def zonalStats(inShp,
inRaster,
bandNum=1,
mask_A=None,
reProj=False,
minVal='',
maxVal='',
verbose=False,
rastType='N',
unqVals=[],
weighted=False,
allTouched=False):
''' Run zonal statistics against an input shapefile. Returns array of SUM, MIN, MAX, and MEAN
INPUT VARIABLES
inShp [string or geopandas object] - path to input shapefile
inRaster [string or rasterio object] - path to input raster
OPTIONAL
bandNum [integer] - band in raster to analyze
reProj [boolean] - whether to reproject data to match, if not, raise an error
minVal/maxVal [number] - if defined, will only calculate statistics on values above or below this number
verbose [boolean] - whether to be loud with technical updates
rastType [string N or C] - N is numeric and C is categorical. Categorical returns counts of numbers
unqVals [array of numbers] - used in categorical zonal statistics, tabulates all these numbers, will report 0 counts
mask_A [numpy boolean mask] - mask the desired band using an identical shape boolean mask. Useful for doing conditional zonal stats
weighted [boolean] - apply weighted zonal calculations. This will determine the % overlap for each
cell in the defined AOI. Will apply weights in calculations of numerical statistics
RETURNS
array of arrays, one for each feature in inShp
'''
if isinstance(inShp, str):
inVector = gpd.read_file(inShp)
else:
inVector = inShp
if isinstance(inRaster, str):
curRaster = rasterio.open(inRaster, 'r')
else:
curRaster = inRaster
# If mask is not none, apply mask
if mask_A is not None:
curRaster.write_mask(mask_A)
outputData = []
if inVector.crs != curRaster.crs:
if reProj:
inVector = inVector.to_crs(curRaster.crs)
else:
raise ValueError("Input CRS do not match")
fCount = 0
tCount = len(inVector['geometry'])
#generate bounding box geometry for raster bbox
b = curRaster.bounds
rBox = box(b[0], b[1], b[2], b[3])
for idx, row in inVector.iterrows():
geometry = row['geometry']
fCount = fCount + 1
try:
#This test is used in case the geometry extends beyond the edge of the raster
# I think it is computationally heavy, but I don't know of an easier way to do it
if not rBox.contains(geometry):
geometry = geometry.intersection(rBox)
try:
if fCount % 1000 == 0 and verbose:
tPrint("Processing %s of %s" % (fCount, tCount))
# get pixel coordinates of the geometry's bounding box
ul = curRaster.index(*geometry.bounds[0:2])
lr = curRaster.index(*geometry.bounds[2:4])
# read the subset of the data into a numpy array
window = ((float(lr[0]), float(ul[0] + 1)), (float(ul[1]),
float(lr[1] + 1)))
if mask_A is not None:
data = curRaster.read(bandNum, window=window, masked=True)
else:
data = curRaster.read(bandNum, window=window, masked=False)
if weighted:
allTouched = True
#Create a grid of the input raster (data)
rGrid = polygonizeArray(data, geometry.bounds, curRaster)
#Clip the grid by the input geometry
rGrid['gArea'] = rGrid.area
rGrid['newArea'] = rGrid.intersection(geometry).area
#Store the percent overlap
rGrid['w'] = rGrid['newArea'] / rGrid['gArea']
newData = data
for idx, row in rGrid.iterrows():
newData[row['row'],
row['col']] = data[row['row'],
row['col']] * row['w']
data = newData
# create an affine transform for the subset data
t = curRaster.transform
shifted_affine = Affine(t.a, t.b, t.c + ul[1] * t.a, t.d, t.e,
t.f + lr[0] * t.e)
# rasterize the geometry
mask = rasterize([(geometry, 0)],
out_shape=data.shape,
transform=shifted_affine,
fill=1,
all_touched=allTouched,
dtype=np.uint8)
# create a masked numpy array
masked_data = np.ma.array(data=data, mask=mask.astype(bool))
