File:300 ma atmosphere near surface temp mean 1.png

2. 1. netcdf downscaler
   3. also habitat test
   4. python 3,GDAL
3. 1. v 0012.0004
   2. 07.01.2022

import numpy as np import scipy import pandas as pd import matplotlib.pyplot as plt import matplotlib.image as mpimg from matplotlib import cm from colorspacious import cspace_converter from collections import OrderedDict

1. import colormaps as cmaps
2. import cmaps

import netCDF4 as nc import os from scipy.interpolate import Rbf

from sklearn.linear_model import LinearRegression from sklearn.preprocessing import PolynomialFeatures from sklearn.pipeline import make_pipeline from sklearn.ensemble import RandomForestRegressor from sklearn.preprocessing import StandardScaler from sklearn.model_selection import train_test_split from sklearn import svm, metrics from pygam import LogisticGAM from pygam import LinearGAM

import pandas as pd import array as arr import scipy.stats

1. if basemap is available, we'll use it.
2. otherwise, we'll improvise later...

try:

   from mpl_toolkits.basemap import Basemap
   basemap = True

except ImportError:

   basemap = False

1. 1. control vars

1. 1. random fotest

RF_estimators1=10 RF_features1=2

1. 1. downscaler

DS_method=0 ## default random forest DS_show=1 ## view downscaled DS_input_log=0 ## convert "b" var to log during downscaling

cache_lons1=[] cache_lats1=[] cache_x=[] cache_y=[] cache_z=[] cache_x2=[] cache_y2=[] cache_z2=[]

def probability_single_var( x1, y1, x2): print ("Specie habitation test")

xa1=np.array(x1) ya1=np.array(y1) xa2=np.array(x2) sha1=np.shape(x1) dim2=1 x=xa1.reshape((-1, dim2)) y=ya1.reshape((-1, 1)) xb=xa2.reshape((-1, dim2))

#y=y*0.0+1 y=x

#print(x) #print(y)

x_mean=np.mean(x) x_std=np.std(x) x_cover_std=(x-x_mean)/x_std

y_mean=np.mean(y) y_std=np.std(y) y_cover_std=(y-y_mean)/y_std

kohde=abs(x-x_mean)/x_std #print (kohde) #print (x-x_mean) #quit()

#plt.plot(x_cover_std)

#plt.show()

#model = LinearRegression().fit(x, kohde) #degree=3 #polyreg=make_pipeline(PolynomialFeatures(degree),LinearRegression()) #model=polyreg.fit(x,kohde) sc = StandardScaler() xa = sc.fit_transform(x) xba = sc.transform(xb)

#model = RandomForestRegressor(n_estimators=10, max_features=2).fit(xa,kohde) #model = LogisticGAM().fit(x, kohde) model = LinearGAM().fit(x, kohde)

y2= model.predict(xb)

#svm1 = svm.OneClassSVM(nu=0.1, kernel="rbf", gamma=0.5) #svm1 = svm.OneClassSVM(nu=0.1, kernel="rbf", gamma=0.5)

#model=svm1.fit(x,y)

#model=svm1.fit(x,kohde)

#y2= model.predict(xb)

y2_mean=np.mean(y2) y2_std=np.std(y2) y2_cover_std=(y2-y2_mean)/y2_std

y3=y2_cover_std

#y5=scipy.stats.norm.cdf(y3,y2_mean,y2_std) y4=scipy.stats.norm.sf(y3,y2_mean,y2_std)

ymin=np.min(y4) deltamax=np.max(y4)-np.min(y4)

y5=((y4-ymin)/deltamax) #zgrid1=np.array(zoutvalue1).reshape(1000, 400) #plt.plot(y5) #cb = plt.scatter(np.arange(1,1000), np.arange(1,400), s=60, c=zoutvalue1)

#plt.show()

#print(np.shape(y5[0])) #stop(-1)

return(y5)

def load_xy(infname1, lonname1, latname1): global cache_lons1 global cache_lats1 global cache_x global cache_y global cache_z global cache_x2 global cache_y2 global cache_z2

indf1=pd.read_csv(infname1, sep=';') #print(indf1)

templons1=indf1[lonname1] templats1=indf1[latname1]

cache_lons1=np.array(templons1) cache_lats1=np.array(templats1)

