【台风预报制图python】误差圈和4象限风圈的画法，以FNV3为例

菜园子 · 发表于 2025-8-13 20:32

本帖最后由菜园子于 2025-8-13 20:50 编辑

我现在已经懒到连日期拆分都用大模型写了哈哈哈，使用大模型写代码真是方便，只要你有想象力，就没有做不出来的功能
以下代码应该可以直接运行，前提是要装好cartopy（运行时需要先把FNV3的相关文件放在指定文件夹中，下载地址为https://deepmind.google.com/science/weatherlab，需要梯子，你也可以自己修改指定文件夹）

代码中不懂的可以问大模型，现在的大模型真是很好的老师，我也是做大模型方向的，一起感受大模型的魅力吧~

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.patches as patches
import cartopy.crs as ccrs
import cartopy.feature as cfeature
from cartopy.io.shapereader import Reader
from cartopy.mpl.gridliner import LONGITUDE_FORMATTER, LATITUDE_FORMATTER
import matplotlib.ticker as mticker
from scipy.interpolate import PchipInterpolator
from shapely.geometry import Polygon, MultiPolygon
from shapely.ops import unary_union
from datetime import datetime, timedelta
import os
import shutil
import re
def extract_init_time(filename):
"""
从文件名中提取init_time（格式：YYYYMMDDHHZ）
示例文件名：FNV3_2025_08_13T06_00_paired.csv → 返回 2025081306Z
"""
try:
# 使用正则表达式匹配日期时间部分
match = re.search(r'(\d{4})_(\d{2})_(\d{2})T(\d{2})_\d{2}', filename)
if match:
year, month, day, hour = match.groups()
return f"{year}{month}{day}{hour}Z"
else:
raise ValueError("文件名格式不符合要求")
except Exception as e:
print(f"Error extracting init_time: {e}")
raise
def parse_lead_time(lead_time_str):
"""Parse lead_time string to get hours"""
# Format: 'X days HH:MM:SS'
parts = lead_time_str.strip().split()
days = int(parts[0])
time_part = parts[2] # HH:MM:SS
hours, minutes, seconds = map(int, time_part.split(':'))
total_hours = days * 24 + hours + minutes / 60 + seconds / 3600
return total_hours
def simplify_track_id(track_id):
"""Simplify track ID (e.g., AL052025 -> 05A, WP162025 -> 16W)"""
if len(track_id) >= 8:
basin = track_id[0] # First letter
number = track_id[2:4] # First two numbers
return f"{number}{basin}"
return track_id
def draw_wind_radii(ax, lat, lon, ne_radius, se_radius, sw_radius, nw_radius, color, zorder=15):
"""Draw wind radii for four quadrants"""
# Convert 0km to 0.01km for robustness
radii = [max(0.01, r) for r in [ne_radius, se_radius, sw_radius, nw_radius]]
# Check if all radii are effectively 0
if all(r <= 0.01 for r in radii):
return
# Convert km to degrees (approximate)
km_to_deg = 1/111.32
radii_deg = [r * km_to_deg for r in radii]
# Define angles for each quadrant
angles = {
'NE': np.linspace(0, np.pi/2, 50), # 0 to 90 degrees
'SE': np.linspace(np.pi/2, np.pi, 50), # 90 to 180 degrees
'SW': np.linspace(np.pi, 3*np.pi/2, 50), # 180 to 270 degrees
'NW': np.linspace(3*np.pi/2, 2*np.pi, 50) # 270 to 360 degrees
}
# Calculate points for each quadrant
all_x, all_y = [], []
for i, (quadrant, theta_range) in enumerate(angles.items()):
radius = radii_deg[i]
x_quad = lon + radius * np.cos(theta_range)
y_quad = lat + radius * np.sin(theta_range)
all_x.extend(x_quad)
all_y.extend(y_quad)
# Close the polygon
if all_x and all_y:
all_x.append(all_x[0])
all_y.append(all_y[0])
ax.plot(all_x, all_y, color=color, linewidth=1.0,
transform=ccrs.PlateCarree(), zorder=zorder)
def create_smooth_track(hours, lons, lats, points_per_segment=20):
"""Create smooth track using interpolation"""
if len(hours) < 2:
return lons, lats
# Create interpolated points
interp_hours = []
for i in range(len(hours)-1):
segment = np.linspace(hours[i], hours[i+1], num=points_per_segment)[:-1]
interp_hours.extend(segment)
interp_hours.append(hours[-1])
interp_hours = np.array(interp_hours)
# Use PCHIP interpolation for smooth curves
smooth_lons = PchipInterpolator(hours, lons)(interp_hours)
