Python转换WRF输出数据为网页可视化JSON格式

纸上星河

发布时间：2026-10-16| 5人浏览过

原创

本文介绍如何将WRF模式输出的非等间距风场数据转换为网页可用的JSON格式。通过NCL脚本进行网格插值生成NetCDF文件，再利用Python脚本读取并转换为包含经纬度信息的JSON数据，最终实现网页端的动态风场展示。

前言

通常网页可视化风场数据采用JSON格式。若需在网页中展示WRF模式模拟的风场，需先将WRF输出数据转换为网页可识别的JSON格式。
主要依赖json库，将读取的风场U、V分量转换为字典并存储至JSON文件。
由于WRF模拟数据通常为不规则网格，需先进行插值处理，将其转换为等间距网格。此过程可通过NCL函数rcm2rgrid_Wrap实现。

例如，将模式中设置为兰伯特投影的网格进行转换。

插值后的等间距网格示例。

主要编程工作分为两部分：

第一部分使用NCL脚本将WRFOUT数据转换为等间距网格，并导出为NetCDF格式。
第二部分使用Python脚本读取第一步导出的NetCDF文件，并将其转换为JSON格式保存。

NCL插值脚本1

需修改路径和变量。以下脚本不仅包含风场U、V，还包含降水、海表面压力、气温等，可根据需要自行修改。

begin
  a = addfile("/Users/WRF/outdata/2022071000/wrfout_d01_2022-07-10_01:00:00","r")
  lat2d = a->XLAT(0,:,:)
  lon2d = a->XLONG(0,:,:)
  lat1d = lat2d(:,0)
  lon1d = lon2d(0,:)
 
  time = wrf_user_getvar(a,"XTIME",-1)
  u10 = wrf_user_getvar(a,"U10",0)
  v10 = wrf_user_getvar(a,"V10",0)
  slp = wrf_user_getvar(a,"slp",0)
  t2  = wrf_user_getvar(a,"T2",0)
  td  = wrf_user_getvar(a,"td",0)
  rainc = wrf_user_getvar(a,"RAINC",0)
  rainnc = wrf_user_getvar(a,"RAINNC",0)
  
  u10@lat2d = lat2d
  u10@lon2d = lon2d
  u10_ip = rcm2rgrid_Wrap(lat2d,lon2d,u10,lat1d,lon1d,0)
  
  v10@lat2d = lat2d
  v10@lon2d = lon2d
  v10_ip = rcm2rgrid_Wrap(lat2d,lon2d,v10,lat1d,lon1d,0)
  
  slp_ip  =   rcm2rgrid_Wrap(lat2d,lon2d,slp,lat1d,lon1d,0)
  t2_ip  =   rcm2rgrid_Wrap(lat2d,lon2d,t2,lat1d,lon1d,0)
  td_ip  =   rcm2rgrid_Wrap(lat2d,lon2d,td,lat1d,lon1d,0)
  rainc_ip  =   rcm2rgrid_Wrap(lat2d,lon2d,rainc,lat1d,lon1d,0)
  rainnc_ip  =   rcm2rgrid_Wrap(lat2d,lon2d,rainnc,lat1d,lon1d,0)
  outf = addfile("/Users/wrfout_d01_2022-07-10_01:00:00.nc","c")
  outf->time =  time
  outf->lat  =  lat2d
  outf->lon  =  lon2d
  outf->u10  =  u10_ip
  outf->v10  =  v10_ip
  outf->slp  =  slp_ip
  outf->t2   =  t2_ip
  outf->td   =  td_ip
  outf->rainc   =  rainc_ip
  outf->rainnc  =  rainnc_ip
end