#print (cache_lons1) #print (cache_lats1)

return(0)

def preprocess_big_raster(infilename1, invarname1, sabluname1, outfilename1, soutfilename1, outvarname1,lon1, lon2, lat1, lat2, width1, height1, roto): gdal_cut_resize_fromnc_tonc(infilename1, sabluname1, outfilename1, invarname1,lon1, lon2, lat1, lat2) gdal_cut_resize_tonc(infilename1,soutfilename1 ,lon1, lon2, lat1, lat2, width1, height1) return(0)

def preprocess_small_single_raster(infilename1, invarname1, outfilename1, outvarname1,lon1, lon2, lat1, lat2, roto): print(infilename1) tempfilename1="./process/temp00.nc" tempfilename2="./process/temp01.nc" loadnetcdf_single_tofile(infilename1, invarname1, tempfilename1, outvarname1) #rotate_netcdf_360_to_180(tempfilename1, outvarname1,tempfilename2, outvarname1) gdal_cut_tonc(tempfilename1,outfilename1 ,lon1, lon2, lat1, lat2) return(0)

def preprocess_small_timed_raster(infilename1, invarname1, intime1, outfilename1, outvarname1,lon1, lon2, lat1, lat2, roto): tempfilename1="./process/temp00.nc" tempfilename2="./process/temp01.nc" loadnetcdf_timed_tofile(infilename1, invarname1, intime1, tempfilename1, outvarname1) rotate_netcdf_360_to_180(tempfilename1, outvarname1,tempfilename2, outvarname1) gdal_cut_tonc(tempfilename2,outfilename1 ,lon1, lon2, lat1, lat2) return(0)

def rotate_netcdf_360_to_180(infilename1, invarname1,outfilename1, outvarname1): global cache_lons1 global cache_lats1 global cache_x global cache_y global cache_z pointsx1=[] pointsy1=[] gdal_get_nc_to_xyz_in_mem(infilename1, invarname1 ) lonsa=cache_lons1 latsa=cache_lats1 pointsz1=np.array(cache_z) pointsx1=np.array(cache_x) pointsy1=np.array(cache_y) nlatsa1=len(latsa) nlonsa1=len(lonsa) pointsz3=np.reshape(pointsz1, (nlatsa1, nlonsa1)) rama1=int(len(lonsa)/2) pointsz4=np.roll(pointsz3,rama1,axis=1) lonsb=lonsa-180 save_points_to_netcdf(outfilename1, outvarname1, lonsb, latsa, pointsz4) return(0)

def gdal_get_nc_to_xyz_in_mem(inname1, invar1): global cache_lons1 global cache_lats1 global cache_x global cache_y global cache_z global cache_x2 global cache_y2 global cache_z2

imga=loadnetcdf_single_tomem(inname1, invar1) #plt.imshow(imga) lonsa=cache_lons1 latsa=cache_lats1

cache_lons1=[] cache_lats1=[] cache_x=[] cache_y=[] cache_z=[] cache_x2=[] cache_y2=[] cache_z2=[]

dim1=imga.shape nlat1=len(latsa) nlon1=len(lonsa)

#plt.plot(imga) #plt.show() #print(nlat1) #print(nlon1)

#quit(-1) #print(inname1) #print (nlat1)

#quit(-1)

for iy in range(0,nlat1): for ix in range(0,nlon1): pz1=imga[iy,ix] if (str(pz1) == '--'): cache_x.append(lonsa[ix]) cache_y.append(latsa[iy]) cache_z.append(0) else: cache_x.append(lonsa[ix]) cache_y.append(latsa[iy]) cache_z.append(pz1)

#print(cache_z) cache_lons1=lonsa cache_lats1=latsa

#print (pz1) return (cache_z)

def average_tables(daata1, daata2): daata3=daata1 pitu=len(daata1) for n in range(0,pitu): daata3[n]=(daata1[n]+daata2[n])/2

return(daata3)

def add_scalar(daata, skalar): pitu=len(daata) for n in range(0,pitu): daata[n]=daata[n]+skalar

return(daata)

def gam_multiple_vars( x1, y1, x2): print ("GAM") xa1=np.array(x1) ya1=np.array(y1) xa2=np.array(x2)

#print (xa1)

1. quit(-1)

sha1=np.shape(x1)

dim2=sha1[1]

x=xa1.reshape((-1, dim2)) y=ya1.reshape((-1, 1)) xb=xa2.reshape((-1, dim2))

