Understanding Heating and Cooling Degree Days

Heating degree days (HDD) and cooling degree days (CDD) are two of the most-used numerical indices in the U.S. natural gas and power industries. They convert raw temperature observations and forecasts into a single number that correlates closely with daily energy demand for space conditioning. Before walking through specific use cases, it is worth understanding exactly how the index is constructed and why every desk in the energy industry uses the same base temperature.

Heating degree days for a single day are computed as the maximum of zero and 65 minus the daily mean temperature. Cooling degree days for the same day are the maximum of zero and the daily mean temperature minus 65. The daily mean is conventionally calculated as the simple average of the daily high and the daily low, although hourly-integrated means produce slightly different results.

If a city in Iowa has a daily high of 30°F and a daily low of 10°F, the daily mean is 20°F. HDD that day is max(0, 65 − 20) = 45. CDD is max(0, 20 − 65) = 0. If a city in Texas has a daily high of 100°F and a daily low of 80°F, the daily mean is 90°F. CDD that day is 25; HDD is 0. Days with means right around 65°F contribute almost nothing to either index.

The 65°F base is a U.S. industry convention dating back to the early twentieth century. It approximates the threshold at which a typical building starts requiring active heating or cooling — heat from occupants, lights, and appliances tends to keep indoor temperatures comfortable when outdoor means are around 65°F. Other countries use different bases (the United Kingdom commonly uses 15.5°C, which is about 60°F), and some specialized analytics use building-stock-specific bases like 60°F or 70°F. For most U.S. natural gas and power applications, 65°F is the default.

Daily degree days are summed across periods to create weekly, monthly, and seasonal totals. A typical U.S. winter month produces 800–1,400 HDD at most northern stations, while a typical summer month produces 300–500 CDD across the Sun Belt. Annual seasonal totals are tracked by NOAA for each weather station and compared against the 1991–2020 climatological normal, which is the official baseline for U.S. climate comparisons.

Regional degree days are aggregated as either simple averages across stations within a region or as load-weighted averages that emphasize population centers. For natural gas demand modeling, weighted averages by population or historical consumption tend to produce tighter regression fits. For ISO power demand, weighted averages by load served at each city often produce the cleanest signal.

The single most-watched derived metric is the departure from normal. A forecast HDD of 50 against a climatological normal of 30 is a +20 departure, meaning the day is forecast to be substantially colder than typical for that calendar date. A weekly forecast that runs 100 HDD above normal across the eastern U.S. is a major bullish signal for natural gas and power prices, because it implies materially higher heating demand than the market may currently be priced for.

Conversely, a forecast period that runs 50–100 HDD below normal in late November or early December is a bearish signal — it implies that early-winter heating demand will not arrive on schedule, and storage drawdown will be slower than expected. Each Thursday morning, the Energy Information Administration releases the weekly natural gas storage report, and weather expectations for the following week are immediately back-tested against the implied storage trajectory.

Weather Workbench computes HDD and CDD for each tracked city from National Weather Service gridpoint forecast data, then averages across all cities within an ISO to produce a single regional metric per day. The 1991–2020 NOAA climate normals at each city point provide the comparison baseline, and the ISO-level departure is shown alongside the daily totals. The forecast extends to roughly 7–10 days, which captures the highest-resolution National Weather Service grid outputs.

The displayed departures are not adjusted for population weighting at the city level — they are simple averages. For trading-quality analysis, most desks build their own population-weighted or load-weighted city baskets and apply their own normals. Weather Workbench is designed for situational awareness rather than as a substitute for that proprietary work. The methodology page on this site walks through every step of the calculation in detail.

Degree days are a useful but imperfect proxy for energy demand. They do not account for humidity (which materially affects perceived temperature and air conditioning load), wind chill (which affects building heat loss), solar radiation (which warms south-facing buildings during the day), or cloud cover. Specialized analytics use derivative indices like cooling-degree-day-equivalents or temperature-humidity index to capture these effects.

Degree days also lag behavioral changes. Long stretches of unusually mild weather can train building occupants to set thermostats differently, which complicates modeling of the next stretch of normal weather. And the 1991–2020 baseline incorporates a warming trend versus older 1981–2010 normals, which is something to remember when comparing across decades.

Used alongside the rest of the Weather Workbench dashboards — NWS forecasts at city scale, CPC outlooks, model comparison maps, and regional alerts — degree days give you a quick read on which way the demand fundamentals are tilting in any given ISO. That is most of what they were designed to do.

Definitions, baselines, and weekly products referenced in this article are documented in the public-use federal sources listed below.