该脚本的局限在于仅能基于模式模拟的固定经纬度区域进行插值，无法动态调整范围。

NCL插值脚本2

NCL还提供ESMF_regrid函数以实现上述过程。其优势在于支持任意经纬度范围的插值，但不足是暂时无法读取含高度层变量的数据。

（也可能因个人技术限制未能发现其他方法）。附相关脚本如下：

load "$NCARG_ROOT/lib/ncarg/nclscripts/esmf/ESMF_regridding.ncl"

begin
  a = addfile("/Users/WRF/outdata/2022071000/wrfout_d01_2022-07-10_01:00:00","r")
  u10 = wrf_user_getvar(a,"U10",0)
  v10 = wrf_user_getvar(a,"V10",0)
  slp = wrf_user_getvar(a,"slp",0)
  t2  = wrf_user_getvar(a,"T2",0)
;  td  = wrf_user_getvar(a,"td",0)
  rainc = wrf_user_getvar(a,"RAINC",0)
  rainnc = wrf_user_getvar(a,"RAINNC",0)
  
  u10@lat2d = a->XLAT(0,:,:) 
  u10@lon2d = a->XLONG(0,:,:)
  v10@lat2d = a->XLAT(0,:,:) 
  v10@lon2d = a->XLONG(0,:,:)
  slp@lat2d = a->XLAT(0,:,:) 
  slp@lon2d = a->XLONG(0,:,:)
  t2@lat2d = a->XLAT(0,:,:) 
  t2@lon2d = a->XLONG(0,:,:)
;  td@lat2d = a->XLAT(0,:,:) 
;  td@lon2d = a->XLONG(0,:,:)
  rainc@lat2d = a->XLAT(0,:,:) 
  rainc@lon2d = a->XLONG(0,:,:)
  rainnc@lat2d = a->XLAT(0,:,:) 
  rainnc@lon2d = a->XLONG(0,:,:)
  
  lat2d = a->XLAT(0,:,:)
  lon2d = a->XLONG(0,:,:)
  lat1d = lat2d(:,0)
  lon1d = lon2d(0,:)
  latS = -20
  latN = 50
  lonW = 95
  lonE = 145

  Opt = True
  Opt@InterpMethod = "bilinear" 
  Opt@ForceOverwrite = True 
  
  Opt@SrcMask2D = where(.not. ismissing(v10),1,0) 
  Opt@DstGridType = "0.1deg"
  Opt@DstLLCorner = (/latS, lonW /) 
  Opt@DstURCorner = (/latN, lonE /) 
  
  u10_regrid = ESMF_regrid(u10,Opt)
  v10_regrid = ESMF_regrid(v10,Opt)
  slp_regrid = ESMF_regrid(slp,Opt)
  t2_regrid = ESMF_regrid(t2,Opt)
;  td_regrid = ESMF_regrid(td,Opt)
  rainc_regrid = ESMF_regrid(rainc,Opt)
  rainnc_regrid = ESMF_regrid(rainnc,Opt)
  
  time = wrf_user_getvar(a,"XTIME",-1)
  
  nlon = dimsizes(v10_regrid&lon)
  nlat = dimsizes(v10_regrid&lat)
  
  ofile = "wrfout_d01_2022-07-10_01:00:00.nc"
  system("rm -rf "+ofile) 
  fout = addfile(ofile,"c") 
  
  dimNames = (/"lat", "lon"/)
  dimSizes = (/nlat, nlon/)
  dimUnlim = (/False, False/)
  
  filedimdef(fout,dimNames,dimSizes,dimUnlim) ;-- define dimensions
  
  filevardef(fout,"lat",typeof(v10_regrid&lat),getvardims(v10_regrid&lat))
  filevardef(fout,"lon",typeof(v10_regrid&lon),getvardims(v10_regrid&lon))
  
  filevardef(fout,"u10",typeof(u10_regrid),getvardims(u10_regrid))
  filevardef(fout,"v10",typeof(v10_regrid),getvardims(v10_regrid))
  filevardef(fout,"slp",typeof(slp_regrid),getvardims(slp_regrid))
  filevardef(fout,"t2",typeof(t2_regrid),getvardims(t2_regrid))
;  filevardef(fout,"td",typeof(td_regrid),getvardims(td_regrid))
  filevardef(fout,"rainc",typeof(rainc_regrid),getvardims(rainc_regrid))
  filevardef(fout,"rainnc",typeof(rainnc_regrid),getvardims(rainnc_regrid))
    