#sc = StandardScaler() #xa = sc.fit_transform(x) #xba = sc.transform(xb)

#model = LogisticGAM().fit(x, y) model = LinearGAM().fit(x, y)

y2= model.predict(xb)

return(y2)

def random_forest_multiple_vars( x1, y1, x2): print ("RF")

global RF_estimators1 global RF_features1

print(RF_estimators1, RF_features1)

#quit(-1)

xa1=np.array(x1) ya1=np.array(y1) xa2=np.array(x2)

#print (xa1)

1. quit(-1)

sha1=np.shape(x1)

dim2=sha1[1]

x=xa1.reshape((-1, dim2)) y=ya1.reshape((-1, 1)) xb=xa2.reshape((-1, dim2))

#model = LinearRegression().fit(x, y) #degree=3 #polyreg=make_pipeline(PolynomialFeatures(degree),LinearRegression()) #model=polyreg.fit(x,y)

sc = StandardScaler() xa = sc.fit_transform(x) xba = sc.transform(xb)

## orig ##model = RandomForestRegressor(n_estimators=10, max_features=2).fit(xa,y) model = RandomForestRegressor(n_estimators=RF_estimators1, max_features=RF_features1).fit(xa,y)

y2= model.predict(xba) return(y2)

def save_points_toxyz(filename, x,y,z): print("Dummy function only") return(0)

def linear_regression_multiple_vars( x1, y1, x2): print ("MLR") xa1=np.array(x1) ya1=np.array(y1) xa2=np.array(x2)

sha1=np.shape(x1)

dim2=sha1[1]

x=xa1.reshape((-1, dim2)) y=ya1.reshape((-1, 1)) xb=xa2.reshape((-1, dim2))

#model = LinearRegression().fit(x, y) degree=3 polyreg=make_pipeline(PolynomialFeatures(degree),LinearRegression()) model=polyreg.fit(x,y)

y2= model.predict(xb) return(y2)

def linear_regression_singe_var( x1, y1, x2): #print (x1) #print(y1) #return(0)

#xa1=np.array(x1) #ya1=np.array(y1) #xa2=np.array(x2) xa1=np.array( x1) ya1=np.array(y1) xa2=np.array(x2)

sha1=np.shape(x1)

dim2=sha1[1]

x=xa1.reshape((-1, dim2)) y=ya1.reshape((-1, 1)) xb=xa2.reshape((-1, dim2)) #x=xa1.reshape((-1, 1)) #y=ya1.reshape((-1, 1)) #xb=xa2.reshape((-1, 1))

#print (x) #print (y)

model = LinearRegression().fit(x, y) #model = RandomForestRegressor(n_estimators=10, max_features=2).fit(x,y) #degree=2 #polyreg=make_pipeline(PolynomialFeatures(degree),LinearRegression()) #polyreg=make_pipeline(PolynomialFeatures(degree), ) #model=polyreg.fit(x,y)

## warning not xb y2= model.predict(xb) #print(y2) #print("LR") return(y2)

def save_points_to_netcdf(outfilename1, outvarname1, xoutlons1, xoutlats1, zoutvalue1):

nlat1=len(xoutlats1) nlon1=len(xoutlons1) #print ("Save ...") #print (nlat1) #print (nlon1) zoutvalue2=np.array(zoutvalue1).reshape(nlat1, nlon1) #indata_set1=indata1 ncout1 = nc.Dataset(outfilename1, 'w', format='NETCDF4') outlat1 = ncout1.createDimension('lat', nlat1) outlon1 = ncout1.createDimension('lon', nlon1) outlats1 = ncout1.createVariable('lat', 'f4', ('lat',)) outlons1 = ncout1.createVariable('lon', 'f4', ('lon',)) outvalue1 = ncout1.createVariable(outvarname1, 'f4', ('lat', 'lon',)) outvalue1.units = 'Unknown' outlats1[:] = xoutlats1 outlons1[:] = xoutlons1 outvalue1[:, :] =zoutvalue2[:] ncout1.close() return 0

def gdal_cut_resize_fromnc_tonc(inname1, inname2, outname2, invar2,lon1, lon2, lat1, lat2): imga=loadnetcdf_single_tomem(inname2, invar2) dim1=imga.shape height1=dim1[0] width1=dim1[1] print (width1) print (height1) jono1="gdalwarp -te "+str(lon1)+" "+str(lat1)+" "+str(lon2)+" "+str(lat2)+" "+"-ts "+str(width1)+" "+str(height1)+ " " +inname1+" "+outname2 print(jono1) os.system(jono1) return

