Beyond well-known risks to person and property, lightning plays an important role in the Earth system. Lightning influences the chemical composition of the atmosphere, is the principle non-anthropogenic cause of wildfire ignitions , and is an important source of high energy radiation that affects atmospheric electricity, e.g., the propagation of radio waves. Quantifying the effects of lightning on the Earth system, and understanding where and how lightning presents hazards, requires an estimate of the timing, geographical distribution, and intensity of lightning strokes at continental to global scales. Large-scale maps of lightning occurrence are as important as those for temperature or precipitation in some land surface and atmospheric chemistry models , and are valuable in their own right for understanding various meteorological phenomena such as the frequency and distribution of extreme precipitation e.g., and for risk and hazard assessment e.g.,. Observing and mapping lightning distribution at large spatial scales has thus been a priority for the community for nearly a century .

While scientific observations of lightning occurrence have been recorded for more than 200 years , only recently has it become possible to create datasets with continuous global coverage. The first global datasets of lightning occurrence were derived from spaceborne remote sensing and remain an important tool for researchers. For nearly two decades, the only global lightning climatology freely available to researchers has been the Lightning Imaging Sensor – Optical Transient Detector (LIS/OTD) dataset. LIS/OTD found wide application in Earth system science, including meteorology and climatology e.g.,, wildfire science e.g.,, atmospheric chemistry e.g.,, and atmospheric physics e.g.,. However, LIS/OTD has several limitations that have made it worthwhile to develop an alternative, free global gridded lightning dataset.

First, LIS/OTD covers the period 1995–2000, with additional data for the tropics (between ±38∘) covering 1998–2010 . Since this release, no update of LIS/OTD has been made. Second, while the LIS/OTD climatology is available at the relatively high spatial resolution of 0.5∘, the time-transient data are only available a low spatial resolution (2.5∘) and have temporal gaps in coverage. Furthermore, the global LIS/OTD time series data are not available beyond 2000. While there is a new International Space Station Lightning Imaging Sensor (ISS-LIS) lightning stroke density and energy product, these data do not cover the high latitudes, nor can the sensor, being mounted on the orbiting space station, detect lightning across the entire globe simultaneously.

At the same time, several very-high-quality lightning datasets have been produced using ground-based sensor networks used for operational lightning monitoring, e.g., by meteorological services for near-real-time hazard warnings. While these datasets have been invaluable for regional studies, they are either not global e.g.,, not free , or both .

Because LIS/OTD is neither updated nor not available at sufficient resolution for many studies, and because other datasets are not free or not global, there is a demand for an open-access, continuously updated global lightning time series and climatology with monthly temporal and 0.5∘ or finer spatial resolution. Over the past decade, steps have been made to develop such a dataset based on the World Wide Lightning Location Network (WWLLN) network of very-low-frequency (VLF) radio sensors.

Based on the observation that lightning strokes emit characteristic VLF radio energy in the 1–24 kHz range, and that the location of strokes could be established through triangulation , the WWLLN was established in 2003 and produced its first set of global observations in August 2004. The WWLLN network has grown steadily over subsequent years and been improved with postprocessing to correct for timing and location inaccuracies, and provide estimates of relative detection efficiency and energy per stroke. Currently, the WWLLN has over 70 participating detector stations that monitor VLF radio waves in real time. The initial specification of the network was to provide global real-time locations of lightning discharges with more than 50 % flash detection efficiency and global mean location accuracy of 3.4 km .

