TC MSE POD Proposal for Incorporation into NOAA's Repo

diagnostics/TC_MSE/Binning_and_compositing.py

@@ -0,0 +1,170 @@
+#Import modules
+import os
+import numpy as np
+import xarray as xr
+
+##########  BINNING/COMPOSITING MODEL DATA #############################################################
+
+######################## MATH FUNCTION(S) ############################################
+def boxavg(thing,lat,lon):
+    coslat_values = np.transpose(np.tile(np.cos(np.deg2rad(lat)),(len(lon),1)))
+    thing1 = thing*coslat_values
+    thing2 = thing1/thing1
+    average = np.nansum(np.nansum(thing1,0))/np.nansum(np.nansum(coslat_values*thing2,0))
+
+    return average
+
+#Lats/Lons
+latres = np.float(os.getenv("latres"))
+lonres = np.float(os.getenv("lonres"))
+lats = np.arange(-5,5+latres,latres)
+lons = np.arange(-5,5+lonres,lonres)
+#Gather the years that were inputted by user
+FIRST_YR = np.int(os.getenv("FIRSTYR"))
+LAST_YR = np.int(os.getenv("LASTYR"))
+ds_all = []
+
+for y in range(FIRST_YR,LAST_YR+1):
+    #Open all the yearly snapshot (budget and regular variable) files
+    ds_reg = xr.open_dataset(os.environ['WK_DIR']+'/model/Model_Regular_Variables_'+str(y)+'.nc')
+    ds_budg = xr.open_dataset(os.environ['WK_DIR']+'/model/Model_Budget_Variables_'+str(y)+'.nc')
+    #Merge the budget and regular variable files by year
+    ds_merge = xr.merge([ds_reg,ds_budg])
+    ds_reg.close()
+    ds_budg.close()
+    #Get all the merged files together so that once all are collected they can be concatenated
+    ds_all.append(ds_merge)
+    ds_merge.close()
+
+#Concatenate the year files together so all variables are combined across all storms
+data = xr.concat(ds_all,dim='numstorms')
+
+#Get a list of the data variables in data to trim the data after lifetime maximum intensity (LMI)
+Model_vars = list(data.keys())
+
+#Grab the vmax variable to get the LMI itself and point of LMI for trimming to account only for intensification period
+maxwinds = data['maxwind']
+winds_list = []
+
+#Loop through the variables to pick out the feedbacks and add a normalized version of that variable
+for var in Model_vars:
+    if(var[0:5]=='hanom' or var[0:10]=='hMoistanom' or var[0:10]=='hTempanom' or var[0:4]=='hvar'):
+        normvar = np.array(data[var])
+        boxavrawvar = np.array(data[var])
+        boxavvar = np.ones((len(maxwinds),len(maxwinds[0]))) * np.nan
+        boxavnormvar = np.ones((len(maxwinds),len(maxwinds[0]))) * np.nan
+        for s in range(len(maxwinds)):
+            for t in range(len(maxwinds[s])):
+                hvar = np.array(data.hvar[s][t][:][:])
+                boxavghvar = boxavg(hvar,np.array(data.latitude[s][t][:]),np.array(data.longitude[s][t][:]))
+                normvar[s][t][:][:] = normvar[s][t][:][:]/boxavghvar
+                boxavvar[s][t] = boxavg(boxavrawvar[s][t][:][:],np.array(data.latitude[s][t][:]),np.array(data.longitude[s][t][:]))
+                boxavnormvar[s][t] = boxavg(np.array(normvar[s][t][:][:]),np.array(data.latitude[s][t][:]),np.array(data.longitude[s][t][:]))
+        data['norm'+var] = (['numstorms','numsteps','latlen','lonlen'],np.array(normvar[:][:][:][:]))
+        data['boxav_'+var] = (['numstorms','numsteps'],np.array(boxavvar[:][:]))
+        data['boxav_norm_'+var] = (['numstorms','numsteps'],np.array(boxavnormvar[:][:]))
+
+#Loop through all model storms and find the LMI, then tag each storm with its LMI for binning later
+for s in range(len(maxwinds)):
+    windmax = float(max(maxwinds[s][:]))
+    windmaxindex = np.squeeze(np.where(maxwinds[s]==windmax))
+    #Check if there are more than one maximum wind speed
+    if windmaxindex.size >= 2:
+        windmaxindex = int(windmaxindex[0])
+    else:
+        windmaxindex = int(np.squeeze(np.where(maxwinds[s]==windmax)))
+    #Loop and have all the indices after the timestep of LMI be NaN for all vars
+    for var in Model_vars:
+        data[var][s,windmaxindex+1:len(maxwinds[s])+1] = np.nan
+
+    vmax_indiv_list = []
+    for t in range(0,len(maxwinds[s])):
+        #First check and NaN all variables at timesteps where TC center is outside 30 N/S 
+        if(data.centerLat[s][t]>30 or data.centerLat[s][t]<-30):
+            for var in Model_vars:
+                if(data[var].ndim==2):
+                    data[var][s][t] = np.nan
+                elif(data[var].ndim==3):
+                    data[var][s][t][:] = np.nan
+                else:
+                    data[var][s][t][:][:] = np.nan
+         #Get max wind at specific step to tag the steps for binning snapshot
+        vmax_sel = maxwinds[s,t].values
