def sza_opencl(scene, time_slot):
dRightAscension, dDeclination, dGreenwichMeanSiderealTime = sunposIntermediate(
time_slot.year, time_slot.month, time_slot.day, time_slot.hour,
time_slot.minute, 0.0)
scale_x = scene["IR_108"].area.pixel_size_x
scale_y = scene["IR_108"].area.pixel_size_y
offset_x = scene["IR_108"].area.pixel_offset_x
offset_y = scene["IR_108"].area.pixel_offset_y
band = scene["IR_108"].data.astype(float32)
outputBand = scipy.zeros_like(band, dtype=float32)
mf = cl.mem_flags
ctx = cl.create_some_context()
outputBuffer = cl.Buffer(ctx,
mf.WRITE_ONLY | mf.USE_HOST_PTR,
hostbuf=outputBand)
prg = loadOpenCLProgram("../util/opencl/solarangle.cl", ctx)
queue = cl.CommandQueue(ctx)
prg.zenithKernel(queue, band.shape, None, outputBuffer,
float32(dGreenwichMeanSiderealTime),
float32(dRightAscension), float32(dDeclination),
float32(scale_x), float32(scale_y), float32(offset_x),
float32(offset_y))
cl.enqueue_copy(queue, outputBand, outputBuffer)
plot2dArray(outputBand,
title="sza_results",
show=True,
outputPath="/home/sebastian/Documents/tests_sza123_full.png")
return
def stratiformity_opencl(scene,
cloudmask,
show_plots=False,
plot_path="",
createPlots=False,
cl_ctx=None,
cl_queue=None):
# Mask band with -999.9-values
noDataValue = -999.9
band = scipy.where(cloudmask == False, noDataValue,
scene["IR_108"].data).astype(numpy.float32)
outputBand = scipy.zeros_like(band, dtype=numpy.float32)
mf = cl.mem_flags
if cl_ctx is None:
platforms = cl.get_platforms()
devices = platforms[1].get_devices(device_type=cl.device_type.CPU)
cl_ctx = cl.Context(devices)
if cl_queue is None:
cl_queue = cl.CommandQueue(cl_ctx)
inputBuffer = cl.Buffer(cl_ctx,
mf.READ_ONLY | mf.USE_HOST_PTR,
hostbuf=band)
outputBuffer = cl.Buffer(cl_ctx,
mf.WRITE_ONLY | mf.USE_HOST_PTR,
hostbuf=outputBand)
prg = loadOpenCLProgram("../util/opencl/std.cl", cl_ctx)
prg.calcSTD(cl_queue, band.shape, None, inputBuffer, outputBuffer)
cl.enqueue_copy(cl_queue, outputBand, outputBuffer)
outputBand = scipy.where(cloudmask == False, numpy.nan,
outputBand).astype(numpy.float32)
if createPlots:
plot2dArray(outputBand,
title="standard deviations of 10.8band",
show=show_plots,
outputPath=plot_path + "strat_stds.png")
#filter clouds of max 3 pixel size
std_band = scipy.where(outputBand <= 2., 1., 0.).astype(numpy.float32)
filtered_outputBand = scipy.zeros_like(std_band, dtype=numpy.float32)
inputBuffer = cl.Buffer(cl_ctx,
mf.READ_ONLY | mf.USE_HOST_PTR,
hostbuf=std_band)
outputBuffer = cl.Buffer(cl_ctx,
mf.WRITE_ONLY | mf.USE_HOST_PTR,
hostbuf=filtered_outputBand)
prg = loadOpenCLProgram("../util/opencl/std.cl", cl_ctx)
queue = cl.CommandQueue(cl_ctx)
prg.filterSmallClouds(queue, band.shape, None, inputBuffer, outputBuffer)
cl.enqueue_copy(cl_queue, filtered_outputBand, outputBuffer)
return scipy.where(filtered_outputBand > 0., True, False)
def stratiformity(scene,
cloudmask,
show_plots=False,
plot_path="",
createPlots=False):
stds = rolling_std(scene["IR_108"].data, cloudmask, 3)
if createPlots:
plot2dArray(stds,
