python scipy interp2d 运行太慢问题

晨曦

我用interp2d把一个（181 * 360）的场插值成（361 * 719）的场，运算怎么那么慢啊！！！！！求大神解释和帮助！

####脚本如下

import numpy as np

import pygrib

from scipy import interpolate

grbs=pygrib.open('perturbation_1220300000120.grib2')

grb=grbs.select(shortName='gh',level=500)[0]

lat,lon=grb.latlons() ###lat，lon 是两维坐标量（181*360）

z=grb.values

latnew,lonnew=np.mgrid[90:-90:361j,0:359:719j]

f=interpolate.interp2d(lat,lon,z)  ##########这一步真的太太慢了。。。。

znew=f(latnew,lonnew)

网络暴民

。。。。。。

阿阿飞飞

Interp2d 主要针对不规则的散点资料，对于格点资料用scipy.ndimage或map_coordinates
ndimage的zoom可以放大一个数组：
“Zooming" an array (scipy.ndimage.zoom)

As a quick example (This will use cubic interpolation. Use order=1 for bilinear, order=0 for nearest, etc.):

import numpy as np
import scipy.ndimage as ndimage

data = np.arange(9).reshape(3,3)

print 'Original:\n', data
print 'Zoomed by 2x:\n', ndimage.zoom(data, 2)

This yields:

Original:
[[0 1 2]
[3 4 5]
[6 7 8]]
Zoomed by 2x:
[[0 0 1 1 2 2]
[1 1 1 2 2 3]
[2 2 3 3 4 4]
[4 4 5 5 6 6]
[5 6 6 7 7 7]
[6 6 7 7 8 8]]

基于这两个函数，JohannesBuchner做了一个regulargrid包，用于插值规则格点资料（下载地址见regulargrid）
用法：
插值函数一：
Cartesian grid regulargrid.cartesiangrid.CartesianGrid (equal spacing between points)

    Uses very fast implementation based on scipy.ndimage.map_coordinates

    Example:

1. # create a 3-dimensional cartesian grid:
   
2. limits = [(0, 1), (0, 1), (0, 1)]
   
3. x = numpy.linspace(0, 1, 8)
   
4. y = numpy.linspace(0, 1, 9)
   
5. z = numpy.linspace(0, 1, 10)
   
6. 
   
7. Z, Y = numpy.meshgrid(z, y)
   
8. X = numpy.array([[x]]).transpose()
   
9. 
   
10. # our grid values
    
11. values = X**2 + Y - Z
    
12. 
    
13. from regulargrid.cartesiangrid import CartesianGrid
    
14. # does linear interpolation
    
15. grid = CartesianGrid(limits, values)
    
16. 
    
17. # interpolate for one point
    
18. print grid([0.1], [0.5], [0.3])
    
19. # interpolate many
    
20. print grid([0.1, 0.3], [0.5, 0.5], [0.3, 0.2])

# create a 3-dimensional cartesian grid:<br /> limits = [(0, 1), (0, 1), (0, 1)]<br /> x = numpy.linspace(0, 1, 8)<br /> y = numpy.linspace(0, 1, 9)<br /> z = numpy.linspace(0, 1, 10)<br /> <br /> Z, Y = numpy.meshgrid(z, y)<br /> X = numpy.array([[x]]).transpose()<br /> <br /> # our grid values<br /> values = X**2 + Y - Z<br /> <br /> from regulargrid.cartesiangrid import CartesianGrid<br /> # does linear interpolation<br /> grid = CartesianGrid(limits, values)<br /> <br /> # interpolate for one point<br /> print grid([0.1], [0.5], [0.3])<br /> # interpolate many<br /> print grid([0.1, 0.3], [0.5, 0.5], [0.3, 0.2])

插值函数二：
Regular grid regulargrid.regulargrid.RegularGrid (unequal spacing between points)
该函数作者没给例程，估计用法跟上述用法一致

晨曦

阿阿飞飞 发表于 2013-9-28 10:25

                               
登录/注册后可看大图

Interp2d 主要针对不规则的散点资料，对于格点资料用scipy.ndimage或map_coordinates
ndimage的zoom可以放 ...

非常感谢你

奔波尔灞

压哈 居然在这能找到你

ginanony

python需要关注优化程序，提升程序的速度，学习了
{:eb502:}{:eb502:}

wanghp

学习了

平凡的幸福

请问lat 和lon都是二维的，用什么插值函数

mozhiting

学到了，不错不错，python用起来