Title: Climate change and deepening of the North Sea fish assemblage: a biotic indicator of warming seas
Abstract: Journal of Applied EcologyVolume 45, Issue 4 p. 1029-1039 Free Access Climate change and deepening of the North Sea fish assemblage: a biotic indicator of warming seas Nicholas K. Dulvy, Corresponding Author Nicholas K. Dulvy Centre for Environment, Fisheries and Aquaculture Science, Lowestoft Laboratory, Lowestoft NR33 0HT, UK; Simon Fraser University, Department of Biological Sciences, Burnaby, BC, Canada V5A 1S6; and *Correspondence author. E-mail: [email protected]Search for more papers by this authorStuart I. Rogers, Stuart I. Rogers Centre for Environment, Fisheries and Aquaculture Science, Lowestoft Laboratory, Lowestoft NR33 0HT, UK;Search for more papers by this authorSimon Jennings, Simon Jennings Centre for Environment, Fisheries and Aquaculture Science, Lowestoft Laboratory, Lowestoft NR33 0HT, UK;Search for more papers by this authorVanessa Stelzenmüller, Vanessa Stelzenmüller Centre for Environment, Fisheries and Aquaculture Science, Lowestoft Laboratory, Lowestoft NR33 0HT, UK;Search for more papers by this authorStephen R. Dye, Stephen R. Dye Centre for Environment, Fisheries and Aquaculture Science, Lowestoft Laboratory, Lowestoft NR33 0HT, UK;Search for more papers by this authorHein R. Skjoldal, Hein R. Skjoldal Institute of Marine Research, Box 1870 Nordnes, N-5817 Bergen, NorwaySearch for more papers by this author Nicholas K. Dulvy, Corresponding Author Nicholas K. Dulvy Centre for Environment, Fisheries and Aquaculture Science, Lowestoft Laboratory, Lowestoft NR33 0HT, UK; Simon Fraser University, Department of Biological Sciences, Burnaby, BC, Canada V5A 1S6; and *Correspondence author. E-mail: [email protected]Search for more papers by this authorStuart I. Rogers, Stuart I. Rogers Centre for Environment, Fisheries and Aquaculture Science, Lowestoft Laboratory, Lowestoft NR33 0HT, UK;Search for more papers by this authorSimon Jennings, Simon Jennings Centre for Environment, Fisheries and Aquaculture Science, Lowestoft Laboratory, Lowestoft NR33 0HT, UK;Search for more papers by this authorVanessa Stelzenmüller, Vanessa Stelzenmüller Centre for Environment, Fisheries and Aquaculture Science, Lowestoft Laboratory, Lowestoft NR33 0HT, UK;Search for more papers by this authorStephen R. Dye, Stephen R. Dye Centre for Environment, Fisheries and Aquaculture Science, Lowestoft Laboratory, Lowestoft NR33 0HT, UK;Search for more papers by this authorHein R. Skjoldal, Hein R. Skjoldal Institute of Marine Research, Box 1870 Nordnes, N-5817 Bergen, NorwaySearch for more papers by this author First published: 09 July 2008 https://doi.org/10.1111/j.1365-2664.2008.01488.xCitations: 516AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Summary 1 Climate change impacts have been observed on individual species and species subsets; however, it remains to be seen whether there are systematic, coherent assemblage-wide responses to climate change that could be used as a representative indicator of changing biological state. 2 European shelf seas are warming faster than the adjacent land masses and faster than the global average. We explore the year-by-year distributional response of North Sea bottom-dwelling (demersal) fishes to temperature change over the 25 years from 1980 to 2004. The centres of latitudinal and depth distributions of 28 fishes were estimated from species-abundance–location data collected on an annual fish monitoring survey. 3 Individual species responses were aggregated into 19 assemblages reflecting physiology (thermal preference and range), ecology (body size and abundance-occupancy patterns), biogeography (northern, southern and presence of range boundaries), and susceptibility to human impact (fishery target, bycatch and non-target species). 4 North Sea winter bottom temperature has increased by 1·6 °C over 25 years, with a 1 °C increase in 1988–1989 alone. During this period, the whole demersal fish assemblage deepened by ~3·6 m decade−1 and the deepening was coherent for most assemblages. 