smooth_lats = PchipInterpolator(hours, lats)(interp_hours)
return smooth_lons, smooth_lats
def create_typhoon_map(df_typhoon, track_id, init_time_str):
global textPairedContent
textPairedContent = textPairedContent + simplify_track_id(track_id) + ','
"""Create typhoon forecast map for a single typhoon"""
# Filter data for forecast times up to 120 hours
df_filtered = df_typhoon[df_typhoon['lead_time_hours'] <= 120].copy()
# Only plot specific time points: 0, 12, 24, 36, 48, 72, 120 hours
target_hours = [0, 12, 24, 36, 48, 72, 120]
df_plot = df_filtered[df_filtered['lead_time_hours'].isin(target_hours)].copy()
df_plot = df_plot.sort_values('lead_time_hours')
if df_plot.empty:
print(f"No data to plot for typhoon {track_id}")
return
# Calculate map extent
lats = df_plot['lat'].values
lons = df_plot['lon'].values
lat_margin = 8
lon_margin = 8
extendLatMin = max(-90, np.min(lats) - lat_margin)
extendLatMax = min(90, np.max(lats) + lat_margin)
extendLonMin = max(-180, np.min(lons) - lon_margin)
extendLonMax = min(180, np.max(lons) + lon_margin)
centerlon = (extendLonMin + extendLonMax) / 2
# Create the map
myproj = ccrs.PlateCarree()
fig = plt.figure(figsize=(12.5, 11))
ax = fig.add_axes([0.01, 0.01, 0.9, 0.9], projection=ccrs.PlateCarree(central_longitude=centerlon))
extend = [extendLonMin, extendLonMax, extendLatMin, extendLatMax]
ax.set_extent(extend, crs=ccrs.Geodetic())
ax.set_adjustable('datalim')
# Add gridlines
gl = ax.gridlines(draw_labels=True, linewidth=0.3, color='k', alpha=0.65, linestyle=':', zorder=90)
gl.xformatter = LONGITUDE_FORMATTER
gl.yformatter = LATITUDE_FORMATTER
gl.xlocator = mticker.FixedLocator(np.arange(-180, 180, 5))
gl.ylocator = mticker.FixedLocator(np.arange(-90, 90, 5))
gl.top_labels = False
gl.right_labels = False
gl.xlines = True
gl.ylines = True
gl.xlabel_style = {'size': 10}
gl.ylabel_style = {'size': 10}
# Add map features
ax.add_feature(cfeature.COASTLINE.with_scale('50m'), edgecolor='#6b6b6b', facecolor='none', lw=0.42)
ax.add_feature(cfeature.OCEAN.with_scale('50m'), facecolor='white')
ax.add_feature(cfeature.LAND.with_scale('50m'), facecolor='#e3e3e3')
# Add shapefiles (if they exist)
shapefiles = [
]
for shapefile in shapefiles:
if os.path.exists(shapefile):
try:
ax.add_geometries(Reader(shapefile).geometries(),
crs=myproj, facecolor='none', edgecolor='k', linewidth=0.80, zorder=10)
except Exception as e:
print(f"Warning: Could not load shapefile {shapefile}: {e}")
# Draw error circles first (lower zorder)
hours = df_plot['lead_time_hours'].values
fore_lon = df_plot['lon'].values
fore_lat = df_plot['lat'].values
# Calculate error radius: 50km per 12 hours, convert to degrees
fore_r = (hours / 12) * 50 / 111.32 # 50km per 12 hours, convert km to degrees
print(f"Debug: Error radii for {track_id}: {fore_r}") # Debug info
if len(hours) > 1:
# Interpolate for smooth error envelope
points_num = 12
interp_hours = []
for i in range(len(hours)-1):
segment = np.linspace(hours[i], hours[i+1], num=points_num)[:-1]
interp_hours.extend(segment)
interp_hours.append(hours[-1])
interp_hours = np.array(interp_hours)
x = PchipInterpolator(hours, fore_lon)(interp_hours)
y = PchipInterpolator(hours, fore_lat)(interp_hours)
r = PchipInterpolator(hours, fore_r)(interp_hours)
# Create error envelope
thetas = np.linspace(0, 2 * np.pi, 360)
polygon_x = x[:,None] + r[:,None] * np.cos(thetas) # Note: cos for longitude
polygon_y = y[:,None] + r[:,None] * np.sin(thetas) # Note: sin for latitude
