README.md

---
license: cc-by-4.0
language:
  - en
  - he
pretty_name: Jerusalem Air Quality — Weekday vs Shabbat
size_categories:
  - 1M<n<10M
tags:
  - air-quality
  - environmental
  - jerusalem
  - israel
  - time-series
  - shabbat
  - sabbath
  - pm25
  - no2
  - environmental-monitoring
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/readings.parquet
  - config_name: aqi_hourly
    data_files:
      - split: train
        path: data/aqi_hourly.parquet
  - config_name: weather_hourly
    data_files:
      - split: train
        path: data/weather_hourly.parquet
  - config_name: jlm_deweathered
    data_files:
      - split: train
        path: data/jlm_deweathered.parquet
  - config_name: control_deweathered
    data_files:
      - split: train
        path: data/control_deweathered.parquet
---

![Banner](banner.png)

# Jerusalem Air Quality — Weekday vs Shabbat

Twelve months of 5-minute-resolution air-quality readings from **12 monitoring
stations across Jerusalem**, formatted for analysing the weekday vs Friday vs
Shabbat (Saturday) pattern in urban pollution.

> 📊 **Companion analysis repository:**
> [`danielrosehill/JLM-Air-Quality-Analysis`](https://github.com/danielrosehill/JLM-Air-Quality-Analysis)
> — exact halachic relabeling, cross-city control analysis (London + New
> York), and the figure set summarised below. Reproducible scripts included.

The motivation: Jerusalem has a uniquely strong weekly traffic cycle — many
streets in observant neighbourhoods empty out from Friday afternoon to
Saturday night. This dataset is intended to support quantitative work on
the resulting air-quality signal.

## Source and attribution

All readings come from the **Israeli Ministry of Environmental Protection
(המשרד להגנת הסביבה)** national air-quality monitoring network, accessed
through the public portal at <https://air.sviva.gov.il/>.

The MoEP / sviva.gov.il site is the canonical source. This dataset is a
convenience repackaging of that public data into a tidy parquet file with
some derived columns (local time, day-of-week, Shabbat/Friday flags) and
12 months of history pre-aggregated for the Jerusalem stations.

If you use this data in research or downstream work, please cite the
**Israeli Ministry of Environmental Protection** as the original source.

## Coverage

- **Window:** 2025-04-28 → 2026-04-28 (12 months, full year)
- **Resolution:** 5 minutes (native sensor cadence)
- **Stations:** 12 active continuous monitoring stations within the
  Jerusalem region (`regionId == 8` in the source system).
- **Total valid readings:** ~7.1 million (only readings flagged `valid`
  by the source system are included; sensor-error / calibration rows are
  dropped).

### Stations

| station_id | Name (English) | Name (Hebrew) | Notes |
|---|---|---|---|
| 5   | Bar Ilan St | רחוב בר אילן | Major arterial through ultra-Orthodox neighbourhood |
| 13  | Baka | בקעה | Residential mixed |
| 36  | Safra Square | כיכר ספרא | City centre / municipal complex |
| 193 | Romema | רוממה | Ultra-Orthodox |
| 310 | Central Bus Station Platform 10 | רציף 10 תחנה מרכזית | Heavy bus traffic |
| 328 | Atarot industrial zone | אזור תעשייה עטרות | Industrial / north |
| 458 | Devorah HaNevia St | רחוב דבורה הנביאה | Residential |
| 547 | Highway 16 / Beit interchange | כביש 16-מחלף בייט | Highway interchange |
| 568 | Rockefeller Museum | מוזיאון רוקפלר | East Jerusalem corridor |
| 607 | Beit Safafa | בית צפאפא | South-east neighbourhood |
| 623 | Atarot Eden | עטרות עדן | North; partial coverage from 2025-08 |
| 635 | Bezalel St | ירושלים, רחוב בצלאל בזק | City centre; partial coverage from 2026-02 |

Full station metadata is in `stations.csv`.

