π₯Ά Peter Sinks#
Demonstrates the following:
Using TimeSeries service
Convert UTC datetime to station local time
Plotting with Seaborn
Rolling mean for each station
Downsample for each station
The Peter Sinks in Northern Utah is one of the coldest places in the contiguous United States (Utah Climate Center, Wikipedia). There is a weather station at the top of the rim and another at the bottom. In is not uncommon for temperatures at the bottom station to be 40 or 50 degrees colder than the rim, and the elevation difference between the two stations is only 272 feet.
[1]:
from datetime import datetime
import matplotlib.dates as mdates
import matplotlib.pyplot as plt
import matplotlib.ticker as mticker
import polars as pl
import seaborn as sns
import synoptic
[2]:
df = synoptic.TimeSeries(
start=datetime(2023, 12, 19),
end=datetime(2023, 12, 27),
vars="air_temp",
stid="psink,psrim",
units="english",
).df()
# Convert date_time to Local Time
df = df.with_columns(pl.col("date_time").dt.convert_time_zone(df["timezone"][0]))
df
ππ¨ Speedy delivery from Synoptic's timeseries service.
π¦ Received data from 2 stations (0.36 seconds).
[2]:
shape: (1_538, 21)
| stid | date_time | variable | sensor_index | is_derived | value | units | id | name | elevation | latitude | longitude | mnet_id | state | timezone | elev_dem | period_of_record_start | period_of_record_end | qc_flagged | is_restricted | is_active |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| str | datetime[ΞΌs, America/Denver] | str | u32 | bool | f64 | str | u32 | str | f64 | f64 | f64 | u32 | str | str | f64 | datetime[ΞΌs, UTC] | datetime[ΞΌs, UTC] | bool | bool | bool |
| "PSINK" | 2023-12-18 17:00:00 MST | "air_temp" | 1 | false | 38.05 | "Fahrenheit" | 26896 | "Peter Sinks" | 8173.0 | 41.91302 | -111.51416 | 26 | "UT" | "America/Denver" | 8175.9 | 2010-01-01 00:00:00 UTC | 2024-12-12 04:30:00 UTC | false | false | true |
| "PSINK" | 2023-12-18 17:15:00 MST | "air_temp" | 1 | false | 37.9 | "Fahrenheit" | 26896 | "Peter Sinks" | 8173.0 | 41.91302 | -111.51416 | 26 | "UT" | "America/Denver" | 8175.9 | 2010-01-01 00:00:00 UTC | 2024-12-12 04:30:00 UTC | false | false | true |
| "PSINK" | 2023-12-18 17:30:00 MST | "air_temp" | 1 | false | 36.51 | "Fahrenheit" | 26896 | "Peter Sinks" | 8173.0 | 41.91302 | -111.51416 | 26 | "UT" | "America/Denver" | 8175.9 | 2010-01-01 00:00:00 UTC | 2024-12-12 04:30:00 UTC | false | false | true |
| "PSINK" | 2023-12-18 17:45:00 MST | "air_temp" | 1 | false | 35.21 | "Fahrenheit" | 26896 | "Peter Sinks" | 8173.0 | 41.91302 | -111.51416 | 26 | "UT" | "America/Denver" | 8175.9 | 2010-01-01 00:00:00 UTC | 2024-12-12 04:30:00 UTC | false | false | true |
| "PSINK" | 2023-12-18 18:00:00 MST | "air_temp" | 1 | false | 34.06 | "Fahrenheit" | 26896 | "Peter Sinks" | 8173.0 | 41.91302 | -111.51416 | 26 | "UT" | "America/Denver" | 8175.9 | 2010-01-01 00:00:00 UTC | 2024-12-12 04:30:00 UTC | false | false | true |
| … | … | … | … | … | … | … | … | … | … | … | … | … | … | … | … | … | … | … | … | … |
| "PSRIM" | 2023-12-26 16:00:00 MST | "air_temp" | 1 | false | 21.01 | "Fahrenheit" | 26897 | "Peter Sinks Rim" | 8445.0 | 41.91909 | -111.51022 | 26 | "UT" | "America/Denver" | 8444.9 | 2010-01-01 00:00:00 UTC | 2024-12-12 04:30:00 UTC | false | false | true |
| "PSRIM" | 2023-12-26 16:15:00 MST | "air_temp" | 1 | false | 21.68 | "Fahrenheit" | 26897 | "Peter Sinks Rim" | 8445.0 | 41.91909 | -111.51022 | 26 | "UT" | "America/Denver" | 8444.9 | 2010-01-01 00:00:00 UTC | 2024-12-12 04:30:00 UTC | false | false | true |
| "PSRIM" | 2023-12-26 16:30:00 MST | "air_temp" | 1 | false | 21.46 | "Fahrenheit" | 26897 | "Peter Sinks Rim" | 8445.0 | 41.91909 | -111.51022 | 26 | "UT" | "America/Denver" | 8444.9 | 2010-01-01 00:00:00 UTC | 2024-12-12 04:30:00 UTC | false | false | true |
| "PSRIM" | 2023-12-26 16:45:00 MST | "air_temp" | 1 | false | 21.48 | "Fahrenheit" | 26897 | "Peter Sinks Rim" | 8445.0 | 41.91909 | -111.51022 | 26 | "UT" | "America/Denver" | 8444.9 | 2010-01-01 00:00:00 UTC | 2024-12-12 04:30:00 UTC | false | false | true |
