These daily weather records were compiled from a subset of stations in the Global Historical Climatological Network (GHCN)-Daily dataset. A weather record is considered broken if the value exceeds the maximum (or minimum) value recorded for an eligible station. A weather record is considered tied if the value is the same as the maximum (or minimum) value recorded for an eligible station. Daily weather parameters include Highest Min/Max Temperature, Lowest Min/Max Temperature, Highest Precipitation, Highest Snowfall and Highest Snow Depth. All stations meet defined eligibility criteria. For this application, a station is defined as the complete daily weather records at a particular location, having a unique identifier in the GHCN-Daily dataset. For a station to be considered for any weather parameter, it must have a minimum of 30 years of data with more than 182 days complete in each year. This is effectively a 30-year record of service requirement, but allows for inclusion of some stations which routinely shut down during certain seasons. Small station moves, such as a move from one property to an adjacent property, may occur within a station history. However, larger moves, such as a station moving from downtown to the city airport, generally result in the commissioning of a new station identifier. This tool treats each of these histories as a different station. In this way, it does not thread the separate histories into one record for a city. Records Timescales are characterized in three ways. In order of increasing noteworthiness, they are Daily Records, Monthly Records and All Time Records. For a given station, Daily Records refers to the specific calendar day: (e.g., the value recorded on March 7th compared to every other March 7th). Monthly Records exceed all values observed within the specified month (e.g., the value recorded on March 7th compared to all values recorded in every March). All-Time Records exceed the record of all observations, for any date, in a station's period of record. The Date Range and Location features are used to define the time and location ranges which are of interest to the user. For example, selecting a date range of March 1, 2012 through March 15, 2012 will return a list of records broken or tied on those 15 days. The Location Category and Country menus allow the user to define the geographic extent of the records of interest. For example, selecting Oklahoma will narrow the returned list of records to those that occurred in the state of Oklahoma, USA. The number of records broken for several recent periods is summarized in the table and updated daily. Due to late-arriving data, the number of recent records is likely underrepresented in all categories, but the ratio of records (warm to cold, for example) should be a fairly strong estimate of a final outcome. There are many more precipitation stations than temperature stations, so the raw number of precipitation records will likely exceed the number of temperature records in most climatic situations.

This dataset contains Raleigh Durham International Airport weather data pulled from the NOAA web service described at Climate Data Online: Web Services Documentation. We have pulled this data and converted it to commonly used units. This dataset is an archive - it is not being updated.

Historical changes of annual temperature and precipitation indices at selected 210 U.S. cities

This dataset provide:

Annual average temperature, total precipitation, and temperature and precipitation extremes calculations for 210 U.S. cities.

Historical rates of changes in annual temperature, precipitation, and the selected temperature and precipitation extreme indices in the 210 U.S. cities.

Estimated thresholds (reference levels) for the calculations of annual extreme indices including warm and cold days, warm and cold nights, and precipitation amount from very wet days in the 210 cities.

Annual average of daily mean temperature, Tmax, and Tmin are included for annual average temperature calculations. Calculations were based on the compiled daily temperature and precipitation records at individual cities.

Temperature and precipitation extreme indices include: warmest daily Tmax and Tmin, coldest daily Tmax and Tmin , warm days and nights, cold days and nights, maximum 1-day precipitation, maximum consecutive 5-day precipitation, precipitation amounts from very wet days.

Number of missing daily Tmax, Tmin, and precipitation values are included for each city.

Rates of change were calculated using linear regression, with some climate indices applied with the Box-Cox transformation prior to the linear regression.

The historical observations from ACIS belong to Global Historical Climatological Network - daily (GHCN-D) datasets. The included stations were based on NRCC’s “ThreadEx” project, which combined daily temperature and precipitation extremes at 255 NOAA Local Climatological Locations, representing all large and medium size cities in U.S. (See Owen et al. (2006) Accessing NOAA Daily Temperature and Precipitation Extremes Based on Combined/Threaded Station Records).

Resources:

See included README file for more information.

