def testB(scores=""):
"""
order/arrange the WRFs according to proximity to COMPREF
"""
"""
USE
from armor.tests import roughwork as rw
reload(rw)
rw.testB('/home/k/ARMOR/documents/2013-final-report/1383883165/testA.pydump')
"""
#from armor.objects3 import kongrey # the data
#kongrey.fix()
from armor.objects3 import kongreywrf2 # the models
kongreywrf2.fix() #fixing the w.name's for w in kongreywrf
testBfolder = outputFolder + 'testB' + timeString + '/'
if not os.path.exists(testBfolder):
os.makedirs(testBfolder)
if scores == "" :
testAfile = [v for v in sorted(os.listdir(outputFolder), reverse=True) if 'testA' in v][0]
print 'loading', outputFolder + testAfile
scores = pickle.load(open(outputFolder + testAfile))
elif isinstance(scores, str):
print 'loading', scores
scores = pickle.load(open(scores))
testAfile = scores
else:
testAfile = ""
print 'sleeping 1 seconds'
time.sleep(1)
Tlist = [v.dataTime for v in kongrey]
top8matches = {}
for T in Tlist:
print '\n...............................\nTime:', T
pairList = [v for v in scores.keys() if T in v[0]]
print pairList[:5]
if len(pairList) == 0:
continue
pairList.sort(reverse=True, key=lambda v:scores[v])
top8matches[T] = [v[1] for v in pairList][:8]
v0 = pairList[0][0] # temporary variable
print 'Top 8 matches:', '\n'.join([v+ '\t' + str(scores[(v0,v)]) for v in top8matches[T]])
pickle.dump(top8matches, open(testBfolder+ 'top8matches.pydump', 'w'))
# construct the 3x3 panel
imList = kongrey(T) + [kongreywrf2(M)[0] for M in top8matches[T]]
#kongrey.load(T)
#for im in imList[1:]:
# im.load()
#debug
#print imList
img9 = construct3by3(imList)
img9.imagePath = testBfolder + 'best8matches' + T + '.png'
print 'saving images to', img9.imagePath
img9.saveImage(dpi=600)
try:
open(testBfolder + 'notes.txt').write('source:\n'+ testAfile + '\n\n' +str(scores))
except:
print 'error in writing ' + testBfolder + 'notes.txt'
pass
return top8matches
def testA(startingFrom='20130828.0600', endsAt='20130830.0900', threshold=-9999):
"""
testA : get all the correlations
"""
"""
1. apply LoG(sigma=100) + threshold to models (M01-M20, +- 6 hours) and observations
2. extract regional "features" and compare
3. output the results to '/home/k/ARMOR/documents/2013-final-report/'
"""
# 1. apply LoG(sigma=100) + threshold to models (M01-M20, +- 6 hours) and observations
# set up
if not os.path.exists(outputFolder):
os.makedirs(outputFolder)
# loading the stuff
from armor.objects3 import kongrey # the data
from armor.objects3 import kongreywrf2 # the models
kongreywrf2.fix() #fixing the w.name's for w in kongreywrf
kongrey.fix()
scores = {}
for count, k in enumerate([v for v in kongrey if (v.dataTime>=startingFrom) and (v.dataTime<=endsAt)]):
# set up
# k = kongrey radar data
# w = wrf output
# to compare w to k, we use w.momentNormalise(k)
k.load()
k.setThreshold(0)
k.backupMatrix()
T0 = k.timeDiff(hours=-6)
T1 = k.timeDiff(hours= 6)
wrfs = [w for w in kongreywrf2 if w.dataTime>=T0 and w.dataTime<=T1]
k.mask = k.gaussianMask(100)
for w in wrfs:
# loading
#k.load()
#k.setThreshold(0)
k.restoreMatrix(0) # 2013-11-11
w.load()
w.setThreshold(0)
if threshold != -9999: # added 2013-11-11
if isinstance(threshold, int):
k.matrix *= (k.matrix>=threshold)
w.matrix *= (w.matrix>=threshold)
elif threshold == "volume":
#k.matrix *= (k.matrix>=threshold)
#vol_k = k.volume(threshold)
#threshold_w =threshold-20
#vol_w = w.volume(threshold_w)
#while vol_w > vol_k:
# theshold_w +=5
# vol_w = w.volume(threshold_w)
#w.matrix *= (w.matrix>=threhold_w)
#print "thresholds for RADAR and WRF:", threshold, threshold_w
#print "volumes for RADAR and WRF:" , vol_k, vol_w
k.matrix *= (k.matrix>=30) # fixed 2013-11-15 ; default threshold=30
vol_k = k.volume(30)
threshold_w =threshold-10
vol_w = w.volume(threshold_w)
while vol_w > vol_k:
theshold_w +=5
vol_w = w.volume(threshold_w)
w.matrix *= (w.matrix>=threhold_w)
print "thresholds for RADAR and WRF:", threshold, threshold_w
print "volumes for RADAR and WRF:" , vol_k, vol_w
# matching:
# 1. filter; 2. get component; 3.construct masking functions;
# 4. mask; 5. normalise ; 6. compare
w.mask = w.gaussianMask(sigma=100, fraction=0.8)
# compare
k.matrix *= k.mask
w.matrix *= w.mask
w2 = w.momentNormalise(k)
scores[(k.name, w.name)] = w2.corr(k)
# ordering
# output
print k.name, ',', w.name, scores[(k.name, w.name)]
pickle.dump(scores, open(outputFolder+ 'testA' + timeString + 'from'+startingFrom + 'to' +endsAt +\
'threshold_' + str(threshold) + '.pydump', 'w'))
kongreywrf2.unload(w.name)
return scores