### Pollutants and parameters

Per-station coverage varies — not every station measures every parameter.
The full set present in the dataset:

| code | pollutant_en | typical units |
|---|---|---|
| `NO` | Nitric oxide | ppb |
| `NOX` | Nitrogen oxides | ppb |
| `NO2` | Nitrogen dioxide | ppb |
| `CO` | Carbon monoxide | ppm |
| `O3` | Ozone | ppb |
| `SO2` | Sulphur dioxide | ppb |
| `PM2.5` | Fine particulates | µg/m³ |
| `PM10` | Coarse particulates | µg/m³ |
| `TSP` | Total suspended particulates | µg/m³ |
| `Benzene` | Benzene | ppb |
| `Toluene` | Toluene | ppb |
| `Temp`, `ITemp`, `RH`, `WS`, `WD` | Meteorological co-variates | — |

## Schema

The dataset ships **two parquet tables**: a long-format table of raw readings
with halachic labels, and a wide-format hourly table with derived AQI and
activity-isolating composite indices.

### `data/readings.parquet` — long form (one row per station × timestamp × pollutant)

| column | type | description |
|---|---|---|
| `station_id` | int | MoEP station identifier |
| `station_name_he` | string | Station name in Hebrew |
| `station_name_en` | string | Station name in English |
| `station_lat`, `station_lon` | float | Station coordinates (WGS-84) |
| `timestamp` | timestamp\[UTC\] | Reading timestamp in UTC |
| `timestamp_local` | timestamp\[Asia/Jerusalem\] | Same instant in local time (handles DST) |
| `dow` | int8 | Day of week, Sun=0 … Sat=6 (Israeli convention) |
| `hour` | int8 | Local hour 0–23 |
| `is_friday` | bool | Calendar Friday |
| **`is_shabbat`** | **bool** | **Exact halachic Shabbat: candle-lighting → tzeit hakochavim, weekly Shabbatot only** (Hebcal, Jerusalem times) |
| `shabbat_or_chag_exact` | bool | Either Shabbat or any chag with driving prohibition, exact halachic windows |
| `is_yom_tov` | bool | Inside a chag window (Pesach, Shavuot, Rosh Hashana, Sukkot, Shemini Atzeret) |
| `is_yom_kippur` | bool | Exact Yom Kippur window — broken out from `is_yom_tov` |
| `is_pre_shabbat` | bool | 2-hour window before candle-lighting (erev-Shabbat shopping rush) |
| `is_post_havdalah` | bool | 2-hour window after havdalah (post-Shabbat rebound) |
| `segment` | string | Mutually-exclusive bucket: `weekday`, `pre_shabbat`, `shabbat`, `post_havdalah`, `yom_tov`, `yom_kippur` (priority-ordered, highest wins). The recommended grouping column. |
| `holiday` | string \| null | Name of any Jewish holiday for that local date (joined with `; ` if multiple) |
| `pollutant` | string | Source code (e.g. `NO2`, `PM2.5`) |
| `pollutant_en` | string | English long name |
| `pollutant_he` | string | Hebrew long name |
| `pollutant_id` | int | MoEP pollutant identifier |
| `value` | float64 | Reading value |
| `units` | string | Reading units |

> ⚠️ **Schema change vs earlier revisions**: `is_shabbat` is now the **exact halachic** flag (candle-lighting → havdalah). The old naïve "calendar Saturday" semantics are no longer carried — use `dow == 6` if you need that. New columns: `is_yom_kippur`, `is_pre_shabbat`, `is_post_havdalah`, `segment`.

### `data/aqi_hourly.parquet` — wide form (one row per station × hour)

Pivots the long parquet to one row per station-hour with all measured
pollutants as columns, then adds EPA AQI sub-indices, the composite EPA
AQI, and two activity-isolating indices.