| "PSRIM" | 2023-12-26 17:00:00 MST | "air_temp" | 1 | false | 19.99 | "Fahrenheit" | 26897 | "Peter Sinks Rim" | 8445.0 | 41.91909 | -111.51022 | 26 | "UT" | "America/Denver" | 8444.9 | 2010-01-01 00:00:00 UTC | 2024-12-12 04:30:00 UTC | false | false | true |
[9]:
# Elevation Difference
df["elevation"].max() - df["elevation"].min()
[9]:
272.0
[3]:
plt.figure(figsize=(15, 5), dpi=100)
ax = plt.gca()
# Freezing Line
ax.axhline(32, ls="--", color="#00d0ff")
ax.text(
df["date_time"].max(),
32,
"Freezing",
va="bottom",
ha="right",
color="#00d0ff",
fontweight="bold",
)
# Station Temperature
sns.lineplot(
df,
ax=ax,
x="date_time",
y="value",
hue="stid",
)
# Cosmetics
ax.set_xlabel("Local Time - Year 2023")
ax.set_ylabel("Air Temperature (F)")
ax.set_title("Air Temperature", loc="left", fontsize=14, fontweight="bold")
ax.set_xlim(df["date_time"].min(), df["date_time"].max())
ax.xaxis.set_major_locator(mdates.DayLocator())
ax.xaxis.set_minor_locator(mdates.HourLocator(byhour=range(0, 24, 6)))
ax.xaxis.set_major_formatter(mdates.DateFormatter("%d %b\n%H:%M"))
ax.grid(color="w", lw=2)
ax.set_facecolor(".92")
ax.legend(title="Station ID", loc=0, bbox_to_anchor=(1, 1))
ax.spines["top"].set_visible(False)
ax.spines["right"].set_visible(False)
ax.spines["left"].set_visible(False)
Rolling Mean#
Plotting the 60-min mean temperature gives a more smooth looking graphic. Doing a rolling average will produce a dataframe with the same length as the original data.
Reference: polars.DataFrame.rolling
[4]:
# Rolling mean
df_60m_mean = df.rolling("date_time", period="60m", group_by="stid").agg(
pl.col("value").mean()
)
plt.figure(figsize=(15, 5), dpi=100)
ax = plt.gca()
# Freezing Line
ax.axhline(32, ls="--", color="#00d0ff")
ax.text(
df["date_time"].max(),
32,
"Freezing",
va="bottom",
ha="right",
color="#00d0ff",
fontweight="bold",
)
# Station Temperature
sns.lineplot(
df_60m_mean,
ax=ax,
x="date_time",
y="value",
hue="stid",
)
# Cosmetics
ax.set_xlabel("Local Time - Year 2023")
ax.set_ylabel("Air Temperature (F)")
ax.set_title("Air Temperature", loc="left", fontsize=14, fontweight="bold")
ax.set_xlim(df["date_time"].min(), df["date_time"].max())
ax.xaxis.set_major_locator(mdates.DayLocator())
ax.xaxis.set_minor_locator(mdates.HourLocator(byhour=range(0, 24, 6)))
ax.xaxis.set_major_formatter(mdates.DateFormatter("%d %b\n%H:%M"))
ax.grid(color="w", lw=2)
ax.set_facecolor(".92")
ax.legend(title="Station ID", loc=0, bbox_to_anchor=(1, 1))
ax.spines["top"].set_visible(False)
ax.spines["right"].set_visible(False)
ax.spines["left"].set_visible(False)
[5]:
# Rolling will produce a dataframe of the same length as the original
len(df), len(df_60m_mean)
[5]:
(1538, 1538)
Downsampled mean#
Using group_by_dynamic resamples your data to a coarser, and consistent, time interval. The dataframe will have fewer rows than the original.
[6]:
# Group by window
df_2h_group = df.group_by_dynamic("date_time", every="2h", group_by="stid").agg(
pl.col("value").mean()
)
plt.figure(figsize=(15, 5), dpi=100)
ax = plt.gca()
# Freezing Line
ax.axhline(32, ls="--", color="#00d0ff")
ax.text(
df["date_time"].max(),
32,
"Freezing",
va="bottom",
ha="right",
color="#00d0ff",
fontweight="bold",
)
# Station Temperature
sns.lineplot(df_2h_group, ax=ax, x="date_time", y="value", hue="stid", marker=".")
# Cosmetics
ax.set_xlabel("Local Time - Year 2023")
ax.set_ylabel("Air Temperature (F)")
ax.set_title("Air Temperature", loc="left", fontsize=14, fontweight="bold")
ax.set_xlim(df["date_time"].min(), df["date_time"].max())
ax.xaxis.set_major_locator(mdates.DayLocator())
ax.xaxis.set_minor_locator(mdates.HourLocator(byhour=range(0, 24, 6)))
ax.xaxis.set_major_formatter(mdates.DateFormatter("%d %b\n%H:%M"))
ax.grid(color="w", lw=2)
ax.set_facecolor(".92")
ax.legend(title="Station ID", loc=0, bbox_to_anchor=(1, 1))
ax.spines["top"].set_visible(False)
ax.spines["right"].set_visible(False)
ax.spines["left"].set_visible(False)
[7]:
# Resampling with group_by_dynamic reduces the dataframe length
len(df), len(df_2h_group)
[7]:
(1538, 194)