Additional technical details and analyses can be found in: Lai, Y., & Dzombak, D. A. (2019). Use of historical data to assess regional climate change. Journal of climate, 32(14), 4299-4320. https://doi.org/10.1175/JCLI-D-18-0630.1

Other datasets from the same project can be accessed at: https://kilthub.cmu.edu/projects/Use_of_historical_data_to_assess_regional_climate_change/61538

ACIS database for historical observations: http://scacis.rcc-acis.org/

GHCN-D datasets can also be accessed at: https://www.ncei.noaa.gov/data/global-historical-climatology-network-daily/

Station information for each city can be accessed at: http://threadex.rcc-acis.org/

• 2024 August updated -

  Annual calculations for 2022 and 2023 were added.

  Linear regression results and thresholds for extremes were updated because of the addition of 2022 and 2023 data.

  Note that future updates may be infrequent.

• 2022 January updated -

  Annual calculations for 2021 were added.

  Linear regression results and thresholds for extremes were updated because of the addition of 2021 data.

• 2021 January updated -

  Annual calculations for 2020 were added.

  Linear regression results and thresholds for extremes were updated because of the addition of 2020 data.

• 2020 January updated -

  Annual calculations for 2019 were added.

  Linear regression results and thresholds for extremes were updated because of the addition of 2019 data.

  Thresholds for all 210 cities were combined into one single file – Thresholds.csv.

• 2019 June updated -

  Baltimore was updated with the 2018 data (previously version shows NA for 2018) and new ID to reflect the GCHN ID of Baltimore-Washington International AP. city_info file was updated accordingly.

  README file was updated to reflect the use of "wet days" index in this study. The 95% thresholds for calculation of wet days utilized all daily precipitation data from the reference period and can be different from the same index from some other studies, where only days with at least 1 mm of precipitation were utilized to calculate the thresholds. Thus the thresholds in this study can be lower than the ones that would've be calculated from the 95% percentiles from wet days (i.e., with at least 1 mm of precipitation).

Climate Data Online (CDO) provides free access to NCDC's archive of global historical weather and climate data in addition to station history information. These data include quality controlled daily, monthly, seasonal, and yearly measurements of temperature, precipitation, wind, and degree days as well as radar data and 30-year Climate Normals.

The average temperature in the contiguous United States reached 55.5 degrees Fahrenheit (13 degrees Celsius) in 2024, approximately 3.5 degrees Fahrenheit higher than the 20th-century average. These levels represented a record since measurements started in ****. Monthly average temperatures in the U.S. were also indicative of this trend. Temperatures and emissions are on the rise The rise in temperatures since 1975 is similar to the increase in carbon dioxide emissions in the U.S. Although CO₂ emissions in recent years were lower than when they peaked in 2007, they were still generally higher than levels recorded before 1990. Carbon dioxide is a greenhouse gas and is the main driver of climate change. Extreme weather Scientists worldwide have found links between the rise in temperatures and changing weather patterns. Extreme weather in the U.S. has resulted in natural disasters such as hurricanes and extreme heat waves becoming more likely. Economic damage caused by extreme temperatures in the U.S. has amounted to hundreds of billions of U.S. dollars over the past few decades.

Dataset from National Environment Agency. For more information, visit https://data.gov.sg/datasets/d_03bb2eb67ad645d0188342fa74ad7066/view

Clean, reliable data – Weather and more!

We’ve spent millions of dollars over the last 25+ years working with global weather data space. Raw data, from a multitude of government sources, private weather networks and Earth Observation platforms, is assimilated into proprietary numerical weather prediction (NWP) models.

This Model output, analyzed and cleansed by the latest AI technologies gives Weather Data AI a clean, proprietary, high resolution global gridded weather database, available to you for whatever purpose you may need.

Hourly geographically aggregated weather data for Europe. This data package contains radiation and temperature data, at hourly resolution, for Europe, aggregated by Renewables.ninja from the NASA MERRA-2 reanalysis. It covers the European countries using a population-weighted mean across all MERRA-2 grid cells within the given country.

