Layout 1 INTRODUCTION The Caspian Sea (CS) is located in Western Asia and Eastern Europe. With a surface area of approximately 370’000 km2, the CS is considered as the world’s largest enclosed water body (Lattuada et al., 2019). Due to its lat- itudinal extension, the CS crosses multiple climatic zones and encounters a strong north to south gradient in geologi- cal, physical, and chemical properties (Barale, 2010). Con- ventionally, the CS has been divided into three major basins according to water depth and geographical location, namely the Northern, Middle, and Southern CS (Fairbridge, 1966; Ibrayev et al., 2010). The Northern Caspian Sea (NCS) is very shallow (with an average depth of less than 5 m and maximum depth of 20 m; Stolberg et al., 2006) and is lo- cated on the extreme southern boundary of sea ice cover de- velopment in the Northern Hemisphere (Kouraev et al., 2004). Every year, its surface area freezes during winter, while most of the middle and southern CS stays ice free year-round (Barale, 2010). Thickness of sea ice may reach up to 75 cm in the western and central NCS and 120 cm in the eastern parts of this basin. The cumulative impacts of climate conditions, wind fields, water currents, and sea mor- phology lead to a significant temporal and spatial variability in the sea ice cover over the CS (Kouraev et al., 2004). To understand spatial and temporal variations in the ex- tent and timing of sea ice in the CS, here, 15 years of the sea ice presence and its phenology (freeze-up, break-up, and duration) in the NCS for a period from 2004 to 2018 were studied. Sea ice data were provided by the Centre for Ice Hydrometeorological Information at the Arctic and Antarc- tic Research Institute in St. Petersburg (AARI), the Hy- drometeorological Centre in Moscow (Hydrometcentre), and local hydrometeorological offices in the Arctic, Far- Eastern Russia, Baltic, Black and Caspian seas, all belong- ing to the Russian Federal Hydrometeorological Service (Roshydromet). The Roshydromet coastal weather polar stations provide visual and instrumental observations on dif- ferent features of sea ice such as sea ice concentration and thickness. The centre also employs visible and infrared satellite images from American (NOAA HRPT, EOS TERRA) and Russian (METEOR, OKEAN) satellites for the sea ice monitoring. All the data are processed and uti- lized for sea ice analysis at the AARI station. The AARI sta- tion is compiling weekly sea ice conditions charts for the CS since winter 2002 (WMO, 2006). METHODS Caspian Sea ice data are provided by AARI in the shape of lines and polygons data (http://wdc.aari.ru). Weekly bi- nary raster maps of sea ice presence and absence in 0.1×0.1 degree resolution were created in ArcMap environment and were written into Netcdf format for further analyses in Mat- lab. Several algorithms were developed in Matlab environ- ment to extract different features of the sea ice in the NCS for the period of study (2004-2018). Percentage of sea ice cover over the NCS in each year was computed as the total number of ice-covered pixels in each year divided by the total number of pixels in the NCS. Algorithms were devel- oped for retrieving sea ice phenology metrics, including the time of sea ice freeze-up and break-up, as well as duration of ice-covered sea for each pixel in the NCS for a period from 2004 to 2018. The time of ice formation (week) was set as the time when the sea ice appeared for the first time in a year in a pixel that was ice free in its previous two weeks. This value was negative for the years when the ice formed in the autumn of the previous year. The time of ice break up (week) was set as the time when ice was observed for the last time in a year in a given pixel, while the pixel remained ice-free for at least two subsequent weeks. The duration of ice coverage in a pixel (weeks) was calculated as the difference between the time of ice in and ice out. The existence of a linear trend in the sea ice phenology (start, ARTICLE Sea ice phenology in the Caspian Sea Forough Fendereski* School of Environment and Sustainability, University of Saskatchewan, Saskatoon, Canada ABSTRACT To understand the spatial and temporal variations in the extent and timing of sea ice in the Caspian Sea (CS), 15 years of sea ice presence and its phenology (freeze-up, break-up, and duration) were studied in the Northern CS (NCS) for