| column group | columns | description |
|---|---|---|
| Identifiers | `station_id`, `station_name_en`, `station_lat`, `station_lon`, `hour_bin`, `dow`, `hour` | One row per station × UTC hour |
| Halachic labels | `is_friday`, `is_shabbat`, `is_pre_shabbat`, `is_post_havdalah`, `is_yom_tov`, `is_yom_kippur`, `segment`, `holiday` | Same semantics as the long parquet |
| Hourly pollutant means | `pm25`, `pm10`, `no2`, `no`, `nox`, `co`, `o3`, `benzene`, `toluene`, `tsp` | Mean over the hour, native units (PM in µg/m³, gases in ppb / CO in ppm) |
| AQI averaging windows | `pm25_24h`, `pm10_24h`, `o3_8h`, `co_8h` | Rolling means as required by EPA AQI breakpoints |
| EPA AQI sub-indices | `aqi_pm25`, `aqi_pm10`, `aqi_o3`, `aqi_no2`, `aqi_co` | Linear-interpolated per-pollutant sub-indices using EPA breakpoints |
| Composite AQI | `aqi_epa`, `aqi_dominant`, `aqi_category` | Max of sub-indices; name of the pollutant driving it; verbal band (Good / Moderate / Unhealthy / …) |
| Canadian AQHI | `aqhi`, `aqhi_band` | Health Canada's additive, mortality-risk-weighted index of NO₂ + O₃ + PM2.5 (3-hour rolling means). Computed at *city* level (no single Jerusalem station measures all three pollutants) and broadcast to every station-row at that hour. Bands: Low 1–3, Moderate 4–6, High 7–10, Very High 11+. |

### `data/weather_hourly.parquet` — meteorological covariates

Hourly weather from [Open-Meteo's Historical Weather API](https://open-meteo.com/en/docs/historical-weather-api) (ERA5-backed reanalysis), one row per (`city`, `timestamp_utc`):

| column | description |
|---|---|
| `city` | `Jerusalem`, `London`, or `New York` |
| `timestamp_utc` | UTC hour |
| `temperature_2m` | °C at 2 m |
| `relative_humidity_2m` | % at 2 m |
| `wind_speed_10m` | m/s at 10 m |
| `wind_direction_10m` | degrees at 10 m |
| `pressure_msl` | hPa, sea-level reduced |
| `boundary_layer_height` | m — most important variable for surface pollution (low BLH traps emissions) |
| `precipitation` | mm |

Used for meteorological normalisation in `jlm_deweathered.parquet` and `control_deweathered.parquet`.

### `data/jlm_deweathered.parquet` and `data/control_deweathered.parquet` — meteorology-controlled series

Per-station, per-pollutant, per-hour pollutant series after Random Forest meteorological normalisation following the openair `deweather::buildMod` workflow. Each row contains:

| column | description |
|---|---|
| `station_id` (or `location_id`+`city` for controls) | Station |
| `timestamp_utc` | UTC hour |
| `pollutant_en` (or `pollutant`) | Pollutant |
| `value_raw` | Raw hourly mean concentration (same units as original) |
| `value_dw` | Deweathered concentration: `raw − RF_predicted + mean(raw)` |
| `units` | Native units |
| (Jerusalem only) | `is_shabbat`, `is_yom_tov`, `is_yom_kippur`, `segment`, `dow`, `hour` |

**Method.** A Random Forest is fit per pollutant per city, regressing hourly concentration on weather covariates (T, RH, wind speed, wind direction sin/cos, pressure, BLH, precipitation), seasonal day-of-year (sin/cos), and station ID. Hour-of-day and day-of-week are deliberately **excluded** from the feature set — they would let the model absorb the diurnal traffic cycle and the Shabbat / weekend effect along with weather, eliminating the very signal the analysis is trying to measure. Day-of-year stays in to capture seasonal trend (winter heating, summer photochemistry).

Subtracting the model's prediction from the raw value gives a residualised series whose Shabbat-vs-weekday contrast is attributable to *non-meteorological* factors. R² typically 0.7–0.95 depending on pollutant.