Historical weather data is essential for understanding environmental trends, assessing climate risk, and building predictive models for infrastructure, agriculture, and sustainability initiatives. Among all variables, temperature and humidity serve as core indicators of environmental change and operational risk.

Ambios offers high-resolution Historical Weather Data focused on temperature and humidity, sourced from over 3,000+ first-party sensors across 20 countries. This dataset provides hyperlocal, verified insights for data-driven decision-making across industries.

-Historical weather records for temperature and humidity -First-party sensor data from a decentralized network -Global coverage across 20 countries and diverse climate zones -Time-stamped, high-frequency measurements with environmental context -Designed to support ESG disclosures, research, risk modeling, and infrastructure planning

Use cases include:

-Long-term climate trend analysis and model validation -Historical baselining for ESG and sustainability frameworks -Resilience planning for heatwaves, humidity spikes, and changing climate conditions -Agricultural research and water management strategy -Infrastructure and energy load forecasting -Academic and scientific studies on regional weather patterns

Backed by Ambios’ decentralized physical infrastructure (DePIN), the data is reliable, traceable, and scalable—empowering organizations to make informed decisions grounded in historical environmental intelligence.

Whether you're building ESG models, planning smart infrastructure, or conducting climate research, Ambios Historical Weather Data offers the precision and credibility needed for long-term environmental insight.

The backbone of CustomWeather's forecasting arm is our proven, high-resolution model, the CW100. The CW100 Model is based on physics, not statistics or airport observations. As a result, it can achieve significantly better accuracy than statistical models, especially for non-airport locations. While other forecast models are designed to forecast the entire atmosphere, the CW100 greatly reduces computational requirements by focusing entirely on conditions near the ground. This reduction of computations allows it to resolve additional physical processes near the ground that are not resolved by other models. It also allows the CW100 to operate at a much higher resolution, typically 100x finer than standard models and other gridded forecasts.

Detailed Forecasts:
Features a detailed 48-hour outlook broken into four segments per day: morning, afternoon, evening, and overnight. Each segment provides condition descriptions, high/low temperatures, wind speed and direction, humidity, comfort level, UV index, expected and probability of precipitation, 6-hr forecasted precip amounts, and miles of visibility. Available for over 85,000 forecast points globally. The information is updated four times per day.

Extended Forecasts Days 1-15:
Features condition descriptions, high/low temperatures, wind speed and direction, humidity, comfort level, UV index, expected and probability of precipitation, and miles of visibility. Available for over 85,000 forecast points globally. The information is updated four times per day.

Hour-by-Hour Forecasts: Features Hour-by-Hour forecasts. The product is available as 12 hour, 48 hour and 168 hour blocks. Each hourly forecast includes weather descriptions, wind conditions, temperature, dew point, humidity, visibility, rainfall totals, snowfall totals, and precipitation probability. Available for over 85,000 forecast points globally. Updated four times per day.

Historical Longer Term Forecasts: Includes historical hourly and/or daily forecast data from 2009 until present date. Data will include condition descriptions, high/low temperatures, wind speed and direction, dew point, humidity, comfort level, UV index, probability of precipitation, rainfall and snowfall amounts. Available for over 85,000 forecast points globally. The information is updated four times per day.

Below are available time periods per each type of forecast from the GFS model and from CustomWeather's proprietary CW100 model:

GFS: 7-day hourly forecasts from August 2nd 2009; 48-hour to 5-day detailed forecasts from August 4th 2009; 15-day forecasts from October 9th 2008.

CW100: 7-day hourly forecasts from November 27, 2012; 48-hour detailed forecasts from November 12, 2011; 7-day forecasts from December 6, 2010, 15-day forecasts from August 6, 2012. CW100 is CustomWeather's proprietary model.

MOS: (Model Output Statistics) for any global location using archive of model and observation data. 0.25 degree resolution. 15-day hourly forecasts from January 1, 2017; 15-day forecasts from April 19, 2017.