a period from 2004 to 2018. This study indicated that the percentage of sea ice covered area in the CS showed fluctuations ranging from 54.2 % in 2004 to more than 89 % in 2006 and 2008. This study, furthermore, found regional differences within the NCS in sea ice pheno- logical features during 2004 to 2018. Specifically, earlier sea ice formation, later ice melt, and longer ice periods were observed in the eastern than the western NCS. Noticeable interannual changes were also observed in the timing and duration of the NCS ice (2004-2018). However, most parts of the NCS did not show significant trends (P>0.05) in the formation, decay, and duration of sea ice and their interannual variability. The observed spatial and temporal patterns in the CS ice can have implications for phyto- plankton blooms and higher trophic levels, such as fish and endangered Caspian seal populations and merits further investigation. Non -co mmerc ial us e o nly Sea ice phenology in the Caspian Sea 25 end, duration) for each pixel during the period of study (2004 to 2018) was analysed using non-parametric Mann- Kendall test (P<0.05). The magnitude of inter-annual vari- ability (iv) in the phenological characteristics of the sea ice (i.e., timing of the ice formation and break-up, and the du- ration of the sea ice coverage) for each individual grid point in the NCS was computed based on the standardized anom- alies in the ice phenology metrics of each pixel in each year. To this aim, the differences between all the grid points for each phenological feature in each year and their climato- logical averages (2004-2018) were computed and divided by the standard deviation (2004-2018; Soppa et al., 2016). Average climatologies of anomalies of the phenological fea- tures over the study period (2004-2018) were calculated and the magnitude of their interannual variability were com- puted as absolute temporal mean of anomalies (2004-2018). RESULTS AND DISCUSSION The percentage of sea ice covered area in the CS from 2004 to 2018 ranged between 54.24 (2004) to 89.79 to 89.97% (2008 and 2006, respectively; Fig. 1). The observed oscillation in the sea ice coverage can be explained by changes in the severity of different winters during the period of this study (Kouraev et al., 2004). The dependence of the ice regime to the air temperature was also shown in other studies (e.g., Baklagin, 2019). A gradual west/southwards gradience was observed in the time of the sea ice formation in the NCS. The shallow eastern parts of the NCS froze ear- lier than other parts of the NCS (as early as the second half of December on average (week -2); Fig. 2b). The sea ice later spread out to the west (because of the Volga runoff im- pact; Kouraev et al., 2004), and the latest ice developed in the southern parts of the NCS (in late February and early March; Fig. 2b), where larger depths and increase in the water exchange with the Middle CS limited further devel- opment of sea ice cover in this area (Kouraev et al., 2004). CS ice (Fig. 2c) decayed earlier in the southern parts of the NCS (from as early as the first half of January; week 2), then in the western parts of this area near the Volga delta, and finally, last CS ice disappeared in the eastern parts of the NCS (as late as April; week 12), where the thickest ice is observed (average of 40-50 cm; Kouraev et al., 2004). Expectedly, longest periods of ice coverage are in the east- ern NCS (8-18 weeks) and as we go further west and south, sea ice lasted for less time (Fig. 2c). Spatial differences in the sea ice timing have been also reported in other parts of the World Ocean, e.g., in the Arctic Ocean (Ji et al., 2013). A previous study in the CS also demonstrates similar spatial patterns in the sea ice timing in the region and stated a cor- respondence between the observed timing of sea ice forma- tion in different regions of the NCS with historical observations of sea ice (Kouraev et al., 2004). An explana- tion for the delay in the sea ice break-up in the eastern part of the NCS can be the weak circulation in the north-eastern part of the CS after melting the sea ice that causes the last Fig. 1. Percentage of sea ice coverage in the north Caspian Sea from 2004 to 2018. Non -co mmerc ial us e o nly F. Fendereski26 cold-water patch in that region to stay undistributed (Fend- ereski et al., 2014). Spatial differences in the sea ice timing