if rastType == 'N':
if minVal != '' or maxVal != '':
if minVal != '':
masked_data = np.ma.masked_where(
masked_data < minVal, masked_data)
if maxVal != '':
masked_data = np.ma.masked_where(
masked_data > maxVal, masked_data)
if masked_data.count() > 0:
results = [
np.nansum(masked_data),
np.nanmin(masked_data),
np.nanmax(masked_data),
np.nanmean(masked_data)
]
else:
results = [-1, -1, -1, -1]
else:
results = [
np.nansum(masked_data),
np.nanmin(masked_data),
np.nanmax(masked_data),
np.nanmean(masked_data)
]
if rastType == 'C':
if len(unqVals) > 0:
xx = dict(Counter(data.flatten()))
results = [xx.get(i, 0) for i in unqVals]
else:
results = np.unique(masked_data, return_counts=True)
outputData.append(results)
except Exception as e:
if verbose:
print(e)
if rastType == 'N':
outputData.append([-1, -1, -1, -1])
else:
outputData.append([-1 for x in unqVals])
except:
print("Error processing %s" % fCount)
return outputData
def zonalStats(inShp,
inRaster,
bandNum=1,
reProj=False,
minVal='',
verbose=False,
rastType='N',
unqVals=[]):
''' Run zonal statistics against an input shapefile
INPUT VARIABLES
inShp [string] - path to input shapefile
inRaster [string] - path to input raster
OPTIONAL
bandNum [integer] - band in raster to analyze
reProj [boolean] - whether to reproject data to match, if not, raise an error
minVal [number] - if defined, will only calculation statistics on values above this number
verbose [boolean] - whether to be loud with responses
rastType [string N or C] - N is numeric and C is categorical. Categorical returns counts of numbers
unqVals [array of numbers] - used in categorical zonal statistics, tabulates all these numbers, will report 0 counts
RETURNS
array of arrays, one for each feature in inShp
'''
outputData = []
with rasterio.open(inRaster, 'r') as curRaster:
inVector = gpd.read_file(inShp)
if inVector.crs != curRaster.crs:
if reProj:
inVector = inVector.to_crs(curRaster.crs)
else:
raise ValueError("Input CRS do not match")
fCount = 0
tCount = len(inVector['geometry'])
for geometry in inVector['geometry']:
fCount = fCount + 1
if fCount % 1000 == 0 and verbose:
tPrint("Processing %s of %s" % (fCount, tCount))
# get pixel coordinates of the geometry's bounding box
ul = curRaster.index(*geometry.bounds[0:2])
lr = curRaster.index(*geometry.bounds[2:4])
# read the subset of the data into a numpy array
window = ((float(lr[0]), float(ul[0] + 1)), (float(ul[1]),
float(lr[1] + 1)))
try:
data = curRaster.read(bandNum, window=window)
# create an affine transform for the subset data
t = curRaster.transform
shifted_affine = Affine(t.a, t.b, t.c + ul[1] * t.a, t.d, t.e,
t.f + lr[0] * t.e)
# rasterize the geometry
mask = rasterize([(geometry, 0)],
out_shape=data.shape,
transform=shifted_affine,
fill=1,
all_touched=True,
dtype=np.uint8)
# create a masked numpy array
masked_data = np.ma.array(data=data, mask=mask.astype(bool))
if rastType == 'N':
if minVal != '':
masked_data = np.ma.masked_where(
masked_data < minVal, masked_data)
if masked_data.count() > 0:
results = [
masked_data.sum(),
masked_data.min(),
masked_data.max(),
masked_data.mean()
]
else:
results = [-1, -1, -1, -1]
else:
results = [
masked_data.sum(),
masked_data.min(),
masked_data.max(),
masked_data.mean()
]
if rastType == 'C':
if len(unqVals) > 0:
xx = dict(Counter(data.flatten()))
results = [xx.get(i, 0) for i in unqVals]
else:
results = np.unique(masked_data, return_counts=True)
outputData.append(results)
except Exception as e:
print(e)
outputData.append([-1, -1, -1, -1])
return outputData