics based on a limited set of well known keys to ensures consistency, compatibility and efficient processing
- Supports depth and time based logging data
- Supports single value and multi-dimensional (image) curves
- Fast: The simple syntax and streaming nature makes parsing extremely efficient
- Omnipresent parsers and generators for just about any system environment available
- Existing ecosystem of NoSQL cluster database support with high volume storage, search and indexing, distribution, scalability and high performance analytics
Example
A JSON Well Log file consists of one or more log sets each containing a log header, curve definitions and the corresponding measurement data. This example contains a single log set with two one-dimensional curves:
[
{
"header": {
"name": "EcoScope Data",
"well": "35/12-6S",
"field": "Fram",
"date": "2019-06-14",
"operator": "Logtek Petroleum",
"startIndex": 2907.79,
"endIndex": 2907.84,
"step": 0.01
},
"curves": [
{
"name": "MD",
"description": "Measured depth",
"quantity": "length",
"unit": "m",
"valueType": "float",
"dimensions": 1
},
{
"name": "A40H",
"description": "Attenuation resistivity 40 inch",
"quantity": "electrical resistivity",
"unit": "ohm.m",
"valueType": "float",
"dimensions": 1
}
],
"data": [
[2907.79, 29.955],
[2907.80, 28.892],
[2907.81, 27.868],
[2907.82, 31.451],
[2907.83, 28.080],
[2907.84, 27.733]
]
}
]
The JSON syntax can be efficiently parsed in any programming environment available. The well log semantics must still be parsed by the client code, but this is far simpler to do navigating in-memory data structures in the programming environment at hand, instead of dealing with external disk resources of obscure proprietary formats.
Data types
The JSON Well Log Format defines the following data types for header data and curves:
Type | Description | Examples |
---|---|---|
float | Floating point decimal numbers | 10.2, 0.014, 3.1e-108, 2.13e12, 0.0, null |
integer | Integer decimal numbers | 10, 42, 1000038233, -501, null |
string | Text strings | "error", "final depth", "message 402", "", null |
datetime | Date/time specifications according to ISO 8601 | "2019-12-19", "2010-02-18T16:23:48,3-06:00", null |
boolean | Logic states | true, false, null |
Numbers must contain values corresponding to a double-precision 64-bit IEEE 754 binary format value. Integer values has the same internal representation in JavaScript as floats and should be limited to 52 bits (+/-9007199254740991) to ensure accuracy.
Also, numeric values that cannot be represented as sequences of digits (such as Infinity and NaN) must be avoided.
Log header
The following header keys are defined as well known:
Key | Type | Description |
---|---|---|
name | string | Log name |
description | string | Log description |
externalIds | object of key,value string pairs | IDs within external storage, key being the storage name, and value being the ID. |
well | string | Well name |
wellbore | string | Wellbore name |
field | string | Field name |
country | string | Country of operation |
date | datetime | Logging date |
operator | string | Operator company name |
serviceCompany | string | Service company name |
runNumber | string | Run number |
elevation | float | Vertical distance between measured depth 0.0 and mean sea level in SI unit (meters). |
source | string | Source system or process of this log |
startIndex | According to index value type | Value of the first index. Unit according to index curve. |
endIndex | According to index value type | Value of the last index. Unit according to index curve. |
step | According to index value type | Distance between indices if regularly sampled. Unit according to index curve. If log is time based, milliseconds assumed. |
All header data are optional.
In addition to the listed entries, clients may add any number of custom header data in any form supported by the JSON syntax. Note that the general informational value of custom header data is low. Some clients may understand the meaning of the entries, but in general such information is not fit for further processing.
Curve definition
The following keys are used for curve definitions:
Key | Type | Description |
---|---|---|
name | string | Curve name or mnemonic. Mandatory. Non-null. |
description | string | Curve description. Optional. |
quantity | string | Curve quantity such as length, pressure, force etc. Optional. |
unit | string | Unit of measurement such as m, ft, bar, etc. Optional. |
valueType | string | Curve value type: float, integer, string, datetime or boolean. Non-null. Optional. float assumed if not present. |
dimensions | integer | Number of dimensions. [1,>. Non-null. Optional. 1 assumed if not present. |
axis | array of curve definition | A detailed description of the multi-dimensional structure of the curve in case this spans multiple axes. One element per axis. The combined product of the axis diemsnsions elements must equal the dimensions of the curve. Optional. |
Quantities and units should follow the Unit of Measure Standard from Energistics. To ease transition from legacy formats this is no requirement.
In addition to the listed, clients may add any number of custom curve definition entries in any form supported by the JSON syntax.
Curve data
Curve data are specified in arrays for each index entry, with one entry per curve. If a curve is multi-dimensional, the entry is itself an array of subentries, one per dimension.
Curve values are according to the value type defined for the curve, or null
for no-values. The index curve is always the first curve listed, and must not contain no-values. It is advised that the index curve is continuously increasing or decreasing.
