In recent years, artificial intelligence, particularly deep learning, has garnered significant attention among practitioners and scholars in meteorology and atmospheric sciences, leading to a substantial body of literature. This study aims to delineate the present research status and trends in climate innovation through CiteSpace visual analysis. To comprehend the current landscape, prevalent terms, and research frontiers of deep learning for climate change research (DLCCR) within meteorology and atmospheric applications, we gathered 256 published papers spanning from 2018 to 2022 from the Web of Science (WOS) core database. Employing these articles, we conducted co-authorship, co-citation, and keyword co-occurrence analyses. The findings unveiled a steady rise in DLCCR publications over the last five years. However, the correlation between high yield and high-citation authorship appears inconsistent and weak. Notably, prolific authors in this domain included Zhang Z.L. and Bonnet P. Furthermore, leading institutions such as the Chinese Academy of Sciences (China), le Centre National de la Recherche Scientifique (France), and Nanjing University of Information Science and Technology (China) have played pivotal roles in advancing DLCCR. The primary contributors among high-yield countries primarily cluster in a select group comprising China, the USA, South Korea, and Germany. Identifying significant information gaps in numerical weather, atmospheric physics and processes, algorithm parametrizations, and extreme events, our study underscores the necessity for future researchers to focus on these and related subjects. This study provides valuable insights into research hotspots, developmental trajectories, and emerging frontiers, thereby delineating the knowledge structure in this field and highlighting directions for further climate innovation research.

The study combined the second-order polynomial and some elements of algebra and trigonometry with the ogive to objectively (mathematically) locate the rainfall onset, peak, and retreat on the ogive, which was visually done. The data used for the investigation are the daily rainfalls of 16 synoptic stations (Ikeja, Calabar, Port-Harcourt, Benin, Ondo, Enugu, Ilorin, Lokoja, Jos, Kaduna, Yola, Kano, Sokoto, Maiduguri, Potiskum, and Nguru) across all the ecological zones of Nigeria. The datasets spanning 50 years (1971–2020) were collected from the Archives of the Nigerian Meteorological Services, Abuja, Nigeria. The ogives were derived from the frequency of rainy days. The peak periods were best detected from the pentad graphs of the rainy-day frequency. The results showed that rainy-day frequency is better than rainfall amount in determining the various rainfall parameters over Nigeria. The second-order polynomial modeled the two curvatures of the rainfall perfectly. The rainy-day frequency in the southern part of the country exhibited double rainfall maxima, while those in the northern part showed a single rainfall maximum. The double rainfall maxima are not peculiar to the southwestern region of Nigeria, as previously widely asserted; they cut across the whole southern region, although the number of days and the depth of the trough between the peaks dimmish from the largest values in the extreme southwest to the least in the extreme southeast. The first rainfall peaks were attained in southern Nigeria in July, except at Ikeja, which was in June. The second rainfall peak was reached in September in all the southern stations. The single peaks were attained in all the extreme northern stations in August and in the other stations south of them in early September. The rainfall onset begins at Ikeja in the extreme southwestern corner of Nigeria around March ending. It spreads eastwards and northwards to cover the entire country by mid-June, reaching Nguru in the northeastern corner. It is generally earlier on the western flank than the eastern flank. Rainfall begins to retreat from the northernmost stations by the third week in September to reach the extreme southern stations between October and early November.