Pre-submission inquiry: CoordinateCleaner

[azizka]

Hi,

I'd like to submit the CoordinateCleaner package. It's a package for automated cleaning of biological and paleontological collection data, useful for conservation, ecology and evolutionary biology. There is overlap with scrubr, but CoordinateCleaner has a lot of additional functions. So hopefully it could still be appropriate?

In particular the package adds:

• additional record-level tests to identify problems (capitals, urban areas, oceans, biodiversity institutions)
• a novel geo-referenced database of ~10,000 biodiversity institutions, with an related app to add more at http://biodiversity-institutions.surge.sh/
• novel algorithms to test for problems unidentifiable on record level, in particular conversion errors in coordinate annotation and rasterized sampling schemes
• functions suited to clean fossil occurrences
• tutorials on presenting pipelines to clean coordinates from the biggest public data-providers in the field (www.gbif.org, www.paleobiodb.org)

CoordianteCleaner is on CRAN and I had discussed it briefly with @sckott when he was in Stockholm in October 2016. There is also a manuscript ready for submission to Methods in Ecology and Evolution linked with the package.

Thanks,
Alex

Summary

• What does this package do? (explain in 50 words or less):
  Scan data sets of recent and fossil species occurrence records for geo-referencing and dating imprecision and data-entry errors in a standardized and reproducible way.

• Paste the full DESCRIPTION file inside a code block below:

Package: CoordinateCleaner
Type: Package
Title: Automated Cleaning of Occurrence Records from Biological Collections
Version: 1.0-7
Date: 2018-03-01
Authors@R: c(person(given = "Alexander", family = "Zizka", email = "alexander.zizka@bioenv.gu.se",
                    role = c("aut", "cre")),
             person(given = "Daniele", family = "Silvestro", role = c("ctb")))
Description: Automated cleaning of geographic species occurrence records by automated flagging of problems common to biodiversity data from biological collections. Includes automated tests to easily flag (and exclude) records assigned to country or province centroid, the open ocean, the headquarters of the Global Biodiversity Information Facility, urban areas or the location of biodiversity institutions (museums, zoos, botanical gardens, universities). Furthermore identifies per species outlier coordinates, zero coordinates, identical latitude/longitude and invalid coordinates. Also implements an algorithm to identify data sets with a significant proportion of rounded coordinates. Especially suited for large data sets. See <https://github.com/azizka/CoordinateCleaner/wiki> for more details and tutorials.
License: GPL-3
Depends: R (>= 3.0.0), sp
Imports: geosphere, ggplot2, methods, raster, rgeos, rnaturalearth, stats
LazyData: true
RoxygenNote: 6.0.1

[karthik]

Thanks @azizka. We will discuss this and get back to you.

[sckott]

hi @azizka - nice to see you here.

As I probably mentioned to you in Stockholm, we do have a overlap policy where we try not to have packages that overlap too much. But here I think we might be okay. What do you think are the areas of overlap for the two packages? Maybe we can just avoid overlapping functionality.

[azizka]

Hi @sckott & @karthik , thanks!

I dug into both packages, after all the overlap seems rather small. See below for a by-function comparison table with scrubr from CRAN (please correct it if I missed something).

In two sentence: The aim of both packages is identical -- to improve quality of occurrence records from large databases, beyond that there is actually little overlap. Few basic functionalities are virtually identical, beyond that scrubr includes date- and taxonomic cleaning, while CoordinateCleaner includes many unique feature for coordinates and fossils, and enables custom gazetteers and custom precision for the match with political centroids and capitals.

Sorry for the slow reply, I was out of office.

Functionality	CoordinateCleaner1.0-7	scrubr 0.1.1	Percent overlap
Missing coordinates	cc_val	coord_incomplete	100%
Coordinates outside CRS	cc_val	coord_impossible	100%
Duplicated records	cc_dupl	dedup	The aim is identical, methods differ
0/0 coordinates	cc_zero	coord_unlikely	100%
Identical lon/lat	cc_equ	-	0%
Country capitals	cc_cap	-	0%
Political unit centroids	cc_cen	"not ready yet"	0%
Coordinates in-congruent with additional location information	cc_count	coord_within	100%
Coordinates assigned to GBIF headquaters	cc_gbif	-	0%
Coordinates assigned to the location of biodiversity institutions	cc_inst	-	0%
Spatial outliers	cc_outl	-	0%
Coordinates within the ocean	cc_sea	-	0%
Coordinates in urban area	cc_urb	-	0%
Coordinate conversion error	dc_ddmm	-	0%
Rounded coordinates/rasterized collection	dc_round	-	0%
Fossils: invalid age range	tc_equal	-	0%
Fossils: excessive age range	tc_range	-	0%
Fossils: temporal outlier	tc_outl	-	0%
Fossils: PyRate interface	WritePyrate	-	0%
Wrapper functions to run all test	CleanCoordinates, CleanCoordinatesDS, CleanCoordiantesFOS	-	0%
Database of biodiversity institutions	institutions	-	0%
Taxonomic cleaning	-	tax_no_epithet	0%
Missing date	-	date_missing	0%
Add date	-	date_create	0%
Date format	-	date_standardize	0%

[sckott]

thanks for this @azizka - Sorry about delay in responding. We agree that the overlap isn't sufficient to warrant concern. We'd like you to submit the package for review. Open a new issue and fill out the issue template you'll see.

[sckott]

closing this, looking forward to your submission