Spaces:
Running
Running
# -*- coding: utf-8 -*- | |
""" | |
Created on Tue Dec 6 09:56:29 2022 | |
@author: mritchey | |
""" | |
#streamlit run "C:\Users\mritchey\.spyder-py3\Python Scripts\streamlit projects\mrms\mrms_all buffer.py" | |
import plotly.express as px | |
from joblib import Parallel, delayed | |
import pandas as pd | |
import streamlit as st | |
from geopy.extra.rate_limiter import RateLimiter | |
from geopy.geocoders import Nominatim | |
import folium | |
from streamlit_folium import st_folium | |
import math | |
import geopandas as gpd | |
from skimage.io import imread | |
from streamlit_plotly_events import plotly_events | |
import requests | |
import rasterio | |
import rioxarray | |
import numpy as np | |
import base64 | |
import re | |
import io | |
def geocode(address, buffer_size): | |
try: | |
address2 = address.replace(' ', '+').replace(',', '%2C') | |
df = pd.read_json( | |
f'https://geocoding.geo.census.gov/geocoder/locations/onelineaddress?address={address2}&benchmark=2020&format=json') | |
results = df.iloc[:1, 0][0][0]['coordinates'] | |
lat, lon = results['y'], results['x'] | |
except: | |
geolocator = Nominatim(user_agent="GTA Lookup") | |
geocode = RateLimiter(geolocator.geocode, min_delay_seconds=1) | |
location = geolocator.geocode(address) | |
lat, lon = location.latitude, location.longitude | |
df = pd.DataFrame({'Lat': [lat], 'Lon': [lon]}) | |
gdf = gpd.GeoDataFrame( | |
df, geometry=gpd.points_from_xy(df.Lon, df.Lat, crs=4326)) | |
gdf['buffer'] = gdf['geometry'].to_crs( | |
3857).buffer(buffer_size/2*2580).to_crs(4326) | |
return gdf | |
def get_pngs(date): | |
year, month, day = date[:4], date[4:6], date[6:] | |
url = f'https://mrms.nssl.noaa.gov/qvs/product_viewer/local/render_multi_domain_product_layer.php?mode=run&cpp_exec_dir=/home/metop/web/specific/opv/&web_resources_dir=/var/www/html/qvs/product_viewer/resources/&prod_root={prod_root}&qperate_pal_option=0&qpe_pal_option=0&year={year}&month={month}&day={day}&hour={hour}&minute={minute}&clon={lon}&clat={lat}&zoom={zoom}&width=920&height=630' | |
response = requests.get(url,verify=False) | |
image_data = io.BytesIO(response.content) | |
data = imread(image_data) | |
# data = imread(url)[:, :, :3] | |
data2 = data.reshape(630*920, 3) | |
data2_df = pd.DataFrame(data2, columns=['R', 'G', 'B']) | |
data2_df2 = pd.merge(data2_df, lut[['R', 'G', 'B', 'Value', ]], on=['R', 'G', 'B'], | |
how='left')[['Value', ]] | |
data2_df2['Date'] = date | |
return data2_df2.reset_index() | |
def get_pngs_parallel(dates): | |
results1 = Parallel(n_jobs=32, prefer="threads")( | |
delayed(get_pngs)(i) for i in dates) | |
return results1 | |
def png_data(date): | |
year, month, day = date[:4], date[4:6], date[6:] | |
url = f'https://mrms.nssl.noaa.gov/qvs/product_viewer/local/render_multi_domain_product_layer.php?mode=run&cpp_exec_dir=/home/metop/web/specific/opv/&web_resources_dir=/var/www/html/qvs/product_viewer/resources/&prod_root={prod_root}&qperate_pal_option=0&qpe_pal_option=0&year={year}&month={month}&day={day}&hour={hour}&minute={minute}&clon={lon}&clat={lat}&zoom={zoom}&width=920&height=630' | |
response = requests.get(url,verify=False) | |
image_data = io.BytesIO(response.content) | |
data = imread(image_data) | |
# data = imread(url) | |
return data | |
def map_folium(data, gdf): | |