+        vmax = xr.full_like(data.h[s,t],float(vmax_sel)).rename('vmax')
+        vmax_indiv_list.append(vmax)
+    vmax_indiv_array = xr.concat(vmax_indiv_list,dim='numsteps')
+    #Create the vmax tag variable
+    winds_list.append(vmax_indiv_array)
+
+#Update Model data with the vmax tag created above
+model_winds_array = xr.concat(winds_list,dim='numstorms')
+model_updated = xr.merge([data,model_winds_array])
+#Stretch the boxav variables to 1 dimension and make a new stretched windmax variable
+newvars = list(model_updated.keys())
+for var in newvars:
+    if(var[0:5]=='boxav'):
+        model_updated['new_'+var] = (['newsteps'],np.squeeze(np.reshape(np.array(model_updated[var]),(len(data.numstorms)*len(data.numsteps)))))
+
+model_updated['new_maxwind'] = (['newsteps'],np.squeeze(np.reshape(np.array(model_updated['maxwind']),(len(data.numstorms)*len(data.numsteps)))))
+
+#Bin snapshots according to max wind speed bins
+bins = np.arange(0,66,3)
+#Set a count array to gather the sample size for each bin and all bins
+count_denom = len(data.latitude[0][0]) * len(data.longitude[0][0])
+bins_count = np.zeros(len(bins))
+vmax2 = model_updated.vmax.copy(deep=True)
+onedvmax = model_updated.new_maxwind.copy(deep=True)
+for b, bin in enumerate(bins):
+    upperbin = bin+3
+    #Variable to get the number of samples for the current bin (divide by the resolution dims multiplied together)
+    count = (len(np.where((model_updated.vmax>=bin)&(model_updated.vmax<upperbin))[0])/count_denom)
+    bins_count[b] = count
+    vmax2 = (xr.where((model_updated.vmax>=bin)&(model_updated.vmax<upperbin), b, vmax2))
+    onedvmax = (xr.where((model_updated.new_maxwind>=bin)&(model_updated.new_maxwind<upperbin), b, onedvmax))
+bin_ds = xr.Dataset(data_vars=dict(bins=(['numstorms','numsteps','latlen','lonlen'],vmax2.values)))
+onedbin_ds = xr.Dataset(data_vars=dict(newbins=(['newsteps'],onedvmax.values)))
+ds = xr.merge([model_updated,bin_ds,onedbin_ds])
+ds = ds.set_coords(['bins'])
+ds = ds.set_coords(['newbins'])
+
+#Get the mean of each bin for one composite image
+bins = np.arange(0,22,1)
+binlabels = np.arange(1.5,66,3)
+dsbins = ds['bins'].values
+dsnewbins = ds['newbins'].values
+binmeans = {}
+binboxavstdevs = {}
+for var_name,values in ds.items():
+    dvar = ds[var_name].values
+    if(len(np.shape(dvar))==4):
+        avg_bin_list = []
+        for b,bin in enumerate(bins):
+            avg_bin_list.append(np.nanmean(np.where(dsbins==bin,np.array(dvar),np.nan),axis=(0,1)))
+        binmeans[var_name] = (['bin','lat','lon'],avg_bin_list)
+    if(len(np.shape(dvar))==1):
+        avg_bin_list = []
+        stdev_bin_list = []
+        for b,bin in enumerate(bins):
+            avg_bin_list.append(np.nanmean(np.where(dsnewbins==bin,np.array(dvar),np.nan),axis=(0)))
+            stdev_bin_list.append(np.nanstd(np.where(dsnewbins==bin,np.array(dvar),np.nan),axis=(0)))
+        binmeans[var_name] = (['bin'],avg_bin_list)
+        binboxavstdevs[var_name] = (['bin'],stdev_bin_list)
+
+#Bin means
+binmeans['bin'] = (['bin'],binlabels)
+binmeans['lat'] = (['lat'],lats)
+binmeans['lon'] = (['lon'],lons)
+binmeans['bincounts'] = ('bin',bins_count)
+#Add relevant raw variables that have not been binned or composited
+binmeans['maxwind'] = (['numstorms','numsteps'],np.array(data['maxwind']))
+binmeans['minSLP'] = (['numstorms','numsteps'],np.array(data['minSLP']))
+
+#Binned boxav stdevs
+binboxavstdevs['bin'] = (['bin'],binlabels)
+binboxavstdevs['bincounts'] = ('bin',bins_count)
+
+modelbinavgdata = xr.Dataset(data_vars=binmeans, attrs={'description':'Mean Binned Data'})
+modelbinboxavstdevdata = xr.Dataset(data_vars=binboxavstdevs, attrs={'description':'Standard Deviations of Binned Box Averaged Variables'})
+
+modelbinavgdata.to_netcdf(os.environ['WK_DIR']+ '/model/Model_Binned_Composites.nc')
+modelbinboxavstdevdata.to_netcdf(os.environ['WK_DIR']+ '/model/Model_Binned_STDEVS_of_BoxAvgs.nc')

diagnostics/TC_MSE/Plotting.py

@@ -0,0 +1,19 @@
+#Import Plotting Functions
+import Plotting_Functions as plot
+
+# Leave or comment out which plots you do not want to make
+
+# This will plot the spatial composite panels (4 plot files saved)
+plot.SpatialCompositePanels()
+
+# This will plot the azimuthal mean line plots (1 plot file saved)
+plot.AzmeanPlotting()
+
+# This will plot the non-normalized and normalized box-averaged feedbacks 
+# as a function of bin. (2 plot files saved)
+plot.BoxAvLinePlotting()
+
+# This will plot the scattering of non-normalized and normalized
+# box-averaged feedbacks and percent of storms intensifying from one
+# bin to the next. (1 plot file saved)
+plot.BoxAvScatter()