HadISD: Sub-daily, non-interpolated weather station data

The following was contributed by Colin Raymond, May 2023:

HadISD is a premier choice for analyses aiming to investigate surface weather or climate with a high degree of precision and without reliance on interpolation algorithms of any sort. These attributes are the source of both its strengths and limitations. On one hand, HadISD data can shed light on diurnal cycles, short-duration events such as extremes, localized phenomena, and variables that are poorly represented in reanalyses and models, even those run at regional scales. For example, it has been used to good effect in examining regional and local characteristics of heat waves and heavy precipitation events, revealing details missed or spatiotemporally smoothed out by gridded products. On the other hand, certain instrumentation and measurement errors can (infrequently) linger despite algorithmic attempts at detecting and removing them, and other sources of uncertainty such as station moves or nearby land-cover change are not accounted for.

Overall, HadISD is an authoritative station-based dataset with frequent updates and a large body of citing literature. Aside from its value as a standalone dataset, it can serve as a benchmark observational reference, particularly for variables with frequent fluctuations or pervasive model biases, such as humidity, clouds, and winds. Unlike the Global Historical Climatology Network [GHCN], HadISD contains all available subdaily data, though it is missing certain GHCN variables like evaporation. GHCN also offers more stations, through inclusion of cooperative observing locations and stations with substantial missing records or short durations; most of the good-quality stations globally are present in both datasets.

No single dataset ever holds complete advantage over all others or provides a complete picture by itself. However, taking into account the above caveats, HadISD's combination of high temporal resolution, general reliability, and frequent usage in peer-reviewed studies make it a strong choice for understanding patterns or trends at anything from a local to a global perspective.

Key Strengths

The station set is comprehensive and routinely updated, enabling an as-close-to-truly-global analysis as is possible without making recourse to a gridded product which of course involves uncertainties of other sorts. Because clearly problematic data present in ISD have largely been filtered out, in areas with good data coverage, calculations can generally proceed straight ahead without the need to delve into further QC efforts. The lack of imposed inhomogeneity corrections or other adjustments allows users to inspect and analyze the raw observations as they best see fit. HadISD has been frequently cited in peer-reviewed work and also serves as the basis for derived products such as HadISDH, a package of monthly-mean humidity-related variables.

Key Limitations

Only temperature, dewpoint, pressure, and wind and cloud information are quality-controlled, and each variable's QC tests are run separately. For this reason and others, detailed analysis can still occasionally encounter QC issues. To aid in understanding how the data have been affected by the QC process, QC flags and the data removed in consequence are also included. Data are not homogenized, so temporal agreement cannot be guaranteed, and this goes for spatial consistency among nearby stations too. Additional information is available at the "online material" page.