m = folium.Map(location=[lat, lon], zoom_start=zoom, height=300) | |
folium.Marker( | |
location=[lat, lon], | |
popup=address).add_to(m) | |
folium.GeoJson(gdf['buffer']).add_to(m) | |
folium.raster_layers.ImageOverlay( | |
data, opacity=0.8, bounds=bounds).add_to(m) | |
return m | |
def to_radians(degrees): | |
return degrees * math.pi / 180 | |
def lat_lon_to_bounds(lat, lng, zoom, width, height): | |
earth_cir_m = 40075016.686 | |
degreesPerMeter = 360 / earth_cir_m | |
m_pixel_ew = earth_cir_m / math.pow(2, zoom + 8) | |
m_pixel_ns = earth_cir_m / \ | |
math.pow(2, zoom + 8) * math.cos(to_radians(lat)) | |
shift_m_ew = width/2 * m_pixel_ew | |
shift_m_ns = height/2 * m_pixel_ns | |
shift_deg_ew = shift_m_ew * degreesPerMeter | |
shift_deg_ns = shift_m_ns * degreesPerMeter | |
return [[lat-shift_deg_ns, lng-shift_deg_ew], [lat+shift_deg_ns, lng+shift_deg_ew]] | |
def image_to_geotiff(bounds, input_file_path, output_file_path='template.tiff'): | |
south, west, north, east = tuple( | |
[item for sublist in bounds for item in sublist]) | |
dataset = rasterio.open(input_file_path, 'r') | |
bands = [1, 2, 3] | |
data = dataset.read(bands) | |
transform = rasterio.transform.from_bounds(west, south, east, north, | |
height=data.shape[1], | |
width=data.shape[2]) | |
crs = {'init': 'epsg:4326'} | |
with rasterio.open(output_file_path, 'w', driver='GTiff', | |
height=data.shape[1], | |
width=data.shape[2], | |
count=3, dtype=data.dtype, nodata=0, | |
transform=transform, crs=crs, | |
compress='lzw') as dst: | |
dst.write(data, indexes=bands) | |
def get_mask(bounds, buffer_size): | |
year, month, day = date[:4], date[4:6], date[6:] | |
url = f'https://mrms.nssl.noaa.gov/qvs/product_viewer/local/render_multi_domain_product_layer.php?mode=run&cpp_exec_dir=/home/metop/web/specific/opv/&web_resources_dir=/var/www/html/qvs/product_viewer/resources/&prod_root={prod_root}&qperate_pal_option=0&qpe_pal_option=0&year={year}&month={month}&day={day}&hour={hour}&minute={minute}&clon={lon}&clat={lat}&zoom={zoom}&width=920&height=630' | |
img_data = requests.get(url, verify=False).content | |
input_file_path = f'image_name_{date}_{var}.png' | |
output_file_path = 'template.tiff' | |
with open(input_file_path, 'wb') as handler: | |
handler.write(img_data) | |
image_to_geotiff(bounds, input_file_path, output_file_path) | |
rds = rioxarray.open_rasterio(output_file_path) | |
# rds.plot.imshow() | |
rds = rds.assign_coords(distance=(haversine(rds.x, rds.y, lon, lat))) | |
mask = rds['distance'].values <= buffer_size | |
mask = np.transpose(np.stack([mask, mask, mask]), (1, 2, 0)) | |
return mask | |
def haversine(lon1, lat1, lon2, lat2): | |
# convert decimal degrees to radians | |
lon1 = np.deg2rad(lon1) | |
lon2 = np.deg2rad(lon2) | |
lat1 = np.deg2rad(lat1) | |
lat2 = np.deg2rad(lat2) | |
# haversine formula | |
dlon = lon2 - lon1 | |
dlat = lat2 - lat1 | |
a = np.sin(dlat/2)**2 + np.cos(lat1) * np.cos(lat2) * np.sin(dlon/2)**2 | |
c = 2 * np.arcsin(np.sqrt(a)) | |
r = 6371 | |
return c * r | |
def render_svg(svg): | |
"""Renders the given svg string.""" | |
b64 = base64.b64encode(svg.encode('utf-8')).decode("utf-8") | |
html = r'<img src="data:image/svg+xml;base64,%s"/>' % b64 | |
st.write(html, unsafe_allow_html=True) | |
def rgb_to_hex(rgb): | |
return '#'+'%02x%02x%02x' % rgb | |
def get_legend_lut(prod_root): | |