Since its inception, the WWLLN has been used in more than 100 publications including local, regional, and global studies on atmospheric electricity e.g.,, and climate phenomena including precipitation and tropical cyclones e.g.,, and to develop regional climatologies e.g.,. A complete list of published studies using WWLLN is cataloged at http://wwlln.net/publications (last access: 5 July 2021). WWLLN has been extensively evaluated against independent observations of lightning occurrence and stroke energy at regional and global scale, including studies assessing the network's detection efficiency , precision in geolocation , and accuracy of the calculated stroke energy .

presented the first global lightning climatology based on WWLLN data covering the period 2005–2012 with 0.25∘ spatial and hourly temporal resolution. In this study, the gridded WWLLN climatology was compared with LIS/OTD and also showed the added value of observing the diurnal cycle of lightning by having a dataset with hourly resolution . In the intervening years, WWLLN has continued to collect data and increased the quality of the retrievals through a build-out of the sensor network. Given these improvements, continued interest in an open-access global gridded lightning dataset, and questions about the relationship between ongoing climate change and lightning occurrence, synthesis of the WWLLN data is due for an update.

Here, we present a new analysis of all of the WWLLN data collected to date, the development of a multi-resolution gridded climatology and time series, and evaluation of the resulting fields with independent observations from surface and spaceborne sensors. We demonstrate that in terms of total number of strokes detected the WWLLN network stabilized around 2014, but that with corrections for relative detection efficiency the dataset can be used back to 2010. We discuss the climatology of lightning over the second decade of the 21st century and interannual variability in lightning distribution and stroke power. The global gridded datasets resulting from this study form the WWLLN Global Lightning Climatology (WGLC). The WGLC is freely available for download at 0.5∘ and 5 arcmin spatial with daily and monthly temporal resolution and will be updated annually in the first quarter of every year.

While the earliest years of the WWLLN data show a strong increase in the number of lightning strokes detected over time, by 2014, total global lightning detected by the network stabilizes around approximately 210 million strokes yr-1. Using an adjustment for WWLLN's reported detection efficiency, the gridded lightning time series and climatology can arguably be extended back to 2010 or at least to 2012. The gridded version of the WWLLN data that we present here, i.e., the WGLC, thus covers the period 2010–2020 as a time series and as a climatological mean over that period. We choose to use as much of the data as possible in generating our climatology as there are places where lightning is rare, e.g., in the Arctic or over the oceans, that benefit from the extra years of observation when building the climatology, because even some positive lightning density, however small, is more realistic than zero values. Nevertheless, even with the adjustment for detection efficiency, the years 2010–2012 in the WGLC should be treated with caution, as they represent global and regional stroke counts that are lower than the mean over subsequent years, suggesting that the ongoing build-out of the sensor network continued to affect detection efficiency over this period. As more years of data are incorporated into the WGLC in the future, it may be preferable to exclude these early years from the climatological mean.

While the spatial pattern of lightning in the WGLC looks similar to other analyses of global lightning made with WWLLN and other sensors over the past decades, the temporal pattern of showed noteworthy variability over the second decade of the 21st century. The WGLC record is remarkable for a period of high lightning density in the tropics and Southern Hemisphere from 2012–2015 and increasing lightning in the midlatitudes of the Northern Hemisphere from 2018–2020. Furthermore, there is a noticeable decline in median stroke power after 2013, which reached a decadal minimum in late 2020. These changes in lightning occurrence may be related to interannual climate variability.

summarized several cyclical climate drivers that have been hypothesized to influence lightning occurrence, including the El Niño–Southern Oscillation (ENSO), the solar cycle, and the stratospheric Quasi-Biennial Oscillation (QBO) (see Figs. S11–S13 in the Supplement). Similar to their analysis that used LIS/OTD , we do not see any evidence of a global-scale relationship between WGLC lightning density and the multivariate ENSO index, total solar irradiance, or the QBO. On the other hand, it appears that stroke energy may be correlated with the solar cycle. Total solar irradiance reached a maximum in 2014–2015 and declined to a minimum in 2019, similar to, though not completely in phase with, the stroke power time series. Although there are plausible physical mechanisms in the solar flux that could influence atmospheric electricity , a longer time series of lightning energy observations would be necessary to confirm this relationship.

