James Geach

Extragalactic Astronomer / Science Writer

Jim Geach @jgeach

j.geach at herts.ac.uk.

Centre for Astrophysics Research (CAR), University of Hertfordshire, Hatfield, Hertfordshire, AL10 9AB, United Kingdom

Research diary

Research diary

30 Apr 2017 » Reawakenings

Another inexcusable hiatus in posts… must do better again


11 Dec 2013 » Paper

Inexcusable hiatus in posts… must do better. First purchases of the URF grant were iPad Air, new MacBook (Air, natch) and 27” Thunderbolt screen. For the former I have started loading up apps, and what I’m most excited about is ‘Paper’ by FiftyThree. It’s the closest thing I’ve seen to emulating an actual sketch book or note pad, with all the benefits of tech (gestures to magnify, rewind option, etc.). I’m going to be using this for sketching out ideas and concepts; I’ll post some here.


09 Sep 2013 » Galaxy Evolution Over Five Decades

Attended Galaxy Evolution Over Five Decades last week in Cambridge, to celebrate Malcolm Longair’s career. Nice set of talks (I talked about CLS) - it was particularly good to see Rashid Sunyaev talk in person.


17 Aug 2013 » CLS progress

Excellent progress this week with CLS: new SMP machine (256GB/32 core) with 30TB local store installed and running nicely. Scheduling jobs with minumum required memory for a single continuous chunk and one smurf thread allows much more efficient processing of many scans in parallel.

Working on LBG stacking and LABs in SSA22.


24 Jul 2013 » kd-tree nearest neighbours

Merging VICS82 J+K[short] with SDSS Stripe 82 photometry and SDSS (III) spectroscopy (mainly BOSS for QSO redshifts, but some SEGUE serendips in there too). Getting tens of thousands of galaxy redshifts (z<1), and thousands of quasar redshifts, typically with z=1-3, but a few very high-z guys in there too. Good news for the blind search.

Came across the scipy.spatial kd-tree module for quick finding duplicate entries (with almost-but-not-quite-the-same co-ordinates). Speed is important, since the VICS82 catalogue has well over a million entries. Don’t worry so much about cos(dec) term here, since we just care about finding (very) nearest neighbours:

from scipy import spatial

tol = 0.1/3600 #some small angle in degrees
coords = zip(ra,dec) #ra, dec arrays of coords
tree =  spatial.KDTree(coords) #KDTree instance
#query to return list of tuples of nearest neighbour (r<tol) indices
matches = tree.query_pairs(tol) 


15 Jul 2013 » Data reduction, etc.

Been churning through more SCUBA-2 CLS data reduction (8TB raw in the can), but pausing a week or so while our new SMP machine (256GB/32-core) and 32TB data store is added to the cluster.

Tried training the SOM on H-ATLAS SPIRE photometry and GAMA redshifts with the aim of producing a non-parametric photo-z estimator for sub-mm colours. Looks promising, but requires more burn-in.

Submitted SPT/WISE QSO cross-correlation Letter

Spent Thursday/Friday working on VICS82 data for survey definition paper. Positional accuracy looks good across the board, less then a quarter of an arcsecond scatter in VICS82-2MASS positions over wide fields. Counts look sane, but need to tweak photometric calibration a little by the look of it. Next job is to produce beta versions of J+K matched catalogues, combined with SDSS. Will allow basic gJK KX quasar selection for a start.