No custom additions to the curve defintion may alter the structure of the data definition as specified above.
Transition objects
To support convertion of legacy formats to JSON a generic table object has been suggested. The table has a set of attributes and it contains a number of named objects with one or more values for each attribute. The table is able to represent metadata from existing well log formats in a consistent and simple manner:
"name": {
"attributes": ["attr1", "attr2", "attr3", ... "attrn"],
"objects": {
"object1": [v11, v12, v13, ... v1n],
"object2": [v21, v22, v23, ... v2n],
"object3": [v31, v32, v33, ... v3n],
:
"objectm": [vm1, vm2, vm3, ... vmn]
}
}
Metedata in LAS files exists as parameters within a section and has the following form:
<name>.<unit> <value> : <description>
A typical example might be:
~PARAMETER INFORMATION
#MNEM.UNIT VALUE DESCRIPTION
#---- ----- -------------------- ------------------------
RUN . 1A : RUN NUMBER
PDAT. MSL : Permanent Datum
EPD .C3 0.000000 : Elevation of Permanent Datum above Mean Sea Level
LMF . DF : Logging Measured From (Name of Logging Elevation Reference)
APD .M 30.000000 : Elevation of Depth Reference (LMF) above Permanent Datum
Using the table object above, this should convert to JSON as follows:
"PARAMETER INFORMATION": {
"attributes": ["value", "unit", "description"],
"objects": {
"RUN": ["1A", null, "RUN NUMBER"],
"PDAT": ["MSL", null, "Permanent Datum"],
"EPD": [0.0, "C3", "Elevation of Permanent Datum above Mean Sea Level"],
"LMF": ["DF", null, "Logging Measured From (Name of Logging Elevation Reference)"],
"APD": [30.0, "M", "Elevation of Depth Reference (LMF) above Permanent Datum"]
}
}
Metedata in DLIS files exists as sets. This is a named entity with a number of attributes and a number of objects with one or more values for each of the attributes. A DLIS set has a binary representation within a DLIS file, but it can be viewed as a matrix as follows:
setName
attr1 attr2 attr3 ... attrn
----------------------------------------
object1 v11 v12 v13 v1n
object2 v21 v22 v23 v2n
object3 v31 v32 v33 v3n
:
objectm vm1 vm2 vm3 vmn
----------------------------------------
A typical example might be:
HzEquipment
LENGTH TRADEMARK-NAME SERIAL-NUMBER WEIGHT
---------------------------------------------------------
APWD 0.0 in APWD25-AA 241408 0.0 kg
ARDC 224.8 in ARC9D-BA 738 1270.0 kg
MSSD900 14.5 in SZR FC-71545 68.0 kg
---------------------------------------------------------
Using the generic table structure, this will convert to JSON as follows:
"HzEquipment": {
"attributes": ["LENGTH", "TRADEMARK-NAME", "SERIAL-NUMBER", "WEIGHT"],
"objects": {
"APWD": ["0.0 in", "APWD25-AA", "241408", "0.0 kg"],
"ARDC": ["224.8 in", "ARC9D-BA", "738", "1270.0 kg"],
"MSSD900": ["14.5 in", "SZR", "FC-71545", "68.0 kg"]
}
}
Writing JSON well logs
Writing JSON well logs can be done in two different formats: condensed or pretty. The condensed format should be without whitespace and newlines and should be used for transmission between computers only.
For well logs that may possibly be viewed by humans the pretty format should always be used. This format should contain whitespace and indentation that emphasizes the logical structure of the content. For the data section in particular, arrays of curve data for each index must be written horizontally and commas between entries should be aligned:
"data": [
[828420, 282.589, 8.6657, 2.202, 2.222, [1.759, 2.31469, 1.33991E-3, 3.75839], 0.52435, ... ],
[828480, 286.239, 9.6601, 2.277, 2.297, [2.219, 2.31189, 1.12295E-3, 3.72152], 0.52387, ... ],
[828540, 276.537, 10.6638, 2.309, null, [2.267, 2.29509, -3.67117E-3, 3.70351], 0.53936, ... ],
[828600, 264.325, 10.6545, 2.324, null, [2.110, 2.27902, -7.77555E-3, 3.67927], 0.55439, ... ],
[828660, 245.938, 9.6937, 2.333, 2.356, [1.525, 2.26512, -1.17965E-2, 3.68386], 0.56211, ... ],
:
]
Schema
Schema for the JSON Well Log Format is available at https://jsonwelllogformat.org/schemas/JsonWellLogFormat.json.
Examples - The Volve data set
Thanks to Equinor all subsurface and production data from the Volve field on the Norwegian continental shelf has been disclosed and made available to the public. It can be downloaded from http://data.equinor.com.
The dataset contains about 15GB of well log data in about 1000 different DLIS, LIS, LAS, ASC and SPWLA files. This part has been converted to the JSON Well Log Format by Petroware and is available here.
An online viewer based on the JavaScript code contributed here is available at https://jsonwelllogformat.org/viewer.