The global maps that form the WGLC are currently distributed at daily and monthly temporal resolution and 0.5∘ and 5 arcmin resolution. In principle, it would also be possible to distribute the WGLC on even finer-resolution grids, subject to the 3.4 km mean uncertainty in the WWLLN geolocation accuracy . As has been shown previously , it would also be possible to generate WGLC grids at higher temporal resolution, such as hourly. However, the resulting data files would become very large, and with daily to monthly resolution the current standard for most global gridded climate datasets e.g.,, the current version of the WGLC may be applied in a range of uses in the community in its current form.

A recurrent characteristic of the comparison of the WGLC with independent observations of lightning from ground-based and spaceborne sensors shows that WGLC detects substantially less lightning during peak seasons and episodes. It appears that, particularly in the early years of WWLLN, the network tended to be saturated at high lightning densities, a characteristic that was reported previously . This saturation of the sensor network means that, in some places and times, that the effective detection efficiency of WWLLN may be 40 % or less. On the other hand, WGLC appears to be better than other sensors at detecting lightning when lightning is rare, e.g., during cold seasons or in places with low overall density. Ground-based sensor networks such as ALDN and NLDN do not detect as much lightning as WGLC in coastal areas or in areas far away from the sensors. This means that the VLF technology behind WWLLN may be appropriate for producing an overall, globally consistent picture of lightning that is not influenced by sensor proximity, but that periods of intense lightning will be underestimated by the network.

Furthermore, our analysis of the WGLC in comparison with the ALDN shows that while WGLC detects as few as 85 % fewer lightning strokes during the peak season in Alaska, those strokes it does detect tend to be the more powerful ones. WGLC is therefore “missing” mainly the weaker lightning strokes. showed that, through comparison with New Zealand Lightning Detection Network data, the detection efficiency of WWLLN for powerful lightning strokes (strokes >50 kA peak current) increases to about 80 %. Global analysis of WWLLN data concluded that the detection efficiency of the network is 60 %–80 % for high-amplitude strokes . Thus, while not capturing all strokes, the WGLC may still be useful in understanding when, where, and how much of the most hazardous lightning occurs, i.e., that which may be more likely to ignite wildfires or damage infrastructure, for example.

As the most widely used global gridded lightning dataset among Earth system scientists, it is worth comparing the LIS/OTD time series and climatology with WGLC. Because LIS/OTD is based on optical sensors on satellites in low Earth orbit to detect lightning flashes, compared with WGLC's network of ground-based VLF receivers, we expect the dataset to be different from WGLC. Our analysis shows that WGLC captures on average 10 % lightning strokes recorded by LIS/OTD on land, with notable differences in the Congo Basin and northwestern Himalaya . Over the oceans, the comparison is more favorable, with WGLC capturing an average of 33 % of the flashes in LIS/OTD. In some oceanic areas, WGLC has a greater annual mean lightning density than LIS/OTD (Fig. ).

In an earlier comparison between LIS/OTD and WWLLN, showed that lightning flashes detected by LIS/OTD could be composed of multiple lightning strokes. They found that, on average, WWLLN captured 1.5 strokes for each LIS/OTD flash, although 71.5 % of the WWLLN-matched LIS/OTD flashes were from a single lightning stroke. Because WGLC nearly always has lower lightning density than LIS/OTD, and because a single LIS/OTD flash may be comprised of multiple lightning strokes, the detection efficiency of WWLLN may be even lower than might be assumed based on a simple comparison of the two datasets. Our simple comparison shows that WGLC is about 3 times more likely to detect LIS/OTD flashes over ocean (33 %) than over land (10 %). Accounting for the mean multiple-stroke-per-flash discrepancy of 1.5 implies that WGLC's implied detection efficiency using LIS/OTD as a standard would be 22 % over ocean and 6.6 % on land, which is in line with the 17.3 % over ocean and 6.4 % over land reported previously .