ps = [Polygon(i) for i in np.dstack((polygon_x, polygon_y))]
n = range(len(ps)-1)
if n: # Only if there are at least 2 circles
convex_hulls = [MultiPolygon([ps[i], ps[i+1]]).convex_hull for i in n]
polygons = unary_union(convex_hulls)
ax.add_geometries([polygons], ccrs.PlateCarree(),
facecolor='yellow', alpha=0.5, edgecolor='black', linewidth=0.5, zorder=5)
print(f"Added error envelope for {track_id}")
else:
# If only one point, draw a simple circle
if len(hours) == 1 and fore_r[0] > 0:
circle = plt.Circle((fore_lon[0], fore_lat[0]), fore_r[0],
facecolor='yellow', alpha=0.5, edgecolor='black',
linewidth=0.5, transform=ccrs.PlateCarree(), zorder=5)
ax.add_patch(circle)
print(f"Added single error circle for {track_id}")
else:
print(f"Not enough points for error envelope: {len(hours)} points")
# ===== 修改部分：绘制平滑的台风路径曲线 =====
track_lons = df_plot['lon'].values
track_lats = df_plot['lat'].values
track_hours = df_plot['lead_time_hours'].values
# 创建平滑曲线
if len(track_hours) >= 2:
smooth_lons, smooth_lats = create_smooth_track(track_hours, track_lons, track_lats)
# 绘制平滑的路径曲线
ax.plot(smooth_lons, smooth_lats, color='hotpink', linewidth=1.1,
transform=ccrs.PlateCarree(), zorder=20, label='Track')
print(f"Drew smooth track with {len(smooth_lons)} interpolated points")
else:
# 如果只有一个点，绘制原来的直线
ax.plot(track_lons, track_lats, color='hotpink', linewidth=1.1,
transform=ccrs.PlateCarree(), zorder=20, label='Track')
print(f"Drew straight track (insufficient points for smoothing)")
global_max_34kt = max(
df_plot['radius_34_knot_winds_ne_km'].max(),
df_plot['radius_34_knot_winds_se_km'].max(),
df_plot['radius_34_knot_winds_sw_km'].max(),
df_plot['radius_34_knot_winds_nw_km'].max()
)
prev_lon, prev_lat = None, None # 记录上一点
# Draw typhoon positions and wind radii
for idx, row in df_plot.iterrows():
lat, lon = row['lat'], row['lon']
# Draw position marker
if row['lead_time_hours'] == 0:
# 第一个点用黑色'x'标记，加粗显示
ax.plot(lon, lat, 'x', color='black', markersize=5.5,
markeredgewidth=1, transform=ccrs.PlateCarree(), zorder=25)
else:
# 其他点用粉色'o'标记
ax.plot(lon, lat, 'o', color='hotpink', markersize=3.5,
transform=ccrs.PlateCarree(), zorder=25)
# Draw wind radii (34, 50, 64 knots)
wind_data = [
# 34-knot winds (7级风圈) - pink
(row['radius_34_knot_winds_ne_km'], row['radius_34_knot_winds_se_km'],
row['radius_34_knot_winds_sw_km'], row['radius_34_knot_winds_nw_km'], 'hotpink'),
# 50-knot winds (10级风圈) - dark red
(row['radius_50_knot_winds_ne_km'], row['radius_50_knot_winds_se_km'],
row['radius_50_knot_winds_sw_km'], row['radius_50_knot_winds_nw_km'], 'darkred'),
# 64-knot winds (12级风圈) - red
(row['radius_64_knot_winds_ne_km'], row['radius_64_knot_winds_se_km'],
row['radius_64_knot_winds_sw_km'], row['radius_64_knot_winds_nw_km'], 'red')
]
for ne, se, sw, nw, color in wind_data:
if pd.notna(ne) and pd.notna(se) and pd.notna(sw) and pd.notna(nw):
draw_wind_radii(ax, lat, lon, ne, se, sw, nw, color)
# 获取风速值并转换为整数
# 解析valid_time获取日期和小时
try:
# 解析valid_time字符串（格式：'2025-08-12 06:00:00'）
valid_time_str = row['valid_time']
# 直接提取月份中的日和小时部分
# 格式：'YYYY-MM-DD HH:MM:SS'
parts = valid_time_str.split()
date_part = parts[0] # '2025-08-12'
time_part = parts[1] # '06:00:00'
# 提取日和小时
day = date_part.split('-')[2] # 从'2025-08-12'提取'12'
hour = time_part.split(':')[0] # 从'06:00:00'提取'06'
# 格式化日期标签：日/小时Z
date_label = f"{day}/{hour}Z"
except Exception as e:
print(f"Error parsing valid_time: {valid_time_str}, using fallback. Error: {e}")
# 如果解析失败，使用备用标签（小时数）
date_label = f"{int(row['lead_time_hours'])}h"