  filevarattdef(fout,"lat",v10_regrid&lat) ;-- copy lat attributes
  filevarattdef(fout,"lon",v10_regrid&lon) ;-- copy lon attributes
  filevarattdef(fout,"u10",u10_regrid)
  filevarattdef(fout,"v10",v10_regrid)
  filevarattdef(fout,"slp",slp_regrid)
  filevarattdef(fout,"t2",t2_regrid)
;  filevarattdef(fout,"td",td_regrid)
  filevarattdef(fout,"rainc",rainc_regrid)
  filevarattdef(fout,"rainnc",rainnc_regrid)
  
  setfileoption(fout,"DefineMode",False)

  fout->u10 = (/u10_regrid/)
  fout->v10 = (/v10_regrid/) 
  fout->slp = (/slp_regrid/) 
  fout->t2 = (/t2_regrid/) 
;  fout->td = (/td_regrid/) 
  fout->rainc  = (/rainc_regrid/) 
  fout->rainnc = (/rainnc_regrid/) 
  
  fout->lat = (/v10_regrid&lat/) ;-- write lat to new netCDF file
  fout->lon = (/v10_regrid&lon/) ;-- write lon to new netCDF file
  fout->time =  time
end

注意：运行该脚本会生成四个NetCDF文件，分别为destination_grid_file.nc、source_grid_file.nc、weights_file.nc以及wrfout_d01_2022-07-10_01:00:00.nc。其中，wrfout_d01_2022-07-10_01:00:00.nc为所需文件，但如何去除其他三个中间文件暂未解决。

Python格式转换脚本1

Python脚本如下所示：

# -*- coding: utf-8 -*-
"""
Created on %(date)s

@author: %(jixianpu)s

Email : 211311040008@hhu.edu.cn

introduction : keep learning althongh walk slowly
"""
"""
用来读取用ncl插值后的wrfoutput.nc 数据,并生成对应文件名的json格式
"""
import pandas as pd
import os
import json
import netCDF4 as nc
import numpy as np
import  datetime
from netCDF4 import Dataset
import argparse
from argparse import RawDescriptionHelpFormatter
import xarray as xr
import sys
import glob

date = sys.argv[1]
date = str(date)
frst = sys.argv[2]
step = sys.argv[3]

path = r'/Users/WRF/outdata/2022071000/'#只能是已经存在的文件目录且有数据才可以进行读取
start = datetime.datetime.strptime(date,'%Y%m%d%H').strftime("%Y-%m-%d_%H:%M:%S")
end = (datetime.datetime.strptime(date,'%Y%m%d%H')+datetime.timedelta(hours=int(frst))).strftime("%Y-%m-%d_%H:%M:%S")
intp = (datetime.datetime.strptime(date,'%Y%m%d%H')+datetime.timedelta(hours=int(step))).strftime("%Y-%m-%d_%H:%M:%S")
fstart = path+'/wrfout_d01_'+start+'*'
fintp  = path+'/wrfout_d01_'+intp+'*'
fend   = path+'/wrfout_d01_'+end+'*'
file = path+'/*'
filestart = glob.glob(fstart)
fileintp  = glob.glob(fintp)
fileend   = glob.glob(fend)
filelist  = glob.glob(file)
filelist.sort()   
rstart = np.array(np.where(np.array(filelist)==filestart))[0][0]
rintp = np.array(np.where(np.array(filelist)==fileintp))[0][0]
rend   = np.array(np.where(np.array(filelist)==fileend))[0][0]
fn = filelist[rstart:rend:rintp]
outroot = 'Users/'    
for i in fn:
    uhdr = {"header":{"discipline":0,"disciplineName":"Meteorological products","gribEdition":2,"gribLength":131858,"center":0,"centerName":"WRF OUTPUT","subcenter":0,"refTime":"2014-01-31T00:00:00.000Z","significanceOfRT":1,"significanceOfRTName":"Start of forecast","productStatus":0,"productStatusName":"Operational products","productType":1,"productTypeName":"Forecast products","productDefinitionTemplate":0,"productDefinitionTemplateName":"Analysis/forecast at horizontal level/layer at a point in time","parameterCategory":2,"parameterCategoryName":"Momentum","parameterNumber":2,"parameterNumberName":"U-component_of_wind","parameterUnit":"m.s-1","genProcessType":2,"genProcessTypeName":"Forecast","forecastTime":3,"surface1Type":103,"surface1TypeName":"Specified height level above ground","surface1Value":10,"surface2Type":255,"surface2TypeName":"Missing","surface2Value":0,"gridDefinitionTemplate":0,"gridDefinitionTemplateName":"Latitude_Longitude","numberPoints":65160,"shape":6,"shapeName":"Earth spherical with radius of 6,371,229.0 m","gridUnits":"degrees","resolution":48,"winds":"true","scanMode":0,"nx":360,"ny":181,"basicAngle":0,"subDivisions":0,"lo1":0,"la1":90,"lo2":359,"la2":-90,"dx":1,"dy":1}}