Saudi Arabia hourly climate integrated surface data with the below data observations, WindSky conditionVisibilityAir temperatureDewSea level pressureNote: The dataset will contain the last 5 years hourly data, however, check the attachments section in this dataset if you need historical data.

WeatherDataAI Single-Point is a simple yet powerful tool for accessing global historical weather data with just a few clicks. Users select the weather variable(s) they need (like temperature, precipitation, etc.), choose the years of interest, and click a location on the interactive map. The platform automatically fetches the data, creates a custom CSV file, and provides a download link while also emailing the file directly to the user. No technical skills required — it’s built for researchers, businesses, and curious minds who need fast, accurate weather insights without coding or complex interfaces.

This dataset contains historical weather observations from various global locations, with its data provided at two temporal resolutions: daily and hourly. It includes core meteorological variables such as temperature, precipitation, wind, humidity, and atmospheric pressure, along with geospatial and temporal metadata for each observation. The dataset covers diverse geographic regions, including cities all arround the world, and supports both short-term event analysis and long-term climate trend exploration.

Year: year MeanLayDate: mean Julian date when first egg of the each clutch was laid ENSOWinter: mean ENSO score from December to March. Hurricanes: total number of hurricanes in the North Atlantic basin DaysBelow18_max: number of days with maximum daily temperature below 18.5 degrees Celsius or with rain during the 28 days after the mean fledging date. A crude measurement of weather conditions post fledging. TimePeriod: population trajectory at the time (growing, declining, post-decline)

Context

Who among us doesn't talk a little about the weather now and then? Will it rain tomorrow and get so cold to shake your chin or will it make that cracking sun? Does global warming exist?

With this dataset, you can apply machine learning tools to predict the average temperature of Detroit city based on historical data collected over 5 years.

Content

The given data set was produced from the Historical Hourly Weather Data [https://www.kaggle.com/selfishgene/historical-hourly-weather-data], which consists of about 5 years of hourly measurements of various weather attributes (eg. temperature, humidity, air pressure) from 30 US and Canadian cities.

From this rich database, a cutout was made by selecting only the city of Detroit (USA), highlighting only the temperature, converting it to Celsius degrees and keeping only one value for each date (corresponding to the average daytime temperature - from 9am to 5pm).

In addition, temperature values ​​were artificially and gradually increased by a few Celsius degrees over the available period. This will simulate a small global warming (or is it local?)...

In summary, the available dataset contains the average daily temperatures (collected during the day), artificially increased by a certain value, for the city of Detroit from October 2012 to November 2017.

The purpose of this dataset is to apply forecasting models in order to predict the value of the artificially warmed average daily temperature of Detroit.

See graph in the following image: black dots refer to the actual data and the blue line represents the predictive model (including a confidence area).

https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F3089313%2Faf9614514242dfb6164a08c013bf6e35%2Fplot-ts2.png?generation=1567827710930876&alt=media" alt="">

Acknowledgements

This dataset wouldn't be possible without the previous work in Historical Hourly Weather Data.

Inspiration

What are the best forecasting models to address this particular problem? TBATS, ARIMA, Prophet? You tell me!

The Alaska Climate Center (ACC) houses the historical weather records for the state of Alaska (mostly official U.S. Government data). These are in the form of raw handwritten data, summarized coded forms, surface and upper air weather map analyses, and selected digital data. They are stored in map sets, report sets, catalogs/indexes, individual maps, bibliographies, file folders, micrographics, and microfiche. It is not a single data base, but rather an extensive repository of historical records from communities and areas throughout the state. These individual data records and reports number in the thousands. The Center publishes a series of Alaska climate technical reports and Arctic climate atlases. In addition, it serves as a repository for publications on Arctic climate-related matters.

