How to calculate the percentile for each cell from timeseries raster
Accepted Solutions
import numpy
import arcpy
flder = r"C:\Temp\Max3days" # where I unzipped your files
arcpy.env.workspace = flder
rasters = arcpy.ListRasters()
arrs = []
for i in rasters:
arrs.append(arcpy.RasterToNumPyArray(i))
a = np.array(arrs)
# ---- create the masked arrays assuming 0.0 is the null/nodata value ----
arrs_m = []
for i in arrs:
m = np.where(a[0] == 0, 1, 0)
am = np.ma.MaskedArray(i, mask=m)
arrs_m.append(am)
a_m = np.array(arrs_m)
n_90 = np.nanpercentile(a_m, 90., axis=0)
""" ---- I would like the coordinates and cell size of the lower left corner !!!
NumPyArrayToRaster (in_array,
{lower_left_corner},
{x_cell_size},
{y_cell_size},
{value_to_nodata})
"""
out = arcpy.NumPyArrayToRaster(n_90, value_to_nodata=0.0)
out_name = "\\".join([flder, "n_90.tif"])
out.save(out_name)
Now if you could fill me in on what the values should look like, I can change stuff.
The attached is the nanpercentile at the 90% percentile for the array stack.
Keep me posted, I haven't had a look in Arc anything yet, so do not add it into a dataframe or map with other data, since it has no coordinate or cell size information yet,.
ADDENDUM
pretty coarse grid or a small area
import numpy
import arcpy
flder = r"C:\Temp\Max3days" # where I unzipped your files
arcpy.env.workspace = flder
rasters = arcpy.ListRasters()
arrs = []
for i in rasters:
arrs.append(arcpy.RasterToNumPyArray(i))
a = np.array(arrs)
# ---- create the masked arrays assuming 0.0 is the null/nodata value ----
arrs_m = []
for i in arrs:
m = np.where(a[0] == 0, 1, 0)
am = np.ma.MaskedArray(i, mask=m)
arrs_m.append(am)
a_m = np.array(arrs_m)
n_90 = np.nanpercentile(a_m, 90., axis=0)
""" ---- I would like the coordinates and cell size of the lower left corner !!!
NumPyArrayToRaster (in_array,
{lower_left_corner},
{x_cell_size},
{y_cell_size},
{value_to_nodata})
"""
out = arcpy.NumPyArrayToRaster(n_90, value_to_nodata=0.0)
out_name = "\\".join([flder, "n_90.tif"])
out.save(out_name)
Now if you could fill me in on what the values should look like, I can change stuff.
The attached is the nanpercentile at the 90% percentile for the array stack.
Keep me posted, I haven't had a look in Arc anything yet, so do not add it into a dataframe or map with other data, since it has no coordinate or cell size information yet,.
ADDENDUM
pretty coarse grid or a small area
Wow excellent!
The result almost same with manual version, slightly different on the minimum value (your script is 109.391 and manual version is 105.947, for the max is same. See pictures attached (map and statistics info from gdal).
The raster extent is not exactly same with yours, because its from excel - convert to point - convert to raster and clip with boundary. Here's the file: n_90_manual.zip
And yes you are correct, its pretty coarse grid with 0.05 deg spatial resolution.
The data came from CHIRPS global precipitation data CHG - Data and this data is the best available global precipitation data with temporal resolution daily, 5days, 10days, monthly, 2monthly, 3monthly and annual from 1981 until now and free for public. So far this data are very useful for me when getting weather station data from govt is very difficult.
Thank you Dan Patterson