def gdal_get_points_from_file(inname1, invar1,pointsx1,pointsy1): global cache_lons1 global cache_lats1 global cache_x global cache_y global cache_z global cache_x2 global cache_y2 global cache_z2

imga=loadnetcdf_single_tomem(inname1, invar1) #plt.imshow(imga) lonsa=cache_lons1 latsa=cache_lats1

cache_lons1=[] cache_lats1=[] cache_x=[] cache_y=[] cache_z=[] cache_x2=[] cache_y2=[] cache_z2=[]

dim1=imga.shape nlat1=dim1[0] nlon1=dim1[1]

pitu=len(pointsx1) #px1=10 #py1=45 for n in range(0,pitu): px1=pointsx1[n] py1=pointsy1[n] #print('.') for iy in range(0,nlat1): if(py1>=latsa[iy]): for ix in range(0,nlon1): if(px1>=lonsa[ix]): pz1=imga[iy,ix] cache_x.append(lonsa[ix]) cache_y.append(latsa[iy]) cache_z.append(pz1)

#print(cache_z) cache_lons1=lonsa cache_lats1=latsa

#print (pz1) return (cache_z)

def gdal_cut_resize_tonc(inname1, outname1, lon1, lon2, lat1, lat2, width1, height1): #gdalwarp -te -5 41 10 51 -ts 1000 0 input.tif output.tif jono1="gdalwarp -of netcdf -te "+str(lon1)+" "+str(lat1)+" "+str(lon2)+" "+str(lat2)+" "+"-ts "+str(width1)+" "+str(height1)+ " " +inname1+" "+outname1 print(jono1) os.system(jono1) return

def interpolate_cache(x_min, y_min, x_max, y_max, reso1): global cache_lons1 global cache_lats1 global cache_x global cache_y global cache_z global cache_x2 global cache_y2 global cache_z2

cache_x2=[] cache_y2=[] cache_z2=[] cache_lons1=[] cache_lats1=[]

pitu1=len(cache_z)

raja1=pitu1

for i in range(0,raja1): #print (i) #print (cache_z[i]) try: xx=cache_x[i] yy=cache_y[i] zz=cache_z[i] if (str(zz) == '--'): raja1=raja1-1 #print (zz) #print(".") else: cache_x2.append(xx) cache_y2.append(yy) cache_z2.append(zz) except IndexError: print("BRK") break

lonsn=(int(x_max-x_min)/reso1) latsn=(int(y_max-y_min)/reso1) cache_lons1=np.linspace(x_min, x_max, lonsn) cache_lats1=np.linspace(y_min, y_max, latsn)

#print (cache_z2) #print (cache_x2) #exit(-1)

grid_x, grid_y = np.mgrid[x_min:x_max:reso1, y_min:y_max:reso1] rbfi = Rbf(cache_x2, cache_y2, cache_z2) di = rbfi(grid_x, grid_y) #plt.figure(figsize=(15, 15)) #plt.imshow(di.T, origin="lower") #cb = plt.scatter(df.x, df.y, s=60, c=df.por, edgecolor='#ffffff66') #plt.colorbar(cb, shrink=0.67) #plt.show() return(di.T)

def makepoints(imgin1, lonsin1, latsin1): global cache_x global cache_y global cache_z cache_x=[] cache_y=[] cache_z=[] dim1=imgin1.shape nlat1=dim1[0] nlon1=dim1[1] k=0 for iy in range(0,nlat1): for ix in range(0,nlon1): zz=imgin1[iy,ix] cache_x.append(lonsin1[ix]) cache_y.append(latsin1[iy]) if (str(zz) == '--'): ## warning there 0 append to equalize grid cache_z.append(0) k=1 else: cache_z.append(zz)