The monthly average temperature in the United States between 2020 and 2025 shows distinct seasonal variation, following similar patterns. For instance, in April 2025, the average temperature across the North American country stood at 12.02 degrees Celsius. Rising temperatures Globally, 2016, 2019, 2021 and 2024 were some of the warmest years ever recorded since 1880. Overall, there has been a dramatic increase in the annual temperature since 1895. Within the U.S. annual temperatures show a great deal of variation depending on region. For instance, Florida tends to record the highest maximum temperatures across the North American country, while Wyoming recorded the lowest minimum average temperature in recent years. Carbon dioxide emissions Carbon dioxide is a known driver of climate change, which impacts average temperatures. Global historical carbon dioxide emissions from fossil fuels have been on the rise since the industrial revolution. In recent years, carbon dioxide emissions from fossil fuel combustion and industrial processes reached over 37 billion metric tons. Among all countries globally, China was the largest emitter of carbon dioxide in 2023.

This dataset provides daily summaries of weather conditions at Norfolk International Airport, sourced from the National Oceanic and Atmospheric Administration (NOAA). NOAA publishes this data as part of their Global Historical Climatology Network – Daily dataset. It includes essential metrics such as maximum temperature, minimum temperature, average temperature, precipitation, snowfall, and average wind speed. The dataset is updated daily.

The National Forest Climate Change Maps project was developed by the Rocky Mountain Research Station (RMRS) and the Office of Sustainability and Climate to meet the needs of national forest managers for information on projected climate changes at a scale relevant to decision making processes, including forest plans. The maps use state-of-the-art science and are available for every national forest in the contiguous United States with relevant data coverage. Currently, the map sets include variables related to precipitation, air temperature, snow (including snow residence time and April 1 snow water equivalent), and stream flow.

Historical (1975-2005) and future (2071-2090) precipitation and temperature data for the contiguous United States are ensemble mean values across 20 global climate models from the CMIP5 experiment (https://journals.ametsoc.org/doi/abs/10.1175/BAMS-D-11-00094.1), downscaled to a 4 km grid. For more information on the downscaling method and to access the data, please see Abatzoglou and Brown, 2012 (https://rmets.onlinelibrary.wiley.com/doi/full/10.1002/joc.2312) and the Northwest Knowledge Network (https://climate.northwestknowledge.net/MACA/). We used the MACAv2- Metdata monthly dataset; average temperature values were calculated as the mean of monthly minimum and maximum air temperature values (degrees C), averaged over the season of interest (annual, winter, or summer). Absolute and percent change were then calculated between the historical and future time periods.

Historical (1975-2005) and future (2071-2090) precipitation and temperature data for the state of Alaska were developed by the Scenarios Network for Alaska and Arctic Planning (SNAP) (https://snap.uaf.edu). These datasets have several important differences from the MACAv2-Metdata (https://climate.northwestknowledge.net/MACA/) products, used in the contiguous U.S. They were developed using different global circulation models and different downscaling methods, and were downscaled to a different scale (771 m instead of 4 km). While these cover the same time periods and use broadly similar approaches, caution should be used when directly comparing values between Alaska and the contiguous United States.

Raster data are also available for download from RMRS site (https://www.fs.usda.gov/rm/boise/AWAE/projects/NFS-regional-climate-change-maps/categories/us-raster-layers.html), along with pdf maps and detailed metadata (https://www.fs.usda.gov/rm/boise/AWAE/projects/NFS-regional-climate-change-maps/downloads/NationalForestClimateChangeMapsMetadata.pdf).

About San Francisco San Francisco is a vibrant and dynamic city located on the west coast of the United States, in the state of California. Known for its hilly terrain, diverse neighborhoods, and iconic landmarks such as the Golden Gate Bridge and Alcatraz Island, San Francisco is a hub of culture, creativity, and innovation. The city is renowned for its world-class restaurants, thriving arts scene, and historic architecture, and is home to many tech companies and startups. With its mild climate, stunning views, and rich history, San Francisco is a must-visit destination for travelers from around the world.

About Dataset This dataset contains daily weather observations for San Francisco, USA from January 1, 1993 to January 1, 2023. The data is collected from Meteostat. The dataset contains 10 columns with 10958 rows.