5 The latitudinal response to warming was heterogeneous, and reflects (i) a northward shift in the mean latitude of abundant, widespread thermal specialists, and (ii) the southward shift of relatively small, abundant southerly species with limited occupancy and a northern range boundary in the North Sea. 6 Synthesis and applications. The deepening of North Sea bottom-dwelling fishes in response to climate change is the marine analogue of the upward movement of terrestrial species to higher altitudes. The assemblage-level depth responses, and both latitudinal responses, covary with temperature and environmental variability in a manner diagnostic of a climate change impact. The deepening of the demersal fish assemblage in response to temperature could be used as a biotic indicator of the effects of climate change in the North Sea and other semi-enclosed seas. Introduction Climate change affects demography, geographic distribution and phenology of populations and species. Demographic effects are manifest as changes in recruitment, growth and survival (O’Brien et al. 2000; Pörtner & Knust 2007), distributional shifts as movements towards the poles or higher altitudes (Walther et al. 2002; Parmesan & Yohe 2003), and phenological effects as advances in the timing of spring-related events by > 2·3 days decade−1, with earlier flowering, egg-laying, plankton blooms and fish migrations creating potential for mismatching between and predator and prey populations (Crick & Sparks 1999; Sims et al. 2001; Parmesan & Yohe 2003; Edwards & Richardson 2004). Climate change-induced habitat loss and changing species distributions are predicted to result in species extinctions on land and population extinctions in the sea (Thomas et al. 2004; Drinkwater 2005). There is an increasing need to summarize the ecological complexity of climate impacts using biological indicators to inform managers, policymakers and society (EEA 2004; MCCIP 2006). Climate variability and longer-term change (hereafter called climate change) have led to marked changes in North East Atlantic conditions over the last century (Cushing 1982; Stenseth et al. 2005). Sea surface temperatures of North Atlantic and UK coastal waters have warmed by 0·2–0·6 °C decade−1 over the past 30 years. These seas are warming faster than the adjacent land and faster than the global average (MacKenzie & Schiedek 2007). Within the North East Atlantic region, warming was fastest in the English Channel, North Sea and Baltic Sea (ICES 2006a; Joyce 2006; Marsh & Kent 2006; Sherman et al. 2007). Some marked changes in North Sea fish distributions have been attributed to climate change: two-thirds of North Sea fishes have shifted mean latitude or depth. Fishes with a northern distributional boundary in the North Sea have shifted northwards and southern boundary species have retracted northwards at rates up to three times faster than terrestrial species (Perry et al. 2005). Exotic fishes with southerly biogeographic affinities are becoming established in the North Sea, including; anchovy Engraulis encrasicolus L., red mullet Mullus surmuletus L., sardine Sardina pilchardus, Walbaum 1792, John Dory Zeus faber, L. and snake pipefish Entelurus aequoreus, L. (Beare et al. 2004; ICES 2006b; Kirby, Johns & Lindley 2006; Enghoff, MacKenzie & Nielsen 2007). A key question is whether the individual responses of species are context-specific phenomena or whether they are symptomatic of a more systematic change in the North Sea ecosystem resulting from climate change. If such an ecosystem-scale change can be detected, this could underpin the development of a biotic indicator of climate change impacts. There is a wide range of desirable indicator properties, including specificity to a single pressure, sensitivity or strength of response, the lag in response and the spatial and taxonomic representativeness of the indicator (Rice & Rochet 2005). Here we summarize the effects of climate change on the demersal fish assemblage and develop an indicator that is taxonomically representative of a wide range of fish species. We search for an assemblage-wide biotic indicator of climate change in the North Sea ecosystem by comparing the distribution changes of fish species and assemblages to temperature and climate change over the past 25 years. For each year, we calculated the distance moved north or south and the deepening and shallowing of each fish species or assemblage relative to the long-term average. Species distributional responses were aggregated into non-mutually exclusive assemblages reflecting