从nc文件中提取风速数据并且进行时间序列分析

                                   本文目的

• 介绍了如何从nc文件中，提取风速数据；
• 介绍如何将风速数据转换成时间序列；
• 简单的时间序列的趋势拆解（首发）。
  
  " o  u" s. T: i# q; U8 s

代码链接

代码我已经放在Github上面了，免费分享使用，https://github.com/yuanzhoulvpi2 ... ree/main/python_GIS。

" h9 _/ p; R/ D- I) \+ F2 ]& Y

过程介绍

* L% P& f- _8 P8 I7 X5 @1. 导入包

6 m/ M, T! |' _, M; l$ y

[Python] 纯文本查看 复制代码

# 基础的数据处理工具
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt # 可视化
import datetime # 处理python时间函数
import netCDF4 as nc  # 处理nc数据
from netCDF4 import num2date  # 处理nc数据
import geopandas as gpd  # 处理网格数据，shp之类的
import rasterio  # 处理tiff文件
from shapely.geometry import Point  # gis的一些逻辑判断
from cartopy import crs as ccrs  # 设置投影坐标系等
from tqdm import tqdm  # 打印进度条
from joblib import Parallel, delayed  # 并行
import platform  # 检测系统

tqdm.pandas()

# matplotlib 显示中文的问题
if platform.system() == 'Darwin':
    plt.rcParams["font.family"] = 'Arial Unicode MS'
elif platform.system() == 'Windows':
    plt.rcParams["font.family"] = 'SimHei'
else:
    pass

! S6 e+ R8 j. N; L

2.导入数据 处理数据

5 G9 e/ d1 R! f* F2 C4 s1 G
/ x8 ^0 P/ s! ~  k; z! H

[Python] 纯文本查看 复制代码

# 导入数据
nc_data = nc.Dataset("./数据集/GIS实践3/2016_2020.nc")

# 处理数据
raw_latitude = np.array(nc_data.variables['latitude'])
raw_longitude = np.array(nc_data.variables['longitude'])
raw_time = np.array(nc_data.variables['time'])
raw_u10 = np.array(nc_data.variables['u10'])
raw_v10 = np.array(nc_data.variables['v10'])
# 提取缺失值，并且将缺失值替换
missing_u10_value = nc_data.variables['u10'].missing_value
missing_v10_value = nc_data.variables['v10'].missing_value
raw_v10[raw_v10 == missing_v10_value] = np.nan
raw_u10[raw_u10 == missing_u10_value] = np.nan


# 处理时间
def cftime2datetime(cftime, units, format='%Y-%m-%d %H:%M:%S'):
    """
    将nc文件里面的时间格式 从cftime 转换到 datetime格式
    :param cftime:
    :param units:
    :param format:
    :return:
    """
    return datetime.datetime.strptime(num2date(times=cftime, units=units).strftime(format), format)

clean_time_data = pd.Series([cftime2datetime(i, units=str(nc_data.variables['time'].units)) for i in tqdm(raw_time)])
clean_time_data[:4]

/ Y! ^5 G& n5 e6 E, ?
3. 计算风速数据

. Z7 m4 ?0 @0 r4 W! F- q- X0 k
/ g  b5 W0 q" G" ^& P3 V

[Python] 纯文本查看 复制代码

windspeed_mean = pd.Series([np.sqrt(raw_v10[i,:, :] ** 2 + raw_u10[i, :, :]**2).mean() for i in tqdm(range(clean_time_data.shape[0]))])

time_windspeed = pd.DataFrame({'time':clean_time_data,'mean_ws':windspeed_mean})
time_windspeed

; A* s. _; q  k! Y# T! t

# _1 D2 ~  v( e) x( ~) s0 _4 h  T
1 I3 @7 S: m3 ~6 J& N
4. 年度数据可视化

[Python] 纯文本查看 复制代码

year_data = time_windspeed.groupby(time_windspeed.time.dt.year).agg(
    mean_ws = ('mean_ws', 'mean')
).reset_index()

# year_data

with plt.style.context('fivethirtyeight') as style:

    fig, ax = plt.subplots(figsize=(10,3), dpi=300)
    ax.plot(year_data['time'], year_data['mean_ws'], '-o',linewidth=3, ms=6)
    ax.set_xticks(year_data['time'])
    #
    #
    for i in range(year_data.shape[0]):
        ax.text(year_data.iloc[/size][/font][i][font=新宋体][size=3]['time']+0.1, year_data.iloc[/size][/font][i][font=新宋体][size=3]['mean_ws'], str(np.around(year_data.iloc[/size][/font][i][font=新宋体][size=3]['mean_ws'], 2)),
                bbox=dict(boxstyle='round', facecolor='white', alpha=0.5))
    #
    for i in ['top', 'right']:
        ax.spines[/size][/font][i][font=新宋体][size=3].set_visible(False)

    ax.set_title("各年平均风速")
    ax.set_ylabel("$Wind Speed / m.s^{-1}$")