### `data/deweathered_comparison.csv` — summary

Headline weather-controlled Shabbat % drops, Jerusalem only:

| Pollutant | Raw drop | After weather control |
|---|---:|---:|
| NO | −80% | **−52%** |
| NOx | −69% | **−43%** |
| NO₂ | −59% | **−34%** |
| Toluene | −52% | **−17%** |
| PM10 | +11% | **−12%** *(paradox flips)* |
| Benzene | −27% | **−9%** |
| PM2.5 | +28% | **−4%** *(paradox dissolves)* |
| CO | −12% | −4% |
| O₃ | +3% | +1% |

Roughly **half the raw NO₂ drop is real**, the rest is weather. Both PM paradoxes (PM2.5 and PM10 going *up* on Shabbat) dissolve once meteorology is controlled — they were dust-event artefacts, not Shabbat effects.
| **Traffic-Combustion Index** | `tci`, `tci_n_inputs` | Mean of per-station z-scores (vs `segment == 'weekday'` daytime baseline) of NO₂, NO, NOx, CO, benzene, toluene. Negative = below the workweek norm. |
| **Photochemical/Dust Index** | `pd_index`, `pd_n_inputs` | Mean of z-scores of O₃ + PM2.5. Moves opposite to TCI on Shabbat — captures the "wrong-way" pollutants. |

The two custom indices are not standards. They exist because the EPA
composite AQI is dominated by particulates here and barely moves on Shabbat,
which obscures the (very real) drop in traffic-attributable pollutants.
TCI is the cars-and-construction signal; PD-Index is the photochemistry +
long-range-transport signal that goes the other way.

**AQI methodology in one paragraph.** Hourly station means are computed
from the long table. PM2.5 and PM10 are 24-hour-rolling-averaged before
applying their breakpoints; O₃ and CO are 8-hour-averaged; NO₂ uses the
1-hour value directly. Each pollutant produces a 0–500 sub-index via
linear interpolation between the EPA breakpoint table; the composite AQI
is the maximum sub-index at that station-hour. EPA's standard category
bands (Good ≤ 50, Moderate ≤ 100, Unhealthy for sensitive ≤ 150,
Unhealthy ≤ 200, Very Unhealthy ≤ 300, Hazardous ≤ 500) are applied to
the composite. SO₂ is not included because Jerusalem stations do not
report it.

## Quick start

```python
from datasets import load_dataset

ds = load_dataset("danielrosehill/Jerusalem-Air-Quality-Shabbat", split="train")
df = ds.to_pandas()

# Mean NO2 by day of week, for Bar Ilan St
import pandas as pd
no2 = df[(df.station_id == 5) & (df.pollutant == "NO2")]
print(no2.groupby("dow")["value"].mean())
```

Or directly with pandas / pyarrow:

```python
import pandas as pd

# Long-format raw readings with halachic labels
df = pd.read_parquet("hf://datasets/danielrosehill/Jerusalem-Air-Quality-Shabbat/data/readings.parquet")

# Wide-format hourly table with EPA AQI + Traffic-Combustion + Photochemical indices
aqi = pd.read_parquet("hf://datasets/danielrosehill/Jerusalem-Air-Quality-Shabbat/data/aqi_hourly.parquet")

# Composite-index summary by halachic segment
print(aqi.groupby("segment")[["aqi_epa", "tci", "pd_index"]].mean())
```

## Headline finding: the Jerusalem amplification factor

The most informative way to read this dataset is not the absolute Shabbat
% drop in Jerusalem (raw or deweathered), but the **ratio of the Jerusalem
off-day drop to the averaged London + NYC weekend drop, after controlling
for weather**. That ratio is the part that isn't captured anywhere else in
the literature: every city has a weekend cycle, but the question is how
much *additional* signal Jerusalem's Shabbat carries beyond it.

After meteorological deweathering (Random Forest with T, RH, wind, BLH,
pressure, precipitation; openair `deweather::buildMod` workflow):

| Pollutant | Jerusalem | West avg (London + NYC) | **Amplification** |
|---|---:|---:|---:|
| **NO₂** | **−34 %** | **−8 %** | **≈ 4 ×** |
| CO | −4 % | −4 % | ≈ 1 × |
| O₃ | +1 % | +2 % | ≈ 0.6 × |
| PM10 | −12 % | −1 % | ≈ 9 × *(small denom)* |
| PM2.5 | −4 % | −1 % | ≈ 7 × *(small denom)* |

**NO₂ at ~4× is the clean, defensible headline** because both the
numerator and denominator are large enough for the ratio to be robust.
The PM ratios are real but the Western denominators are small (London +
NYC weekend PM barely drops at all), so the ratios are inflated by a
near-zero divisor and should be reported with that caveat.