title="standard deviations of 10.8band",
show=show_plots,
outputPath=plot_path + "strat_stds.png")
return scipy.where(stds <= 2., True, False)
def cloud_phase(scene,
cloudmask,
show_plots=False,
plot_path="",
createPlots=False):
test1_data = scipy.where(scene["IR_108"].data < 230., True,
False) # Everything <230K: Ice
if createPlots:
plot2dArray(test1_data,
title="cloud phase below 230K",
show=show_plots,
outputPath=plot_path + "clphase_belowTemp.png")
test4_data = scipy.where(scene["IR_087"].data - scene["IR_108"].data > 0,
True, False) # Where 8.7 - 10.8 > 0: Ice
if createPlots:
plot2dArray(test4_data,
title="cloud phase: 8.7 > 10.8",
show=show_plots,
outputPath=plot_path + "clphase_87-108diff.png")
result = cloudmask & ~(test1_data | test4_data)
return result
def snow_identification(scene,
show_plots=False,
plot_path="",
createPlots=False):
test1_data = scipy.where(scene["VIS008"].data > 0.11, 1.,
0.) # Snow has a minimal reflectance (0.11)
if createPlots:
plot2dArray(test1_data,
title="snow minrefl",
show=show_plots,
outputPath=plot_path + "snow_minrefl.png")
test2_data = scipy.where(
scene["IR_108"].data > 256., 1.,
0.) # Snow has a minimal Brightness Temperature (-17,15′C)
if createPlots:
plot2dArray(test2_data,
title="snow mintemp",
show=show_plots,
outputPath=plot_path + "snow_mintemp.png")
ndsi = (scene["VIS006"].data - scene["IR_016"].data) / (
scene["VIS006"].data + scene["IR_016"].data)
ndsi = scipy.where(ndsi != None, ndsi, 0.)
if createPlots:
plot2dArray(ndsi,
title="NDSI",
show=show_plots,
outputPath=plot_path + "snow_ndsi.png")
test3_data = scipy.where(ndsi > 0.4, 1., 0.) # NDSI > 0.4 --> Snow
return scipy.where(test1_data + test2_data + test3_data == 3, True,
False) # Where all 3 tests are true: Snow-Pixel!
def small_droplet_proxy(scene,
cloud_mask,
not_day_land_mask,
not_day_mask,
show_plots=False,
plot_path="",
createPlots=False):
array = np.ma.array(scene["IR_039"].data)
if createPlots:
plot2dArray(array,
title="03.9 array",
show=show_plots,
outputPath=plot_path + "sdp_array.png")
mask = cloud_mask & not_day_land_mask
if createPlots:
plot2dArray(mask,
title="sdp mask",
show=show_plots,
outputPath=plot_path + "sdp_land_cloud_mask.png")
array.mask = mask
blocks = block_by_size_generator(array.shape, array.shape[1],
array.shape[0], 0)
result_blocks = find_block_threshold(array, blocks)
#print result_blocks
threshold_array = scipy.empty_like(array)
blocks_into_threshold_array(result_blocks, threshold_array)
if createPlots:
plot2dArray(threshold_array,
title="sdp thresholds",
show=show_plots,
outputPath=plot_path + "sdp_thresholds.png")
array.mask = 0
result = (array > threshold_array) & cloud_mask & ~not_day_mask
if createPlots:
plot2dArray(result,
title="sdp result",
show=show_plots,
outputPath=plot_path + "sdp_result.png")
return result
def very_low_cloud_plausibility(scene,
mask_allPreviousSteps,
mask_cloudIdentification,
dem,
lapserate=0.007,
plot_path="",
show_plots=False):
nanmask = scipy.where(
mask_cloudIdentification == True, numpy.nan, 1.