, r% q; p7 d+ q1 G% V

. l" w' e3 K4 {$ A7 S

, M0 I3 p9 A& H7 ]* O1 k# H5. 月维度数据可视化

[Python] 纯文本查看 复制代码

month_data = time_windspeed.groupby(time_windspeed.time.dt.month).agg(
    mean_ws = ('mean_ws', 'mean')
).reset_index()


with plt.style.context('fivethirtyeight') as style:

    fig, ax = plt.subplots(figsize=(10,3), dpi=300)
    ax.plot(month_data['time'], month_data['mean_ws'], '-o',linewidth=3, ms=6)
    ax.set_xticks(month_data['time'])
    _ = ax.set_xticklabels(labels=[f'{i}月' for i in month_data['time']])


    for i in range(month_data.shape[0]):
        ax.text(month_data.iloc[/size][/font][i][font=新宋体][size=3]['time'], month_data.iloc[/size][/font][i][font=新宋体][size=3]['mean_ws']+0.05, str(np.around(month_data.iloc[/size][/font][i][font=新宋体][size=3]['mean_ws'], 2)),
                bbox=dict(boxstyle='round', facecolor='white', alpha=0.5))

    for i in ['top', 'right']:
        ax.spines[/size][/font][i][font=新宋体][size=3].set_visible(False)

    ax.set_title("各月平均风速")
    ax.set_ylabel("$Wind Speed / m.s^{-1}$")
    fig.savefig("month_plot.png")

; p4 m. ?" a) N9 f

6.天维度数据可视化
3 |7 c6 t. f4 e& X5 [/ T

• 计算天数据
  4 C5 b7 _1 K" t: ?' w, H% D
  

[Python] 纯文本查看 复制代码

day_data = time_windspeed.groupby(time_windspeed.time.apply(lambda x: x.strftime('%Y-%m-%d'))).agg(
    mean_ws = ('mean_ws', 'mean')
).reset_index()

day_data['time'] = pd.to_datetime(day_data['time'])

day_data = day_data.set_index('time')
day_data.head()

• 可视化
  ' U# y4 q( Z/ x" \7 P
  5 }  h  J% B7 j: U4 Y) ^2 r! D- ^

[Python] 纯文本查看 复制代码

# day_data.dtypes
fig, ax = plt.subplots(figsize=(20,4), dpi=300)
ax.plot(day_data.index, day_data['mean_ws'], '-o')
# ax.xaxis.set_ticks_position('none')
# ax.tick_params(axis="x", labelbottom=False)
ax.set_title("每天平均风速")
ax.set_ylabel("$Wind Speed / m.s^{-1}$")
ax.set_xlabel("date")
fig.savefig('day_plot.png')

4 _' \" j4 F. F$ t

$ Q& K. o5 _( U4 H" E

1.天维度数据做趋势拆解

- ~+ f* t. Z9 B1 J5 u6 c& Y

[Python] 纯文本查看 复制代码

# 导入包
from statsmodels.tsa.seasonal import seasonal_decompose
from dateutil.parser import parse
# 乘法模型
result_mul = seasonal_decompose(day_data['mean_ws'], model="multilicative", extrapolate_trend='freq')
result_add = seasonal_decompose(day_data['mean_ws'], model="additive", extrapolate_trend='freq')
font = {'family': 'serif',
        'color': 'darkred',
        'weight': 'normal',
        'size': 16,
        }
# 画图

with plt.style.context('classic'):
    fig, ax = plt.subplots(ncols=2, nrows=4, figsize=(22, 15), sharex=True, dpi=300)


    def plot_decompose(result, ax, index, title, fontdict=font):
        ax[0, index].set_title(title, fontdict=fontdict)
        result.observed.plot(ax=ax[0, index])
        ax[0, index].set_ylabel("Observed")

        result.trend.plot(ax=ax[1, index])
        ax[1, index].set_ylabel("Trend")

        result.seasonal.plot(ax=ax[2, index])
        ax[2, index].set_ylabel("Seasonal")

        result.resid.plot(ax=ax[3, index])
        ax[3, index].set_ylabel("Resid")


    plot_decompose(result=result_add, ax=ax, index=0, title="Additive Decompose", fontdict=font)
    plot_decompose(result=result_mul, ax=ax, index=1, title="Multiplicative Decompose", fontdict=font)
    fig.savefig('decompose.png')

               

. j1 C: d9 T6 E# w7 y


& x7 d' O) ?: F, N. w9 a