Rainfall day-ending sensitivity and six-hour event-gap analysis

This analysis combines four ASOS stations and three Pennsylvania Environmental Monitoring Network stations. It tests how changing the 24-hour observation ending time redistributes rainfall among days, then examines the internal dry breaks of events defined by six observed dry hours.

Methods

Hydrologic days are 24-hour windows ending at local civil midnight, fixed 05Z, 11Z, 12Z, 13Z, 23Z, 00Z, or 01Z. A day is valid when at least 90% of its expected observations are present and it is not a partial edge window. A wet day contains at least 0.01 inch. Daily amount quintiles are fixed from the equal-station-weighted pooled distribution of wet days ending at 05Z, then applied unchanged to every ending time. This prevents the bins themselves from moving while the clock boundary is being tested.

PEMN slow-drain episodes identified by the earlier timing QC are removed from timing calculations. ASOS accumulation cycles are reconstructed from p01i; frozen-only events are excluded from the rain-compatible timing analysis, while mixed and unclassified measurable events remain. A primary event contains at least 0.02 inch. Events end after six observed dry hours or at a long data outage.

A dry break is the fully dry interval between two measurable observations. For a one-minute PEMN record, tips six minutes apart contain a five-minute break. For a five-minute ASOS record, reports ten minutes apart contain one five-minute dry slot. The continuation curve asks: after a break has already lasted t minutes, what fraction of eligible breaks still resume before reaching the six-hour event separator? Each complete event contributes one terminal event-ending spell.

Effect of day ending

The largest network-level change relative to fixed 05Z occurs for PEMN at 13Z: -4.65 wet days per 365.25 valid days. Total annualized depth changes little; the principal effect is whether one meteorological episode is concentrated into one daily total or split across two nominal days.

Six-hour event structure

Event length is measured from the first measurable rainfall observation to the last measurable observation in the same six-hour event. Mean intensity uses the inclusive active window so that one-tip events do not require division by zero, another modest victory over arithmetic.

When does a dry break become more likely to end the event?

The table below gives the first dry-break length at which the empirical chance of additional rain before six hours falls to 50% or lower. Values are assigned by the month in which the break begins, not necessarily the month in which the event began.

Seasonal probability thresholds

These include the requested 5%, 10%, 25%, 50%, 75%, 90%, and 95% resumption-probability crossing points. <5m means the probability was already below that level at the first evaluated five-minute break. >6h means it never fell that low before the event separator.

Interpretation

Day-ending sensitivity is not a change in rainfall climate. It is a bookkeeping transformation that changes the number of wet labels, the size of daily totals, and the frequency assigned to the upper bins. Morning endings tend to keep overnight and early-morning stratiform rainfall together, while evening or midnight endings more often divide long cool-season events. Convective events can be split whenever a late-day storm continues across the selected boundary.

The dry-break continuation thresholds are empirical properties of this station sample and six-hour definition. They are not universal physical constants. They depend on gauge resolution, measurable increment size, season, storm type, and the terminal-event rule. The station-level tables should be used before applying a network pooled threshold operationally.

Package contents

The tables directory contains every valid hydrologic day, fixed pooled/network/station quintile thresholds, wet-day and bin-frequency summaries by station, network, month, season, and ending time, full six-hour event catalogs, break-count statistics, dry-spell records, continuation curves, probability crossing thresholds, and internal-gap percentiles. The Python script reproduces the analysis from the uploaded archives.