#cache_z.append(imgin1[iy,ix])

return(0)

def gdal_reso(inname1, outname1, reso1): jono1="gdalwarp -tr "+str(reso1)+" "+str(reso1)+" "+inname1+" "+outname1 os.system(jono1) return(0)

def gdal_cut_tonc(inname1, outname1, lon1, lon2, lat1, lat2):

jono1="gdalwarp -te "+str(lon1)+" "+str(lat1)+" "+str(lon2)+" "+str(lat2)+" "+inname1+" "+outname1 print(jono1) os.system(jono1)

return(0)

def savenetcdf_single_frommem(outfilename1, outvarname1, xoutvalue1,xoutlats1,xoutlons1): nlat1=len(xoutlats1) nlon1=len(xoutlons1) #indata_set1=indata1 ncout1 = nc.Dataset(outfilename1, 'w', format='NETCDF4') outlat1 = ncout1.createDimension('lat', nlat1) outlon1 = ncout1.createDimension('lon', nlon1) outlats1 = ncout1.createVariable('lat', 'f4', ('lat',)) outlons1 = ncout1.createVariable('lon', 'f4', ('lon',)) outvalue1 = ncout1.createVariable(outvarname1, 'f4', ('lat', 'lon',)) outvalue1.units = 'Unknown' outlats1[:] = xoutlats1 outlons1[:] = xoutlons1 outvalue1[:, :] =xoutvalue1[:] ncout1.close() return 0

def loadnetcdf_single_tomem(infilename1, invarname1): global cache_lons1 global cache_lats1 print(infilename1) inc1 = nc.Dataset(infilename1) inlatname1="lat" inlonname1="lon" inlats1=inc1[inlatname1][:] inlons1=inc1[inlonname1][:] cache_lons1=inlons1 cache_lats1=inlats1 indata1_set1 = inc1[invarname1][:] dim1=indata1_set1.shape nlat1=dim1[0] nlon1=dim1[1] inc1.close() return (indata1_set1)

def loadnetcdf_single_tofile(infilename1, invarname1, outfilename1, outvarname1): inc1 = nc.Dataset(infilename1) inlatname1="lat" inlonname1="lon" inlats1=inc1[inlatname1][:] inlons1=inc1[inlonname1][:] indata1_set1 = inc1[invarname1][:] dim1=indata1_set1.shape nlat1=dim1[0] nlon1=dim1[1] #indata_set1=indata1 ncout1 = nc.Dataset(outfilename1, 'w', format='NETCDF4') outlat1 = ncout1.createDimension('lat', nlat1) outlon1 = ncout1.createDimension('lon', nlon1) outlats1 = ncout1.createVariable('lat', 'f4', ('lat',)) outlons1 = ncout1.createVariable('lon', 'f4', ('lon',)) outvalue1 = ncout1.createVariable(outvarname1, 'f4', ('lat', 'lon',)) outvalue1.units = 'Unknown' outlats1[:] = inlats1 outlons1[:] = inlons1 outvalue1[:, :] = indata1_set1[:] ncout1.close() return 0

def loadnetcdf_timed_tofile(infilename1, invarname1, intime1, outfilename1, outvarname1): inc1 = nc.Dataset(infilename1) inlatname1="lat" inlonname1="lon" inlats1=inc1[inlatname1][:] inlons1=inc1[inlonname1][:] indata1 = inc1[invarname1][:] dim1=indata1.shape nlat1=dim1[1] nlon1=dim1[2] indata_set1=indata1[intime1] #img01=indata_set1 #img1.replace(np.nan, 0, inplace=True) #img1 = np.where(isna(img10), 0, img10) #where_are_NaNs = np.isnan(img1) #img1[where_are_NaNs] = 99 #img1 = np.where(np.isnan(img01), 0, img01) ncout1 = nc.Dataset(outfilename1, 'w', format='NETCDF4') outlat1 = ncout1.createDimension('lat', nlat1) outlon1 = ncout1.createDimension('lon', nlon1) outlats1 = ncout1.createVariable('lat', 'f4', ('lat',)) outlons1 = ncout1.createVariable('lon', 'f4', ('lon',)) outvalue1 = ncout1.createVariable(outvarname1, 'f4', ('lat', 'lon',)) outvalue1.units = 'Unknown' outlats1[:] = inlats1 outlons1[:] = inlons1 outvalue1[:, :] = indata_set1 ncout1.close() return 0

1. 1. downscaler funktzione !

