Category Archives: structure-from-motion

Open Structure-from-Motion Data (pt.2)

I wrote a post last summer about kite-based structure-from-motion data released on figshare in collaboration with a few colleagues (a fun side project). At the time, I couldn’t find a more specific repository for the raw images, GPS data, orthophotomosaic, and digital surface model. During AGU 2016 I stopped by the OpenTopography booth and apparently I can contribute my topographic data to the OpenTopography repository. A good way to make sure that small bespoke structure-from-motion datasets are stored in a place where people might find the data (i.e., find the SfM data when scanning through other relevant lidar data). My continued hope is that others may find these datasets useful, even if it seems that most data in repositories remains uncited (see here and here)

I originally published the data with the hopes of citing it in my own coastal dune research — and  I did manage use and cite this data on my AGU 2016 poster. This idea of citing a published dataset in a paper seems to be part of the ‘Geoscience Paper of the Future‘ concept (Gil et al., 2016David et al., 2016) where datasets, code, models, data analysis, etc. are all published seperately (i.e., given a DOI) and cited in the final manuscript. Among other benefits, these open science methods aid reproduceability and reuse by other researchers.

IMG_2863_edited.jpg

Photo credit: Mary Lide Parker / UNC Research

p.s.— check out Mary Lide’s work, it’s awesome.

 

Advertisements

Citation Distributions for Geomorphology Journals

The title of this post could also be “Where should I publish my geomorphology research?” or “Does it matter where you publish your geomorphology article?”

Nature (Callaway, 2016) recently reported on a bioRxiv preprint about Journal Impact Factor and the distribution of citations for journal articles (Larivière et al. 2016). A key point of the preprint is simple: A journal’s impact factor is a poor metric for individual papers, so journals should publish the citation distribution (for all papers in the journal) for greater ‘transparency’.

Just to get us all up to speed, the Journal Impact Factor is proprietary, but calculated by taking the mean citations (in a given year) for all the papers published by a given journal in the previous two years.

Larivière et al. 2016 presented citation distributions, mean citations per paper, and Journal Impact Factors for 11 journals: eLife, EMBO, J. Informetrics, Nature, Nature Communications, PLOS Bio, PLOS Genetics, PLOS ONE, Proceedings of the Royal Society B, Science, and Scientific Reports. The authors of the preprint state that they echo previous work by journal editors and scholars of citations: 1) citation distributions are skewed, 2) most papers have fewer citations than the impact factor, 3) citation spreads for journal can span several orders of magnitude, and  4) journal citation distributions tend to look similar to one another. 

I was curious to see how what this distribution looked like for geomorphology journals. The supplementary material of Larivière et al. 2016 describes how to perform the analysis with subscription services that I have access to (i.e., Scopus; see appendix 2 of the preprint). The Journal Impact Factor, median citations per paper, and citation distribution are shown for the 2014 and 2015 Impact Factor calculations of four journals: Journal of Geophysical Research – Earth Surface, Geomorphology, Earth Surface Processes and Landforms, and Earth Surface Dynamics. It should be noted that the impact factor for JGR-ES is calculated from all the papers in the entire JGR suite of journals (Oceans, Biogeoscience, Solid Earth, Earth Surface, Planets, Space Physics, and Atmosphere).

Here is the data:20142015GJ.jpg

Just with a cursory glance, here are some things to note:

  • ESurf was given its first Impact Factor in 2015, so the 2014 analysis is moot (and not shown).
  • The median citation per paper is identical in all journals for 2014-2015 (except Esurf, but it is still new). As is the shape, skewness, and range of the data (i.e., no journal seems to have a fatter tail)
  • From these four journals,  two most highly cited papers from 2014 are both about ‘Structure-from-Motion’ photogrammetry (James and Robson, JGR-ES; Westoby et al. Geomorphology)
  • From these four journals, two of the three most highly cited papers in 2015 are about ‘Structure-from-Motion’ or high resolution topography (Javernick et al., Geomorphology; Tarolli et al., Geomorphology;….reviews were also highly cited in the 2015 data)

 

I’m sure there is lots more to glean from this data… and I’m sure there is bibliometric research/analysis/techniques that would help… but I am pleading ignorance here…let me know if you have ideas…

Open Structure-from-Motion Data

I am on the cusp of releasing a third structure-from-motion dataset on figshare (I’m uploading the final bits). For each of these datasets, coworkers and I have released the raw images (taken from a hacked canon camera (CHDK) suspended from a kite), XYZ positions of ground control points, a cm-scale digital surface model, a mm-scale orthophotograph, and a ‘report’ on the error and processing parameters Agisoft Photoscan so others could recreate the analysis.

With various coauthors I have ~ 8 more datasets to release from Hog Island, VA, USA at some later data (this work was done in conjunction with a project at the VCR-LTER).

In all of these ‘kite maps’ the observed area is very small compared to traditional remote sensing products (the kite maps are 0.025 km^2) — the maps are focused on what I am interested in: coastal dune systems. My hope is that other might find the data useful as a supplement to the available lidar data and aerial photography.

I bet there is more of this ‘dark’ topographic data lying around on hard drives at universities around the world….