    vhdr = {"header":{"discipline":0,"disciplineName":"Meteorological products","gribEdition":2,"gribLength":131858,"center":0,"centerName":"WRF OUTPUT","subcenter":0,"refTime":"2014-01-31T00:00:00.000Z","significanceOfRT":1,"significanceOfRTName":"Start of forecast","productStatus":0,"productStatusName":"Operational products","productType":1,"productTypeName":"Forecast products","productDefinitionTemplate":0,"productDefinitionTemplateName":"Analysis/forecast at horizontal level/layer at a point in time","parameterCategory":2,"parameterCategoryName":"Momentum","parameterNumber":3,"parameterNumberName":"V-component_of_wind","parameterUnit":"m.s-1","genProcessType":2,"genProcessTypeName":"Forecast","forecastTime":3,"surface1Type":103,"surface1TypeName":"Specified height level above ground","surface1Value":10,"surface2Type":255,"surface2TypeName":"Missing","surface2Value":0,"gridDefinitionTemplate":0,"gridDefinitionTemplateName":"Latitude_Longitude","numberPoints":65160,"shape":6,"shapeName":"Earth spherical with radius of 6,371,229.0 m","gridUnits":"degrees","resolution":48,"winds":"true","scanMode":0,"nx":360,"ny":181,"basicAngle":0,"subDivisions":0,"lo1":0,"la1":90,"lo2":359,"la2":-90,"dx":1,"dy":1}}

    data = [uhdr, vhdr]
    newf = Dataset(i)
    lat = np.array(newf.variables['lat'])
    # print(fn,lat)
    lon = np.array(newf.variables['lon'])
    dys = np.diff(lat, axis = 0).mean(1)
    dy = float(dys.mean())
    dxs = np.diff(lon, axis = 1).mean(0)
    dx = float(dxs.mean())
    nx = float(lon.shape[1])
    ny = float(lat.shape[0])
    la1 = float(lat[-1, -1])
    la2 = float(lat[0, 0])
    lo1 = float(lon[0, 0])
    lo2 = float(lon[-1, -1])

    time =(newf.variables['time'])

    dates = nc.num2date(time[:],units=time.units)

    dt = pd.to_datetime(np.array(dates, dtype='datetime64[s]')).strftime("%Y%m%d%H%M%S")

    tms =pd.to_datetime(np.array(dates, dtype='datetime64[s]')).strftime("%Y-%m-%d_%H:%M:%S")
    for ti, time in enumerate(dt):

        datestr = (dt[0][:8])
        timestr = (dt[0][8:10])+'00'

        dirpath = outroot + date
        os.makedirs(dirpath, exist_ok = True)
        outpath = os.path.join(dirpath, '%s.json' % (i[-19:]))
        for u0_or_v1 in [0, 1]:

            h = data[u0_or_v1]['header']
            h['la1'] = la1
            h['la2'] = la2
            h['lo1'] = lo1
            h['lo2'] = lo2
            h['nx'] = nx
            h['ny'] = ny
            h['dx'] = dx
            h['dy'] = dy
            h['forecastTime'] = 0
            h['refTime'] = tms[0] + '.000Z'

            h['gribLength'] = 1538 + nx * ny * 2
            if u0_or_v1 == 0:
                data[u0_or_v1]['data'] = np.array(newf.variables['u10']).ravel().tolist()
            elif u0_or_v1 == 1:
                data[u0_or_v1]['data'] = np.array(newf.variables['v10']).ravel().tolist()
        if ti == 0:
            outf = open(outpath, 'w')
            json.dump(data, outf)
            outf.close()
        outf = open(outpath, 'w')
        json.dump(data, outf)
        outf.close()