![Jerusalem amplification factor](figures/17_jerusalem_amplification.png)

### Why ~4× is plausible — what stops on Shabbat that doesn't stop on a Western weekend

The 4× ratio isn't well described as "Jerusalem has a stronger weekend
cycle." It's better described as **Shabbat being closer to a full
traffic-plus-construction ban than a Western weekend is**.

The Shabbat intervention removes essentially the entire urban NO₂
inventory at once:

- **Private cars** — sharp reduction city-wide, near-total in observant neighbourhoods
- **Diesel buses** — Egged, Superbus, Dan, intercity coaches all stop completely
- **Light-rail and high-rise construction** — Jerusalem is in an unusually active build-out phase (Green and Blue line works, dozens of high-rise sites), all of which run diesel excavators, drills, and generators six days a week and go silent on Shabbat. Non-road mobile machinery (NRMM) is a known top-3 NO₂ source in growing cities.
- **Trucks and freight delivery** — near-zero
- **Sherut / shared taxis** — stopped in observant areas

A typical Western weekend removes only part of this:

- **London**: cars and freight reduce ~25–30% (not zero), buses run reduced schedule, *some* central-zone construction restrictions, but domestic gas heating, Heathrow aviation, and most infrastructure works continue.
- **New York**: traffic and freight reduce ~25–30%, but **building heating with #4/#6 fuel oil** is uniquely large in NYC's NO₂ inventory and runs continuously regardless of the day of week. JFK/LGA aviation, peaker power generation, and bus service are all weekend-unchanged.

So Jerusalem's Shabbat removes *commuter traffic + buses + freight +
construction simultaneously*, while Western weekends remove only commuter
traffic and some freight. A factor of ~4 between those two interventions
is what a back-of-envelope source apportionment would predict — and it's
what the deweathered data shows.

### Why the other ratios sit where they do

- **CO ≈ 1×** because catalytic converters have pushed urban CO close to the noise floor in all three cities; whatever traffic CO remains is roughly proportional to the day's traffic reduction, and the magnitudes are all small.
- **O₃ ≈ 0.6×** is the *opposite* direction — Jerusalem's ozone bump on Shabbat is *smaller* than the Western weekend ozone bump despite a much larger NOx drop. This is the signature of NOx-saturated urban photochemistry: in this regime, a deeper NOx cut doesn't translate to a proportionally larger O₃ rise, because the local atmosphere flips into a different chemical regime at low NOx.
- **PM10 ≈ 9× and PM2.5 ≈ 7×** are real in direction but the Western denominators (London/NYC weekend PM drops of −1 to 0%) are too small for the ratio to be quantitatively stable. The right read is "Western PM doesn't have a weekend cycle, Jerusalem PM has a small but real Shabbat cycle once dust is controlled" — the construction halt is probably the dominant mechanism.

### What the table predicts about Yom Kippur

Yom Kippur removes the residual that Shabbat doesn't reach: through-traffic
on Highway 1, traffic in non-observant West Jerusalem, and Arab-sector
traffic in East Jerusalem and the corridors that pass through it. The
amplification ratio for Yom Kippur should therefore be even larger than
for Shabbat — and the data agrees: combined PM falls 74% on YK, NOx
collapses to a near-floor, and composite EPA AQI itself drops to 28 (Good
band) which it never does on a regular Shabbat. YK is the natural
extension of the same source-removal logic, with the last few categories
finally peeled away.

## Pre-deweathering Shabbat drops, Jerusalem only

For transparency, these are the *raw* (uncontrolled) Shabbat % drops, with
Friday sitting roughly half-way:

| Pollutant | Friday % drop vs Sun-Thu | Shabbat % drop vs Sun-Thu |
|---|---|---|
| NO    | ~40 % | ~55 % (range 19–76 %) |
| NOx   | ~34 % | ~52 % |
| NO₂   | ~30 % | ~47 % |
| Toluene | ~44 % | ~38 % |
| Benzene | ~25 % | ~21 % |
| CO    | ~7 %  | ~8 %  |

Biggest Shabbat drops are in observant West-Jerusalem neighbourhoods
(Bezalel ~73 %, Bar Ilan ~55 %, Romema ~52 %); smallest along corridors
carrying significant non-observant traffic (Rockefeller Museum ~13 %,
Safra Square ~33 %).