1. def downscale_data_1(input_rasters,loadfiles,intime1,lon1,lon2,lat1,lat2,width1,height1,roto,areaname,sresultfilename1,sresultvarname1,infilenames1, outfilenames1,soutfilenames1,invarnames1,outvarnames1, soutvarnames1):

def downscale_data_1(input_rasters,loadfiles,intime1,lon1,lon2,lat1,lat2,width1,height1,roto,areaname,sresultfilename1,sresultvarname1,infilenames1, invarnames1):

global DS_method global DS_show global DS_input_log

debug=0

#if(input_rasters==1): # os.system("del *.nc")

outfilenames1=[] outvarnames1=[] soutfilenames1=[] soutvarnames1=[]

huba0=len(infilenames1) for n in range (0,huba0): sandersson1=invarnames1[n] outvarnames1.append(sandersson1) soutvarnames1.append(sandersson1)

## big raster?? outfilenames1.append("./process/"+areaname+"_"+invarnames1[0]+".nc")

if(input_rasters==1): #preprocess_small_timed_raster(infilenames1[0], invarnames1[0], intime1, outfilenames1[0], outvarnames1[0], lon1, lon2, lat1, lat2,roto) preprocess_small_single_raster(infilenames1[0], invarnames1[0], outfilenames1[0], outvarnames1[0], lon1, lon2, lat1, lat2,roto)

#quit(-1)

huba=len(infilenames1) for n in range (1,huba): ofnamee="./process/"+areaname+"_"+outvarnames1[n]+"_"+str(n)+".nc" sofnamee="./process/"+areaname+"_"+outvarnames1[n]+"_"+str(n)+"_s.nc" print(ofnamee) print(sofnamee) outfilenames1.append(ofnamee) soutfilenames1.append(sofnamee)

if(input_rasters==1): print("PP ",infilenames1[n]) preprocess_big_raster(infilenames1[n], invarnames1[n], outfilenames1[0], outfilenames1[n], soutfilenames1[n-1], outvarnames1[n], lon1, lon2, lat1, lat2, width1, height1, roto)

imgs=[] pointsx=[] pointsy=[] pointsz=[] mlats=[] mlons=[]

spointsx=[] spointsy=[] spointsz=[] simgs=[] slats=[] slons=[]

len1=len(outfilenames1)

for n in range(0,len1): imgs.append(loadnetcdf_single_tomem(outfilenames1[n], "Band1")) mlons.append(cache_lons1) mlats.append(cache_lats1) makepoints(imgs[n], mlons[n], mlats[n]) pointsx.append(cache_x) pointsy.append(cache_y) pointsz.append(cache_z)

len1=len(soutfilenames1)

for n in range(0,len1): simgs.append(loadnetcdf_single_tomem(soutfilenames1[n], "Band1")) slons.append(cache_lons1) slats.append(cache_lats1) makepoints(simgs[n], slons[n], slats[n]) spointsx.append(cache_x) spointsy.append(cache_y) spointsz.append(cache_z)

if(debug==1): print("Specie habitation.") load_xy("humanlgm.txt","Lon", "Lat") klats1=cache_lats1 klons1=cache_lons1

spointszout=[] ppointsx=[] ppointsy=[] ppointsz=[]

gdal_get_points_from_file(outfilenames1[1], invarnames1[1], klons1, klats1) ppointsx.append(cache_x) ppointsy.append(cache_y) ppointsz.append(cache_z)

gdal_get_points_from_file(outfilenames1[2], invarnames1[2], klons1, klats1) ppointsx.append(cache_x) ppointsy.append(cache_y) ppointsz.append(cache_z)

#gdal_get_points_from_file(outfilenames1[2], invarnames1[2], klons1, klats1) #ppointsx.append(cache_x) #ppointsy.append(cache_y) #ppointsz.append(cache_z) #bpoints1=ppointsz[0] #apoints1=ppointsz[0] #cpoints1=spointsz[0]

bpoints1=ppointsz[0] apoints1=ppointsz[0] cpoints1=spointsz[0]

spointszout.append(probability_single_var(apoints1, bpoints1, cpoints1))

bpoints1=ppointsz[1] apoints1=ppointsz[1] cpoints1=spointsz[1]

spointszout.append(probability_single_var(apoints1, bpoints1, cpoints1))

#bpoints1=ppointsz[2] #apoints1=ppointsz[2] #cpoints1=spointsz[2]