url_legend = f'https://mrms.nssl.noaa.gov/qvs/product_viewer/shared/fetch_svg_legend_via_config.php?web_resources_dir=/var/www/html/qvs/product_viewer/resources/&config_name=title_and_legend_config.txt&product={prod_root}' | |
r = requests.get(url_legend) # Get the webpage | |
svg = r.content.decode() # Decoded response content with the svg string | |
if svg.find('size="16">mm</text>') > 0: | |
svg = svg.replace('size="16">mm</text>', 'size="16">in</text>') | |
beg_string = '"13">' | |
end_string = '</text>' | |
res = re.findall('%s(.*)%s' % (beg_string, end_string), svg) | |
for mm in res: | |
inc = round(float(mm)*0.0393701, 2) | |
svg = svg.replace(f'{beg_string}{mm}{end_string}', | |
f'{beg_string}{str(inc)}{end_string}') | |
elif svg.find('font-size="12">') > 0: | |
beg_string = '"12">' | |
end_string = '</text>' | |
else: | |
beg_string = '"13">' | |
end_string = '</text>' | |
#Make LUT | |
values = re.findall('%s(.*)%s' % (beg_string, end_string), svg) | |
beg_string, end_string = 'fill="rgb(', ')" />' | |
rgb = re.findall('%s(.*)%s' % (beg_string, end_string), svg) | |
rgb = [eval(i[0]) for i in rgb] | |
beg_string, end_string = 'style="fill:rgb(', ');" />' | |
rgb2 = re.findall('%s(.*)%s' % (beg_string, end_string), svg) | |
rgb2 = [eval(i[0]) for i in rgb2] | |
rgb = rgb2+rgb | |
lut = pd.DataFrame({'Value': values, | |
'RGB': rgb}) | |
# lut['R'], lut['G'], lut['B'] = lut['RGB'].str | |
lut[['R','G','B']]=lut['RGB'].apply(pd.Series) | |
lut[['R', 'G', 'B']] = lut[['R', 'G', 'B']].astype('uint8') | |
lut['Value'] = lut['Value'].astype(float) | |
lut['hex'] = lut['RGB'].apply(rgb_to_hex) | |
return svg, lut | |
#Set Columns | |
st.set_page_config(layout="wide") | |
#Input Data | |
zoom = 10 | |
address = st.sidebar.text_input( | |
"Address", "1000 Main St, Cincinnati, OH 45202") | |
var = st.sidebar.selectbox( | |
'Product:', ('Hail', 'Flooding', 'Rain: Radar', 'Rain: Multi Sensor', 'Tornado')) | |
date = st.sidebar.date_input("Date", pd.Timestamp( | |
2022, 9, 8), key='date').strftime('%Y%m%d') | |
d = pd.Timestamp(date) | |
days_within = st.sidebar.selectbox('Within Days:', (5, 30, 60)) | |
mask_select = st.sidebar.radio('Only Show Buffer Data:', ("No", "Yes")) | |
buffer_size = st.sidebar.radio('Buffer Size (miles):', (5, 10, 15)) | |
year, month, day = date[:4], date[4:6], date[6:] | |
hour = 23 | |
minute = 0 | |
#Select Variable | |
if var == 'Hail': | |
var_input = 'hails&product=MESHMAX1440M' | |
elif var == 'Flooding': | |
var_input = 'flash&product=FL_ARI24H' | |
elif var == 'Rain: Radar': | |
var_input = 'q3rads&product=Q3EVAP24H' | |
elif var == 'Rain: Multi Sensor': | |
var_input = 'q3mss&product=P1_Q3MS24H' | |
elif var == 'Tornado': | |
var_input = 'azsh&product=RT1440M' | |
prod_root = var_input[var_input.find('=')+1:] | |
#Geocode | |
gdf = geocode(address, buffer_size) | |
lat, lon = tuple(gdf[['Lat', 'Lon']].values[0]) | |
#Get Value | |
url = 'https://mrms.nssl.noaa.gov/qvs/product_viewer/local/get_multi_domain_rect_binary_value.php?mode=run&cpp_exec_dir=/home/metop/web/specific/opv/&web_resources_dir=/var/www/html/qvs/product_viewer/resources/'\ | |
+ f'&prod_root={prod_root}&lon={lon}&lat={lat}&year={year}&month={month}&day={day}&hour={hour}&minute={minute}' | |
response = requests.get(url, verify=False).json() | |
qvs_values = pd.DataFrame(response, index=[0])[ | |