# 获取风速值并转换为整数
try:
wind_speed = int(row['maximum_sustained_wind_speed_knots'])
except (ValueError, TypeError):
wind_speed = "N/A"
# 计算动态c_r = 最大34kt风圈半径（转换为地理坐标偏移量）
c_r = global_max_34kt/ 111.32 # 约1°纬度=111.32km
global_max_34kt = global_max_34kt + 10
offset_deg = c_r * 1.5 # 维持原有的偏移系数
# 计算连线方向
if prev_lon is not None and prev_lat is not None:
dx = lon - prev_lon
dy = lat - prev_lat
# 判定斜向
if abs(dx) < 1e-6 or abs(dy) < 1e-6:
# 水平或竖直
angle_deg = 30
else:
if dx > 0 and dy < 0: # 左上到右下
angle_deg = 30
elif dx > 0 and dy > 0: # 左下到右上
angle_deg = -30
else:
# 其它情况（比如从右到左），也可以用默认值
angle_deg = 30
else:
# 第一个点
angle_deg = 30
# 偏移计算
theta = np.radians(angle_deg)
dlat = offset_deg * np.cos(theta)
dlon = (offset_deg * np.sin(theta)) / np.cos(np.radians(lat))
label_lon = lon + dlon
label_lat = lat + dlat
ax.text(label_lon, label_lat, f"{date_label} {wind_speed}KTS",
transform=ccrs.PlateCarree(), fontsize=9, ha='left', va='bottom', zorder=30)
ax.plot([lon, label_lon], [lat, label_lat], color='black', linewidth=0.6,
transform=ccrs.PlateCarree(), zorder=29)
prev_lon, prev_lat = lon, lat
# Add title and legend
simplified_id = simplify_track_id(track_id)
plt.title(f'FNV3 Ensemble Mean Track Forecast for Tropical Cyclone [{simplified_id}]\nInitTime: {init_time_str}',
fontsize=14, fontweight='bold')
# Add legend for wind radii
legend_elements = [
plt.Line2D([0], [0], color='hotpink', linewidth=2, label='34kt winds'),
plt.Line2D([0], [0], color='darkred', linewidth=2, label='50kt winds'),
plt.Line2D([0], [0], color='red', linewidth=2, label='64kt winds'),
patches.Rectangle((0, 0), 1, 1, facecolor='yellow', alpha=0.3, label='Error envelope')
]
ax.legend(handles=legend_elements, loc='upper right', bbox_to_anchor=(0.991, 0.991))
plt.figtext(0.775, 0.0276, "Copyright\xa9 SMCA", fontsize = 11,color="black",zorder=101)
# Save the figure
filename = f"{simplified_id}_paired_{init_time_str[:-1]}.png"
filepath = f"/NWP/AI_TC_pic/{init_time_str[:-1]}/{filename}"
plt.savefig(filepath, dpi=300, bbox_inches='tight')
plt.close()
import os
import pandas as pd
textPairedContent = ''
def main():
# 定义目录路径将FNV3的文件放在该目录下即可例如 FNV3_2025_08_13T00_00_paired.csv
directory = "/NWP/AI_TC_ENS/FNV3_paired_circle/"
try:
# 获取目录下所有CSV文件
csv_files = [f for f in os.listdir(directory) if f.endswith('.csv')]
if not csv_files:
print("No CSV files found in the directory.")
return
print(f"Found {len(csv_files)} CSV files to process.")
# 处理每个CSV文件
for csv_file in csv_files:
file_path = os.path.join(directory, csv_file)
print(f"\nProcessing file: {csv_file}")
try:
init_time = extract_init_time(csv_file) # 修改点：从文件名获取
# 读取CSV文件，跳过前6行
df = pd.read_csv(file_path, skiprows=6)
# 解析lead_time获取小时数
df['lead_time_hours'] = df['lead_time'].apply(parse_lead_time)
# 获取唯一的台风track_id
unique_typhoons = df['track_id'].unique()
print(f"Found {len(unique_typhoons)} typhoons in this file: {unique_typhoons}")
# 处理每个台风
for track_id in unique_typhoons:
df_typhoon = df[df['track_id'] == track_id].copy()
# 获取init_time（同一台风的所有行应该相同）
print(f"Processing typhoon {track_id} with {len(df_typhoon)} data points")
# 为这个台风创建地图
create_typhoon_map(df_typhoon, track_id, init_time)
print(f"Finished processing file: {csv_file}")
except Exception as e:
print(f"Error processing file {csv_file}: {e}")
import traceback
traceback.print_exc()
continue # 继续处理下一个文件
print("\n全部台风路径文件制作完成")
except Exception as e:
print(f"Error accessing directory: {e}")
import traceback
traceback.print_exc()
if __name__ == "__main__":
main()