Based on our comparisons of WGLC with ALDN and NLDN, we ascribe the differences in lightning density between WGLC and LIS/OTD to (1) the known behavior for WWLLN to saturate at high lightning densities , (2) LIS/OTD capturing more weak cloud-to-cloud lightning activity , particularly near cloud tops, that would not necessarily be detected by WWLLN, as shown in our analysis for Alaska, and (3) LIS/OTD being subject to a number of postprocessing steps and adjustments for detection efficiency that add uncertainty to the final data product.

Nevertheless, from all of the comparisons between WGLC and other datasets, it is clear that WGLC has less lightning than independent observations. However, this does not mean that the WGLC is not useful. Other datasets either cover only a limited period, have limited spatial coverage or resolution, are not free, or all of the above. WGLC, in contrast, is based on a single methodology from a global sensor network (WWLLN) that has continuously observed lightning since 2005. The WGLC will be updated annually, which makes it valuable for understanding changes in lightning over time and how climate change is affecting lightning frequency and distribution. Furthermore, WGLC is the only free source that can provide data on lightning at any nearly any spatial and temporal resolution, limited only by the properties of the sensor network, i.e., milliseconds in time and approximately 3.4 km in space. Finally, WGLC provides stroke energy estimates as well as lightning location. The WGLC may therefore be a valuable tool for a range of research applications.

Among the applications for which the WGLC may be suitable, a few that stand out include modeling global wildfire and atmospheric chemistry, and risk and hazard assessment, particularly in regions of the world where no other lightning detection networks exist. Because the WGLC is a homogeneous global dataset, it is possible to directly compare lightning observations between locations without the inter-network calibration that would be required to analyze data from different regional networks perhaps using different technologies. WGLC will be particularly useful for understanding the patterns of natural, i.e., non-human-caused, wildfire ignitions in remote locations such as boreal and tropical forests and in areas of the developing world currently undergoing rapid land use change. WGLC may also be useful in parameterizing atmospheric chemistry models to estimate lightning NOx production e.g., and to better understand the relationship between extraterrestrial radiation and atmospheric electricity e.g.,.

WGLC may be a useful tool for assessing lightning hazards to persons and property e.g.,. While the WGLC geolocation accuracy is probably too low to identify individual buildings or other types of infrastructure, WGLC may be used in a probabilistic way to understand seasonal, diurnal, and climatological lightning patterns. This capability of the WGLC will be particularly useful in regions of the world that are not currently served by high-sensitivity regional lightning detection networks, including much of the developing world and oceanic areas. For example, WGLC's generally high detection efficiency over the oceans may make it particularly useful for assessing lightning risks to shipping in open-ocean regions. Finally, as a completely free, open-access dataset, the WGLC can serve researchers, governments, NGOs, and communities that do not have resources to purchase commercial lightning data.

The WGLC gridded lightning density and power fields are distributed as a time series at daily and monthly resolution and as a climatological monthly mean. The data are stored in NetCDF format and are archived with Zenodo 10.5281/zenodo.4774528.

The code used for gridding the raw stroke counts and progressive updates to the gridded data are archived at https://github.com/ARVE-Research/WGLC .

Since its inception in 2005, the WWLLN has grown to the point where the network is capable of producing a globally consistent, spatially resolved picture of lightning activity over land and the oceans. These raw data, gridded, adjusted for detection efficiency, and continuously updated over time, form the WGLC . With more than a decade of reliable data on stroke density and power in the dataset, we can now start to investigate changes in the spatiotemporal pattern of lightning. Lightning strokes appear to show important variability on interannual timescales, and while these patterns may be attributable to interannual climate variability, they require further study to clearly identify the drivers. The WGLC is distributed for free online in two standard spatial resolutions (0.5∘ and 5 arcmin) and with daily and monthly temporal resolution. Other versions of the WGLC could be prepared in response to community demand. While the WGLC does not capture every lightning stroke and appears to underestimate stroke density at peak periods, it represents a unique, open-access global gridded lightning climatology and time series that may be a valuable tool for researchers in a range of fields including wildfire ignitions, atmospheric chemistry, and assessment of lightning risks to humans, animals, and infrastructure.