**Particulates (PM2.5, PM10) do _not_ drop on Shabbat — they often rise.**
A real signal but confounded by Saharan dust intrusions and weekend
grilling; needs its own controlled analysis.

![Day-of-week × hour heatmap](figures/03_jerusalem_dow_hour_heatmap.png)

The Saturday row is dramatically lighter than Sun–Thu; Friday is clearly
intermediate.

## Cross-city comparison

The same 12-month window was pulled from [OpenAQ](https://openaq.org/) for
**London** (3 stations) and **New York** (4 stations) and aggregated using
each city's *own* off-day convention.

**Off-day vs workweek % change, stratified by pollutant:**

| Pollutant | Jerusalem (Shabbat exact vs Sun–Thu) | London (Sat+Sun vs Mon–Fri) | New York (Sat+Sun vs Mon–Fri) |
|---|---:|---:|---:|
| NO₂  | **−62.4 %** | −20.0 % | −31.3 % |
| NO   | **−62.5 %** | — | −57.1 % |
| NOx  | **−63.4 %** | — | −39.1 % |
| CO   |   0.0 %    | −13.7 % | 0.0 % |
| O₃   |  +5.3 %   | +8.3 % | +7.1 % |
| PM10 | **−16.0 %** | −7.1 % | — |
| PM2.5 | −6.8 %  | +1.4 % | +1.6 % |
| SO₂  | —          | −45.0 % | — |

(Cells marked `—` are missing in OpenAQ's coverage for that station/pollutant
combination.)

**What the table shows:**

- **NOx-family (NO, NO₂, NOx) is where Jerusalem stands apart.** All three
  drop ~62–63 % on Shabbat. New York's weekend drop in this family runs
  ~31–57 % depending on species; London's NO₂ weekend drop is only −20 %.
  Jerusalem's signal is roughly 2× New York's and 3× London's — far larger
  than the generic secular weekend cycle.
- **O₃ rises slightly on off-days in every city** (~+5 to +8 %). Consistent
  with reduced NOx scavenging when traffic emissions fall — a universal
  photochemical signature, not a Shabbat-specific effect.
- **PM2.5 does not follow the traffic story.** Jerusalem shows only a
  small drop on Shabbat (and the source's `is_shabbat` calendar-day flag
  often shows PM2.5 *rising*). London and NYC weekend PM2.5 actually rises
  slightly — the same paradox plays out cross-city, suggesting weekend PM
  is dominated by non-traffic sources (cooking, dust events, secondary
  formation) everywhere.
- **PM10 drops modestly in Jerusalem (−16 %) and London (−7 %)**, intermediate
  between the strong NOx signal and the flat PM2.5 signal — coarser
  particulates carry more road-dust and resuspended traffic dust component
  than PM2.5.

![Cross-city off-day drop](figures/09_cross_city_offday_drop.png)

**Bottom line:** The Shabbat effect in Jerusalem is real, large, and
specifically a **traffic-emissions** story — the gap between Jerusalem and
the secular control cities is widest in NO/NO₂/NOx and narrowest (or
flipped) in PM2.5 and O₃. This pattern is consistent with a step-change
reduction in vehicle activity rather than any kind of generalised industrial
or commercial slowdown.

## AQI vs Traffic-Combustion Index

A reading of `aqi_hourly.parquet` by halachic segment exposes the headline
paradox: composite EPA AQI **barely moves** between workweek and Shabbat
because particulate matter dominates the index, and PM in Jerusalem is
driven by long-range dust as much as by local emissions.

| Segment | EPA AQI | Canadian AQHI | TCI (z) | PD-Index (z) |
|---|---:|---:|---:|---:|
| Workweek (Sun–Thu) | 64 | 4.7 | −0.10 | −0.05 |
| Pre-Shabbat (2h) | 63 | 4.7 | −0.58 | +0.20 |
| **Shabbat** | **67** | **4.4** | **−0.61** | **+0.15** |
| Post-Havdalah (2h) | 69 | 4.5 | −0.17 | +0.06 |
| Yom Tov | 71 | 4.1 | −0.66 | +0.03 |
| **Yom Kippur** | **28** | **3.0** | **−0.90** | **−0.55** |