复制代码

爱丽丝女儿镇楼

菜园子 · 发表于 2025-8-13 20:33

本帖最后由菜园子于 2025-8-13 20:51 编辑

效果图如下，因为刚做出来，还比较粗糙，你们可以试试进行美化，砸门相互学习，相互借鉴

菜园子 · 发表于 2025-8-13 20:37

@eoj329 @alexctk

Momo240309 · 发表于 2025-8-13 21:44

学习一下，感觉cartopy比自己从自然地球导数据要方便

Enceladus · 发表于 2025-8-14 00:57

滋磁公开教程（本来我也不会画误差x

菜园子 · 发表于 2025-8-14 01:15

Enceladus 发表于 2025-8-14 00:57
滋磁公开教程（本来我也不会画误差x

其实是比较早的功能了，在Q群里相互毒奶的时候用的XD

Enceladus · 发表于 2025-8-14 02:55

现学现用（

菜园子 · 发表于 2025-8-14 09:33

Enceladus 发表于 2025-8-14 02:55
现学现用（

好好好，配色做的也很好看

		自动登录	找回密码
密码			立即注册

【台风预报制图python】误差圈和4象限风圈的画法，以FNV3为例

本帖子中包含更多资源

本帖子中包含更多资源

点评

本帖子中包含更多资源