上述脚本适用于Linux系统，运行方式如下：

python xx.py 起报时间时长间隔

例如：

我的WRFOUT数据名称如下：

python  convert_to_json.py 2022071000 12 06

根据所需的模式起始时间、起报时长（小时）及预报时间间隔（小时）进行自动化转换。

Python格式转换脚本2

此外，准备了一个Windows下的简易脚本，转换出的信息较为简单：

# -*- coding: utf-8 -*-
"""
Created on %(date)s

@author: %(jixianpu)s

Email : 211311040008@hhu.edu.cn

introduction : keep learning althongh walk slowly
"""
from __future__ import print_function, unicode_literals
import pandas as pd
import os
import json
import netCDF4 as nc
import numpy as np
import  datetime
from netCDF4 import Dataset
import argparse
from argparse import RawDescriptionHelpFormatter
import xarray as xr
# parser = argparse.ArgumentParser(description = """
# """, formatter_class = RawDescriptionHelpFormatter)

args = r'J:/wrf自动化/wrfout_d01_2022-07-10_01_00_00.nc'

outroot = r'D:/'

uhdr = {"header":{
                  "nx":360,
                  "ny":181,
                  "max":11,
                  }}

data = [uhdr]
newf = Dataset(args)
lat = np.array(newf.variables['lat'])
lon = np.array(newf.variables['lon'])
u10 = np.array(newf.variables['u10'])
v10 = np.array(newf.variables['v10'])

# indx = u10>1000

# u10[indx] = np.nan
# v10[indx] = np.nan

w10 = np.nanmax(np.sqrt(u10*u10+v10*v10))

dys = np.diff(lat, axis = 0).mean(1)
dy =    float(dys.mean())
print('Latitude Error:', np.abs((dy / dys) - 1).max())
print('Latitude Sum Error:', (dy / dys - 1).sum())
dxs = np.diff(lon, axis = 1).mean(0)
dx =    float(dxs.mean())
print('Longitude Error:', np.abs(dx / dxs - 1).max())
print('Longitude Sum Error:', (dx / dxs - 1).sum())

nx =    float(lon.shape[1])
ny =    float(lat.shape[0])

la1 =    float(lat[-1, -1])
la2 =   float(lat[0, 0])
lo1 =   float(lon[0, 0])
lo2 =   float(lon[-1, -1])
time =(newf.variables['time'])
dates = nc.num2date(time[:],units=time.units)
dt = pd.to_datetime(np.array(dates, dtype='datetime64[s]')).strftime("%Y%m%d%H%M%S")

ds= {
                      "nx":360,
                      "ny":181,
                      "max":11,
                      # "lo1":0,
                      # "la1":90,
                      # "lo2":359,
                      # "la2":-90,
                      # "dx":1,
                      # "dy":1,
                      # "parameterUnit":"m.s-1",
                      'data':{}
        }

ds['max']   =    float(w10)
ds['nx']    =    (nx)
ds['ny']    =    (ny)
for ti, time in enumerate(dt):
    #2012/02/07/0100Z/wind/surface/level/orthographic=-74.01,4.38,29184
    datestr = (dt[0][:8])
    timestr = (dt[0][8:10])+'00'
    print('Add "#' + datestr + '/' + timestr + 'Z/wind/surface/level/orthographic" to url to see this time')
    dirpath = os.path.join('D:', *datestr.split('/'))
    os.makedirs(dirpath, exist_ok = True)
    outpath = os.path.join(dirpath, '%s-wind-surface-level-gfs-1.0.json' % (timestr,))
    udata=u10.ravel()
    data[0]['data']=[]
    for i in range(len(udata)):

        data[0]['data'].append([
        u10.ravel().tolist()[i],
        v10.ravel().tolist()[i]])

    ds['data'] = data[0]['data']

outf = open(outpath, 'w')
json.dump(ds,outf)
outf.close()

该脚本可在编辑器中直接运行。

运行结束后，将在输出路径下生成一个文件夹：