can partly explain spatial changes in the phytoplankton bloom features in the region. The later increase in water temperature in the eastern NCS compared to that in the western parts of this basin can be one reason for the delay in the occurrence of the phytoplankton bloom in the eastern parts of the NCS (Fendereski et al., 2014). The formation, break-up, and duration of sea ice in the CS changed from year to year during the study period. Most parts of the NCS did not show significant trends in the tim- ing and duration of their sea ice (P>0.05; Fig. 3). Only 12.5% of the NCS showed a significant trend in the timing of the ice freeze-up from 2004 to 2018 towards earlier ice formation dates (P<0.05; Fig. 3a). 10.9% of the NCS shows significant trend in the length of sea ice period and towards longer ice cover periods (P<0.05; Fig. 3b). No significant trend was observed in the time of ice decay in the NCS (P>0.05). For the interannual changes in the sea ice phenol- ogy in the NCS, generally, timing of the ice formation and the duration of ice coverage showed a higher degree of in- terannual variability from 2004 to 2018 than the timing of ice disappearance (Fig. 4). More than 90% of the NCS showed more than half a week of interannual variability in the timing of the beginning/end and duration of their ice season (Fig. 4). Similar to the sea ice phenology metrics, no pronounced spatial gradient in the interannual changes of ice phenology characteristics were observed during the study period (2004 to 2018; Fig. 4). A previous study on the CS ice for a period from 1998 to 2002 showed changes in the dates of ice start and break-up and gradual reduction in the ice cover (Kouraev et al., 2004), attributed to changes in the ther- mal regime towards mild winters in this region; how- ever, this study (Kouraev et al., 2004) left the open question whether the observed changes are indicative of Fig. 2. Climatological mean of the timing of the (a) sea ice for- mation (week; negative values indicate weeks falling into the previous year of a given year), (b) sea ice melt (week), and (c) duration of the sea ice coverage (weeks) in the Caspian Sea from 2004 to 2018. Fig. 3. Trend in the (a) sea ice formation, and (b) duration of the sea ice coverage in the Caspian Sea during 2004 to 2018 (MannK- endall test; P<0.05). White pixels indicate no significant trend in the sea ice phenology in the Caspian Sea during the period of study (2004 to 2018; P>0.05). No significant trend was observed in the timing of the sea ice break-up in the CS (P>0.05). Non -co mmerc ial us e o nly Sea ice phenology in the Caspian Sea 27 a long-term warming trend or just a series of warm win- ters and suggested that to answer this question a longer time series is needed. Covering the period from 2004 to 2018, the current study did not show similar patterns as observed between 1998 and 2002. In contrast, rather than a decreasing trend, an interannual fluctuations were observed in the sea ice extent with no significant trends in the sea ice formation, break-up, and duration in most parts of the NCS (P>0.05). The highly productive shallow fresh waters of the NCS, in particular the eastern parts of this basin, provide an essential breeding, spawning, nursery, foraging, and transit grounds for commercial fish species such as en- demic sturgeons (Acipenser sp.; Karpinsky et al., 2005; Zonn, 2005) and other unique and vulnerable CS species such as the Caspian seal (Phoca caspica; Dmitrieva et al., 2015) that is listed on the IUCN Red List of Threatened Species. Given the ecological importance of the NCB and considering the importance of the sea ice timing on phy- toplankton bloom phenology (Ji et al., 2013) and thereby the whole food web (match-mismatch hypothesis; Cush- ing, 1990; Platt et al., 2003; Hague and Vichi, 2018; Tedesco et al., 2019), the observed changes in the NCS ice may have consequences for the CS ecosystem function and ecosystem service provision. Results of the current study suggested the necessity of a comprehensive study on the impact of changes in the CS ice, in particular in the context of global warming, on primary producers’ phe- nology and their implications for higher trophic levels’ re- cruitment success and population dynamics in this area. CONCLUSIONS In this paper, 15 years (2004 to 2018) of spatial and tem- poral variability in the sea ice presence and