Both standard composite indices — EPA's max-based AQI and Canada's
additive AQHI — are dominated by the Jerusalem PM2.5 baseline (which is
itself dominated by long-range dust transport). EPA AQI is essentially
flat on Shabbat (64→67); AQHI moves a little (4.7→4.4) but not in
proportion to the underlying NOx drop, because PM2.5 is the largest term
in its additive sum here. The Traffic-Combustion Index plunges by half a
standard deviation on Shabbat (and almost a full standard deviation on
Yom Kippur), while the Photochemical/Dust Index drifts the other way —
exactly as the NOx-saturated urban photochemistry would predict.

**The takeaway**: standard composites are calibrated for cities where
PM2.5 sources are mostly local. In Jerusalem they become "dust meters"
that are structurally blind to traffic-emission interventions. An
activity-attributable index like TCI is the right tool for evaluating
this kind of natural experiment.

## Halachic time-window methodology

The dataset ships two Shabbat flags: a coarse calendar one (`is_shabbat`,
all of Saturday) and an exact halachic one (`shabbat_or_chag_exact`,
candle-lighting → havdalah). Real analysis should use the latter, and
extend it. The companion analysis repo derives a 6-flag schema from the
Hebcal export bundled in this dataset (`hebcal-jerusalem.json`):

| Flag | Window |
|------|--------|
| `is_shabbat` | candle-lighting → havdalah, weekly Shabbatot only |
| `is_yom_tov` | onset → conclusion of biblical chagim with driving prohibition (incl. Yom Kippur) |
| `is_yom_kippur` | exact YK window — broken out, since YK is the cleanest natural experiment in the year |
| `is_pre_shabbat` | 2 hours **before** candle-lighting — captures the erev-Shabbat shopping rush |
| `is_post_havdalah` | 2 hours **after** havdalah — captures the post-Shabbat rebound |

Plus a derived `segment` column resolving these into a single
mutually-exclusive bucket: `weekday`, `pre_shabbat`, `shabbat`,
`post_havdalah`, `yom_tov`, `yom_kippur` (priority order, highest wins).

**Time anchors (Jerusalem, via Hebcal):**

- Candle-lighting (Shabbat & chag onset): 40 minutes before sunset (Jerusalem custom)
- Havdalah (Shabbat & chag conclusion): *tzeit hakochavim* (~3 medium stars)

**Why the custom segmentation matters:** the naïve "Saturday vs Tuesday"
comparison under-states the effect because (a) calendar Saturday includes
the post-Shabbat rebound after tzeit, and (b) calendar Friday includes a
full normal weekday morning rush. The exact halachic windows eliminate
both contamination sources.

**Chol HaMoed is intentionally not flagged** — driving is permitted, so
those days are indistinguishable from weekdays for the traffic hypothesis.

**Yom Kippur is broken out** because near-universal observance among
Israelis (including secular Israelis who drive on regular Shabbat) means
the traffic drop is steeper than for any other day — see the 72-hour case
study figure in the analysis repo.

In the 12-month window: **50 Shabbatot, 9 Yom Tov windows, 1 Yom Kippur**.

## Cross-city methodology

Because the secular weekly cycle differs from Jerusalem's, the control
cities use their own convention:

| City | Workweek | Off-days |
|------|----------|----------|
| Jerusalem | Sun–Thu (segment = `weekday`, daytime 07–22) | Shabbat (segment = `shabbat`, exact halachic window) |
| London / New York | Mon–Fri | Sat + Sun |

The cross-city headline metric — **off-day % change vs workweek** — uses
each city's own convention. This is conceptually apples-to-apples ("off-days
vs work-days for the local population"), even though calendar boundaries
differ. NO₂ values were converted from ppb (Jerusalem source) to µg/m³
(OpenAQ convention; 1 ppb ≈ 1.88 µg/m³ at 25 °C, 1 atm) for overlay charts.