['qvs_value', 'qvs_units']].values[0] | |
qvs_value = qvs_values[0] | |
qvs_unit = qvs_values[1] | |
#Get PNG Focus | |
data = png_data(date) | |
#Get PNG Max | |
start_date, end_date = d - \ | |
pd.Timedelta(days=days_within), d+pd.Timedelta(days=days_within) | |
dates = pd.date_range(start_date, | |
end_date).strftime('%Y%m%d') | |
#Get SVG and Lut | |
svg, lut = get_legend_lut(prod_root) | |
bounds = lat_lon_to_bounds(lat, lon, zoom, 920, 630) | |
results1 = get_pngs_parallel(dates) | |
# results1 = Parallel(n_jobs=32, prefer="threads")(delayed(get_pngs)(i) for i in dates) | |
results = pd.concat(results1).fillna(0) | |
max_data = results.groupby('index')[['Value']].max() | |
max_data2 = pd.merge(max_data, | |
lut[['R', 'G', 'B', 'Value']], | |
on=['Value'], | |
how='left')[['R', 'G', 'B']] | |
data_max = max_data2.values.reshape(630, 920, 3) | |
#Masked Data | |
if mask_select == "Yes": | |
mask = get_mask(bounds, buffer_size) | |
mask1 = mask[:, :, 0].reshape(630*920) | |
results = pd.concat([i[mask1] for i in results1]) | |
data_max = data_max*mask | |
else: | |
pass | |
#Bar | |
if var == 'Tornado': | |
bar = results.query("Value>.006").groupby( | |
['Date', 'Value'])['index'].count().reset_index() | |
else: | |
bar = results.query("Value>.2").groupby(['Date', 'Value'])[ | |
'index'].count().reset_index() | |
bar['Date'] = pd.to_datetime(bar['Date']) | |
bar = bar.reset_index() | |
bar.columns = ['level_0', 'Date', 'Value', 'count'] | |
bar = bar.sort_values('Value', ascending=True) | |
bar['Value'] = bar['Value'].astype(str) | |
color_discrete_map = lut[['Value', 'hex']].sort_values( | |
'Value', ascending=True).astype(str) | |
color_discrete_map = color_discrete_map.set_index( | |
'Value').to_dict()['hex'] | |
fig = px.bar(bar, x="Date", y="count", color="Value", | |
barmode='stack', | |
color_discrete_map=color_discrete_map) | |
#Submit Url to New Tab | |
url = f'https://mrms.nssl.noaa.gov/qvs/product_viewer/index.php?web_exec_mode=run&menu=menu_config.txt&year={year}&month={month}&day={day}&hour=23&minute=30&time_mode=static&zoom=9&clon={lon}&clat={lat}&base=0&overlays=1&mping_mode=0&product_type={var_input}&qpe_pal_option=0&opacity=.75&looping_active=off&num_frames=6&frame_step=200&seconds_step=600' | |
#Map Focus | |
m = map_folium(data, gdf) | |
#Map Max | |
m_max = map_folium(data_max, gdf) | |
with st.container(): | |
col1, col2 = st.columns(2) | |
with col1: | |
st.header(f'{var} on {pd.Timestamp(date).strftime("%D")}') | |
st_folium(m, height=300) | |
with col2: | |
st.header( | |
f'Max from {start_date.strftime("%D")} to {end_date.strftime("%D")}') | |
st_folium(m_max, height=300) | |
with st.container(): | |
col1, col2, col3 = st.columns((1, 10, 6)) | |
with col1: | |
render_svg(svg) | |
with col2: | |
link = f'[Go To MRMS Site]({url})' | |
st.markdown(link, unsafe_allow_html=True) | |
selected_points = plotly_events( | |
fig, click_event=True, hover_event=False) | |
with col3: | |
try: | |
date2 = pd.Timestamp(selected_points[0]['x']).strftime('%Y%m%d') | |
data2 = png_data(date2) | |
m3 = map_folium(data2, gdf) | |
st.header(f'{var} on {pd.Timestamp(date2).strftime("%D")}') | |
st_folium(m3, height=300) | |
except: | |
pass | |
st.markdown(""" <style> | |
#MainMenu {visibility: hidden;} | |
footer {visibility: hidden;} | |
</style> """, unsafe_allow_html=True) |