Historical climate change is recorded
Executive summary
Historical climate change is documented through overlapping streams of instrumental records and proxy archives that together show long-term variability across centuries to millions of years and an unprecedented recent rise in global surface temperature; instrumental datasets indicate about a 2°F increase since the mid-1800s and the 10 warmest years on record have occurred in 2015–2024 [1] [2]. Multiple consolidated databases and recent digitization projects have expanded the archive of daily and longer-term records, but methodological differences and gaps mean raw historical records have underestimated some warming without careful correction [3] [4].
1. What the record actually is: instrumental versus proxy archives
The historical climate record is not a single continuous thermometer reading but a mosaic: direct instrumental measurements from land stations, ships and satellites cover roughly the last 150–200 years, while proxy archives—tree rings, ice cores, corals, sediments, cave deposits and even historical documents—extend reconstructions back hundreds to millions of years [5] [2] [6].
2. How scientists build the record: databases and cross-checks
Large curated datasets stitch together observations from tens of thousands of stations and many proxy series; examples include the Global Historical Climatology Network (GHCNd) and NOAA’s Climate Data Online for instrumental daily and monthly data, and global efforts to digitize historical meteorological series submitted to Copernicus and GLAMOD [7] [8] [3]. Paleoclimatologists reconstruct past climates by comparing multiple proxy records to identify consistent patterns and thereby increase confidence in long-term trends [2] [3].
3. What the data show about recent warming
Instrumental records show Earth's surface temperature has risen roughly 2°F (about 1.1°C) since the start of the NOAA record in the mid-19th century, and NOAA reports the 10 warmest years in the historical record occurred in 2015–2024 with 2024 setting a new anomaly record above 2023 [1]. Analyses of longer-term reconstructions place this recent rise as unusually rapid compared with typical pre‑industrial centennial variability, and many studies find an acceleration of warming through the 20th and 21st centuries [2] [9].
4. Limits, biases and the parts of warming that can be “missed”
Historical records are subject to systematic quirks—mixing of sea-surface and air temperatures, sparse Arctic coverage early in the record, changes in instrumentation and station siting—that can hide warming if not accounted for; a NASA-led study estimated historical records miss about 19% of global air-temperature warming since the 1860s if these issues are not corrected [4]. Inhomogeneities remain a challenge in raw daily compilations and require homogenization and careful metadata work to correct biases from relocations, instrument changes, and urbanization [10] [7].
5. Why multiple methods and models matter for confidence
Because different archives and techniques carry different errors and time resolutions, scientists combine proxy reconstructions, instrumental datasets, and climate models to cross-validate trends; when apples-to-apples adjustments are made (for example comparing air temperatures to model air temperatures), observational and modeled warming align much more closely than raw comparisons suggest [4] [3]. Independent reconstructions that converge on similar large-scale features increase confidence, while areas with sparse records remain more uncertain and are an explicit limitation of current reporting [2] [11].
6. Bottom line and open questions
Historical climate change is robustly recorded: multiple independent archives show long-term variability and a rapid, recent warming that stands out in the context of the past centuries and millennia, yet the archive requires ongoing digitization, homogenization and methodological transparency because raw historical data have systematically underestimated some warming and gaps persist in space and time [1] [3] [4]. Where data are thin—ancient intervals, remote regions, or before standardized instrumentation—scientists are explicit about uncertainty and rely on proxy-model synthesis rather than single-line narratives [2] [11].