#spointszout.append(probability_single_var(apoints1, bpoints1, cpoints1))

odex1=0 sdataout=spointszout[0]*spointszout[1]

pointsxout1=spointsx[0] pointsyout1=spointsy[0]

slonsout1=slons[0] slatsout1=slats[0]

save_points_to_netcdf(sresultfilename1, sresultvarname1, slonsout1, slatsout1, sdataout)

plt.imshow( np.array(sdataout).reshape(len(slatsout1), len(slonsout1) ) ) plt.show() return(1)

## main sektion of ds

sla=[]

len1=len(pointsz)

for n in range(1,len1): sla.append(pointsz[n])

slb=[]

for n in range(0,len1-1): print (n) slb.append(spointsz[n])

apoints1=list(zip(*sla)) bpoints1=pointsz[0] cpoints1=list(zip(*slb))

spointszout=[]

#if(DS_input_log==0):

#if(DS_input_log==1): # spointszout.append(np.exp(random_forest_multiple_vars(apoints1, np.log(bpoints1), cpoints1))) # spointszout.append(np.exp(gam_multiple_vars(apoints1, np.log(bpoints1), cpoints1))) # spointszout.append(np.exp(linear_regression_multiple_vars(apoints1, np.log(bpoints1), cpoints1)))

#sdataout=average_tables(spointszout[0],spointszout[1]) if(DS_method==0): spointszout.append(random_forest_multiple_vars(apoints1, bpoints1, cpoints1)) sdataout=spointszout[0] if(DS_method==1): spointszout.append(gam_multiple_vars(apoints1, bpoints1, cpoints1)) sdataout=spointszout[0] if(DS_method==2): spointszout.append(linear_regression_multiple_vars(apoints1, bpoints1, cpoints1)) sdataout=spointszout[0] if(DS_method==77): spointszout.append(random_forest_multiple_vars(apoints1, bpoints1, cpoints1)) spointszout.append(gam_multiple_vars(apoints1, bpoints1, cpoints1)) spointszout.append(linear_regression_multiple_vars(apoints1, bpoints1, cpoints1)) #sdataout=average_tables(spointszout[0],spointszout[1] ) #sdataout=average_tables(spointszout[0],spointszout[1], spointszout[2]) sdataout1=average_tables(spointszout[0],spointszout[1] ) sdataout2=average_tables(spointszout[1],spointszout[2] ) sdataout=average_tables(sdataout1, sdataout2 ) if(DS_method==88): spointszout.append(random_forest_multiple_vars(apoints1, bpoints1, cpoints1)) spointszout.append(gam_multiple_vars(apoints1, bpoints1, cpoints1)) spointszout.append(linear_regression_multiple_vars(apoints1, bpoints1, cpoints1)) #sdataout=average_tables(spointszout[0],spointszout[1] ) #sdataout=average_tables(spointszout[0],spointszout[1], spointszout[2]) sdataout1=average_tables(spointszout[0],spointszout[1] ) sdataout2=average_tables(10*spointszout[1],1*spointszout[2]/10 )/10 sdataout=average_tables(sdataout1, sdataout2 )

pointsxout1=spointsx[0] pointsyout1=spointsy[0] slonsout1=slons[0] slatsout1=slats[0] save_points_to_netcdf(sresultfilename1, sresultvarname1, slonsout1, slatsout1, sdataout)

#plt.register_cmap(name='viridis', cmap=cmaps.viridis) #plt.set_cmap(cm.viridis)

#cmaps = OrderedDict() #kolormap='jet' #kolormap='Spectral_r' #kolormap='gist_rainbow_r' #kolormap='BrBG' #kolormap='rainbow' kolormap='viridis_r'

if(DS_show==1): plt.imshow(np.array(sdataout).reshape(len(slatsout1), len(slonsout1)) , cmap=kolormap) plt.ylim(0, len(slatsout1)) plt.show()

return(0)

1. 1. attempt to post process rain in mountains!
   2. POST PROCESSORI warning experiment only

def manipulate_rainfall_data(demname, rainname, oname2): # try to exaggerate rainfall in mountains dem1=loadnetcdf_single_tomem(demname,"Band1") rain0=loadnetcdf_single_tomem(rainname,"Rain") rain1=np.flipud(rain0) shape1=np.shape(rain1) print(shape1)

dem2=np.ravel(dem1) rain2=np.ravel(rain1) len1=len(rain2) #print (len1) outta1=rain2*0 limith1=1200 for n in range(0,len1): r=rain2[n] o=r d=dem2[n] rk=d/limith1 if (d>limith1): o=r*rk outta1[n]=o

out00=np.reshape(outta1,shape1)