OpenAQ control stations used:
- **London**: 148, 154, 155
- **New York**: 384, 626, 857, 1122

## Figure set

All figures available in [`figures/`](figures/) and in the analysis repo:

1. `01_jerusalem_hour_of_day.png` — NOx / NO₂ / PM2.5 hourly curves
2. `02_jerusalem_segment_bars.png` — Per-station % change vs workweek by segment
3. `03_jerusalem_dow_hour_heatmap.png` — NOx by day-of-week × hour
4. `04_jerusalem_yom_kippur.png` — 72-hour line plot bracketing YK 2025
5. `05_jerusalem_pm_paradox.png` — NOx falls, PM2.5 doesn't
6. `06_jerusalem_friday_split.png` — Friday's intermediate signal concentrates in afternoon
7. `07_jerusalem_segment_distribution.png` — Box plot by halachic segment
8. `08_cross_city_no2_weekly.png` — NO₂ day-of-week curve, three cities overlaid
9. `09_cross_city_offday_drop.png` — Headline cross-city panel

Reproducible via the scripts in
[`danielrosehill/JLM-Air-Quality-Analysis`](https://github.com/danielrosehill/JLM-Air-Quality-Analysis):

```bash
python scripts/relabel.py          # 6-flag halachic relabel
python scripts/build_controls.py   # OpenAQ -> tidy control parquet
python scripts/generate_figures.py # All 9 figures
```

Per-station summary tables are also in `analysis/` of this dataset; full
preliminary write-up is in `analysis/FINDINGS.md`.

## Caveats and known limitations

- **Two Shabbat flags are provided.** `is_shabbat` is the naïve calendar
  Saturday (Sun=6, Sat=5) and is included for backwards compatibility.
  `shabbat_or_chag_exact` is the **recommended** flag — it uses Hebcal's
  Jerusalem-specific candle-lighting (Friday eve, 40 minutes before
  sunset) and havdalah (≈42 minutes after sunset, three medium stars)
  times, and also covers Yom Tov days. There is a single set of
  Hebcal times applied to all stations; intra-Jerusalem variation is
  under one minute and not material for hourly analysis.
- **Jewish holidays are flagged but not removed.** Use `is_yom_tov` to
  exclude the seven major Israeli yamim tovim from your weekday
  baseline; use `holiday` to filter by name. **Yom Kippur in particular
  is the cleanest air day of the year** (Wed 2025-10-01 was YK 2025);
  leaving it in inflates the apparent cleanness of the weekday baseline.
- **Sensor validity flag is trusted as-is.** Only `valid: true` readings
  are included; the source system's QC decisions have not been
  re-verified.
- **Two stations have partial coverage**: 623 (Atarot Eden) starts
  2025-08-06; 635 (Bezalel St) starts 2026-02-05. Their statistics use
  fewer Saturdays.
- **Station 509** (Malchei Israel) is listed as active in the source
  metadata but returned no historical data over this window and is
  omitted from `data/readings.parquet`.
- **Region filter:** stations were selected by `regionId == 8` in the
  source system. Two stations on the periphery of the region (Highway 16
  and Atarot) sit outside the historical city core but are functionally
  part of the Jerusalem network.

## License

The underlying data is published by the Israeli Ministry of Environmental
Protection as a public-information service. This dataset is shared under
**CC-BY-4.0** with attribution to the Ministry as the original data
source.

## Versioning

- **v1**   (2026-04-28): initial release. 12 months × 5-min × 12 stations,
  long-form parquet.
- **v1.1** (2026-04-28): adds Hebcal-derived columns
  `shabbat_or_chag_exact`, `is_yom_tov`, and `holiday` for
  Jerusalem-precise temporal filtering. Underlying readings unchanged.

## External data sources

- **Air-quality readings:** Israeli Ministry of Environmental Protection
  (<https://air.sviva.gov.il/>).
- **Shabbat candle / havdalah times and Jewish-calendar metadata:**
  [Hebcal](https://www.hebcal.com/), public REST API,
  Jerusalem (geonameid 281184).