phenology (freeze-up, break-up, and duration) in the North Caspian Sea (NCS) were studied. Results showed year-to-year os- cillations in the percentage of sea ice cover and its pheno- logical metrics (timing of start, break-up, and duration of sea ice) in the NCS in the period of study. The study also showed spatial patterns in the sea ice phenology in the NCS, with later formation, earlier break-up, and shorter ice in the western parts of the NCS. ACKNOWLEDGMENTS The author is grateful to the Centre for Ice Hydrome- teorological Information at the Arctic and Antarctic Re- search Institute in St. Petersburg (AARI), the Hydrometeorological Centre in Moscow (Hydrometcen- tre), and the Russian Federal Hydrometeorological Serv- ice (Roshydromet) local office in the Caspian Sea for providing the sea ice data. Corresponding author: forough.fendereski@gmail.com Key words: Sea ice; phenology; spatial variation; temporal variation. Received: 24 February 2021. Accepted: 24 August 2021. This work is licensed under a Creative Commons Attribution Non- Commercial 4.0 License (CC BY-NC 4.0). ©Copyright: the Author(s), 2021 Licensee PAGEPress, Italy Advances in Oceanography and Limnology, 2021; 12:9704 DOI: 10.4081/aiol.2021.9704 Fig. 4. Climatological mean of the standardized anomaly of the magnitude of interannual variability (iv) in the (a) sea ice for- mation (ice form; weeks), (b) sea ice melt (weeks), and (c) du- ration of the sea ice coverage (weeks) in the Caspian Sea during 2004 to 2018. Non -co mmerc ial us e o nly F. Fendereski28 REFERENCES Baklagin VN, 2019. Variations of indicative dates of ice regime on Lake Onego based on ground air temperature. Adv Ocean Limnol 10:24-32. Barale V, 2010. The European and marginal seas: An overview, p. 3-22. In: V. Barale, M. Gade (eds.), Remote Sensing of the European Sea. Springer. Cushing DH, 1990. Plankton production and year-class strength in fish populations – an update of the match mismatch hy- pothesis. Adv Mar Biol 26:249-293. Dmitrieva L, Härkönen T, Baimukanov M, Bignert A, Jüssi I, Jüssi M, Goodman S, 2015. Inter-year variation in pup pro- duction of Caspian seals Pusa caspica 2005–2012 determined from aerial surveys. Endanger Species Res 28:209-223. Fairbridge RW, 1966. Encyclopedia of oceanography. Reinhold Publishing Co., New York: 1021 pp. Fendereski F, Vogt M, Payne MR, Lachkar Z, Gruber N, Salmanmahiny A, Hosseini SA, 2014. Biogeographic clas- sification of the Caspian Sea. Biogeosciences 11:6451-6470. Hague M, Vichi M, 2018. A link between CMIP5 phytoplankton phenology and sea ice in the Atlantic Southern Ocean. Geo- phys Res 45:6566-6575. Ibrayev RA, Ozsoy E, Schrum C, Sur HI, 2010. Seasonal vari- ability of the Caspian Sea three-dimensional circulation, sea level and air-sea interaction. Ocean Sci 6:311-329. Ji R, Jin M, Varpe Ø, 2013. Sea ice phenology and timing of pri- mary production pulses in the Arctic Ocean. Glo. Change Biol 19:734-741. Karpinsky MG, Mikhail G, Damir NK, Goryunova VB, Shiganova T, Katunin Vera N, 2005. Biological features and resources, p. 191-210. In: A.N. Kosarev and A G. Kostianoy (eds.), The Caspian Sea environment: The Handbook of en- vironmental chemistry. Springer. Kouraev AV, Papa F, Mognard NM, Buharizin PI, Cazenave A, Cretaux JF, Dozortseva J, Remy F, 2004. Sea ice cover in the Caspian and Aral Seas from historical and satellite data. J Marine Syst 47:89-100. Lattuada M, Christian A, Wilke T, 2019. Differential impact of anthropogenic pressures on Caspian Sea ecoregions. Mar Pollut Bull 142:274-281. Platt T, Fuentes-Yaco C, Frank KT, 2003. Spring algal bloom and larval fish survival. Nature 423:398-399. Soppa MA, Völker C, Bracher A, 2016. Diatom phenology in the Southern Ocean: mean patterns, trends and the role of climate oscillations. Remote Sens 8:420-437. Stolberg F, Borysova O, Mitrofanov I, Barannik V, Eghtesadi P, 2006. Caspian Sea, GIWA Regional assessment 23, UNEP report. University of Kalmar, Sweden. Tedesco L, Vichi M, Scoccimarro E, 2019. Sea-ice algal phe- nology in a warmer Arctic. Sci Adv 5:eaav4830. World Meteorological Organization (WMO), 2006. Sea Ice In- formation Services in the World, Report. WMO, Geneva. Zonn IS, 2005. Economic and international legal dimensions, p. 243-256. In: A.N. Kosarev and A G. Kostianoy (eds.), The Caspian Sea environment: The Handbook of environmental chemistry. Springer. Non -co mmerc ial us e o nly