) # Converting True/False mask to nan/1. mask (using mask_cloudIdentification because only this mask definitely has NO cloud pixels for "False"
BT_108_masked = nanmask * scene[
"IR_108"].data # Multiplying mask with scene["IR_108"].data so only non-masked pixels keep their values
dem = scipy.where(dem == -9999, 0.,
dem) # Exchanging nanvalues (-9999) (sea-pixels) with 0
DEM_masked = nanmask * dem # Multiplying mask with DEM so only non-masked pixels keep their values
xsize = len(BT_108_masked[0])
ysize = len(BT_108_masked)
cloudTopHeights = numpy.empty(
(ysize, xsize)) # Initializing the result-2darray...
cloudTopHeights.fill(numpy.nan) # ... and filling it with nans
# For each pixel that is a liquid-smalldroplet-stratiform-nonsnow-cloudy pixel
# get direct neighbouring pixels and calculate the mean for BT108 and the DEM-values
# but only for those neighbour pixels where the mask_cloudIdentification shows NO cloud!
# and calculate the height of the clouds of these pixels...
for i in range(ysize):
for j in range(xsize):
if mask_allPreviousSteps[
i, j] == True: # If the current pixel is a cloud pixel do:
avSurroundingBT = numpy.nan
N = 1
while numpy.isnan(
avSurroundingBT
): # as long as no non-cloud-neighbour is found: increase search radius by 1
iminus = 0
iplus = ysize
jminus = 0
jplus = xsize
if i - N > 0: iminus = i - N
if i + N <= ysize: iplus = i + N + 1
if j - N > 0: jminus = j - N
if j + N <= xsize: jplus = j + N + 1
avSurroundingBT = numpy.nanmean(
BT_108_masked[iminus:iplus, jminus:jplus])
avSurroundingElevation = numpy.nanmean(
DEM_masked[iminus:iplus, jminus:jplus])
cloudTopHeights[i,
j] = func(avSurroundingBT,
scene["IR_108"].data[i, j],
lapserate,
avSurroundingElevation, dem[i,
j])
cloudTopHeights[i, j] = func(avSurroundingBT,
scene["IR_108"].data[i, j],
lapserate, 0, 0)
N += 1
plot2dArray(cloudTopHeights,
title="Cloud Top Heights",
outputPath=plot_path + "vlc_cth.png",
show=show_plots)
return scipy.where(cloudTopHeights <= 1000., True, False)
#write_scene(scene, "/home/sebastian/Documents/")
#plot2dArray(lats,title="lats",show=True, outputPath="/home/sebastian/Documents/tests_sza123_full_lats.png")
#plot2dArray(lons,title="lons",show=True, outputPath="/home/sebastian/Documents/tests_sza123_full_lons.png")
# Correct nodata-value and value type of elevation model:
demWorldclim = gdal.Open(
"/home/sebastian/git/cloudmask/data/dem/dem_worldclim_fulldisk_noNegativeElevations.tif",
GA_ReadOnly)
demWorldclim_band = demWorldclim.GetRasterBand(1)
demWorldclim_data = demWorldclim_band.ReadAsArray(0, 0,
demWorldclim.RasterXSize,
demWorldclim.RasterYSize)
data = scipy.where(demWorldclim_data < 0., -999., demWorldclim_data)
plot2dArray(data, show=True)
proj = demWorldclim.GetProjection()
xsize = demWorldclim.RasterXSize
ysize = demWorldclim.RasterYSize
driver = gdal.GetDriverByName('GTiff')
dsO = driver.Create("/home/sebastian/Documents/dem_final_nonegativeValues.tif",
3712, 3712, 1, gdal.GDT_Int16)
dsO.SetProjection(proj)
dsO.GetRasterBand(1).WriteArray(int16(data))
dsO.FlushCache() # Write to disk.