#out1=np.flipud(out00) out1=out00

savenetcdf_single_frommem(oname2, "Rain", out1,cache_lats1, cache_lons1) return(out1)

def match_raster(iname,matchname, oname): loadnetcdf_single_tomem(iname, "Band1") lon1=np.min(cache_lons1) lon2=np.max(cache_lons1) lat1=np.min(cache_lats1) lat2=np.max(cache_lats1) gdal_cut_resize_fromnc_tonc(iname, matchname, oname, "Rain",lon1, lon2, lat1, lat2)

1. 1. attempt to post process rain in mountains!
   2. POST PROCESSORI warning experiment 2 only
   3. enhances big rainfalls, diminishes small rainfalls
   4. based on y=kx+b

def manipulate_rainfall_data_2(dname1, dname2, oname2, ofset1, ofset2, k1): # try to exaggerate rainfall in mountains dat10=loadnetcdf_single_tomem(dname1,"Band1") dat20=loadnetcdf_single_tomem(dname2,"Rain") shape1=np.shape(dat10)

dee1=np.ravel(dat10) dee2=np.ravel(dat20)

len1=len(dee1) len2=len(dee2)

outta1=dee2*0

a=ofset1 b=ofset2 k=k1

for n in range(0,len1): d1=dee1[n] d2=dee2[n] o0=(d1+d2)/2 o1=o0+b o2=(o1-a)*k o3=o2+a outta1[n]=o3

out00=np.reshape(outta1,shape1) #out1=np.flipud(out00) out1=out00

if(DS_show==1): kolormap='viridis_r' plt.imshow(out1, cmap=kolormap) #plt.ylim(0, len(slatsout1)) plt.show()

savenetcdf_single_frommem(oname2, "Rain", out1,cache_lats1, cache_lons1) return(out1)

1. 1. 1. 1. 1. 1. 2. main program

input_rasters=1 debug=0 ## human habitation test

1. 1. acquire basic data, loadfiles=1

loadfiles=1

intime1=1

1. lon1=-4.5
2. lon2=10.0
3. lat1=42.0
4. lat2=50.5

lon1=-180 lon2=180 lat1=-90 lat2=90

width1=640 height1=320

1. jn
2. width1=112
3. height1=130

1. JN WARNING
2. width1=200
3. height1=200

roto=1

areaname="planet"

sresultfilename1="./output/dskaled1.nc" sresultvarname1="atm_temp"

infilenames1=[] invarnames1=[]

infilenames1.append('atm_temp.nc')

1. infilenames1.append('./indata/Map19_PALEOMAP_1deg_Late_Cretaceous_80Ma.nc')

infilenames1.append('dem.nc') infilenames1.append('lons.nc') infilenames1.append('lats.nc')

1. infilenames1.append('./process/accu_dem_1.nc')
2. infilenames1.append('./process/accu_windir.nc')

invarnames1.append("atm_temp") invarnames1.append("Band1") invarnames1.append("Band1") invarnames1.append("Band1")

RF_estimators1=100

1. RF_features1=4

RF_features1=1

DS_method=2 DS_show=1 DS_input_log=0

downscale_data_1(input_rasters,loadfiles,intime1,lon1,lon2,lat1,lat2,width1,height1,roto,areaname,sresultfilename1,sresultvarname1,infilenames1, invarnames1)

print(".")

quit(-1)

1. ext1<- extent(0, 360, -90, 90)

ext1<- extent(-180, 180, -90, 90)

extent(r1) <- ext1

str(r1)

ur1<-r1

1. ur2<-rotate(r1)

1. extent(ur2) <- ext1

1. sabluna1<-raster(nrows=360, ncols=360, xmn=0, xmx=180, ymn=-90.5, ymx=90.5)
2. sabluna1<-raster(nrows=360, ncols=360)

1. extent(sabluna1) <- ext1

1. ur1<-resample(ur2,sabluna1)

1. extent(ur1) <- ext1

1. image(ur1)

crs(ur1)<-"+proj=longlat +datum=WGS84 +no_defs +ellps=WGS84 +towgs84=0,0,0"

writeRaster(ur1, "./output/dskaled2.nc", overwrite=TRUE, format="CDF", varname="T", varunit="degC",

       longname="T", xname="lon",   yname="lat")