differences in physiology, ecology, biogeographic origin and human impact. We demonstrate a coherent deepening of fish species in response to climate change and two distinct latitudinal responses to climate change: a northward shift in mean latitude and southward extension of minimum latitude. Methods We used the North Sea English groundfish survey data to assess changes in the geographic distribution of 28 demersal fish species. The English Groundfish Survey (EGFS) samples a grid of trawl stations typically covering up to 84 statistical rectangles (between 51·75 to 61·75° N latitude) and has been fished annually throughout the North Sea as part of the International Council for the Exploration of the Sea (ICES) international bottom trawl survey in autumn (August–October). All fishes caught were identified and measured. Catch rates were raised to number of individuals caught per 60-min tow (for more details see Maxwell & Jennings 2005). Species were included if they were reliably identified throughout the time period and effectively sampled by the net (Sparholt 1990; Knijn et al. 1993; Maxwell & Jennings 2005; Dulvy et al. 2006). Pelagic fish were excluded because of the likelihood that they were captured in the water column during net shooting or hauling. The 28 species retained for analysis were representative of the breadth of morphology, life histories, ecology and taxonomic diversity of the bottom-dwelling fishes sampled by the survey (Table 1) and represent most of the numerical abundance and biomass of the demersal fish assemblage (Jennings et al. 2002). The Latin names for all study species are presented in Table 1; hereafter, only common names will be used. Table 1. Demersal North Sea fish species surveyed by the English Groundfish Survey, body size (cm), biogeographic affinity and thermal characteristics (°C), exploitation status, categorical numerical abundance and spatial occupancy and presence of a northern or southern range boundary. Thermal classification: W, warm thermal preference; C, cold thermal preference; g, generalist with broader thermal range; s, specialist with narrow thermal range. Numerical abundance: LA, less abundant; A, abundant. Spatial occupancy: LC, less common; W, widespread Common name Latin binomial Body size Biogeographic affinity Mean temperature Temperature range Thermal classification Exploitation status Abundance category Spatial occupancy category Range boundary Pogge Agonus cataphractus L. 20 Boreal 15·4 3·9 Cs Non-target LA LC N Wolffish Anarhichas lupus L. 125 Boreal 13·4 2·4 Cs Bycatch LA W S Scaldfish Arnoglossus laterna (Walbaum, 1792) 25 Lusitanian 16·4 3·2 Ws Non-target LA LC N Solenette Buglossidium luteum (Risso, 1810) 13 Lusitanian 16·4 4·5 Wg Non-target A LC N Grey gurnard Eutrigla gurnardus L. 45 Lusitanian 17·0 3·6 Ws Bycatch A W – Cod Gadus morhua L. 132 Boreal 13·8 2·3 Cs Target A W – Witch Glyptocephalus cynoglossus L. 60 Boreal 13·3 3·3 Cs Bycatch LA W S Long rough dab Hippoglossoides platessoides (Fabricius, 1780) 30 Boreal 13·8 2·4 Cs Bycatch A W – Megrim Lepidorhombus whiffiagonis (Walbaum, 1792) 61 Lusitanian 13·3 2·7 Cs Bycatch LA LC S Dab Limanda limanda L. 42 Boreal 17·0 4·5 Wg Bycatch A W S Angler Lophius piscatorius L. 75 Lusitanian 13·4 2·2 Cs Target LA W S Haddock Melanogrammus aeglefinus L. 76 Boreal 13·7 2·4 Cs Target A W – Whiting Merlangius merlangus L. 45 Lusitanian 13·2 3·2 Cs Target A W S Hake Merluccius merluccius L. 110 Lusitanian 13·8 5·8 Cg Target LA W S Lemon sole Microstomus kitt (Walbaum, 1792) 60 Boreal 15·2 2·3 Cs Target A W S Ling Molva molva L. 200 Boreal 13·1 2·1 Cs Target LA LC – Plaice Pleuronectes platessa L. 95 Lusitanian 17·0 4·4 Wg Target A W – Saithe Pollachius virens L. 130 Boreal 13·4 2·7 Cs Target A W S Cuckoo ray Leucoraja naevus (Müller & Henle, 1841) 70 Lusitanian 12·6 1·8 Cs Bycatch LA LC S Starry ray Amblyraja radiata (Donovan, 1808) 60 Boreal 13·7 2·4 Cs Non-target A W – Four-beard rockling Rhinonemus cimbrius L. 41 Boreal 13·9 4·6 Cg Non-target LA LC S Lesser spotted dogfish Scyliorhinus canicula L. 75 Lusitanian 12·4 2·1 Cs Bycatch LA LC S Sole Solea solea L. 60 Lusitanian 17·2 4·8 Wg Target LA LC S Spurdog Squalus acanthias L. 105 Atlantic 15·0 2·5 Cs Bycatch LA W – Lesser weaver Trachinus vipera (Cuvier, 1829) 15 Lusitanian 17·2 4·5 Wg Non-target A W S Norway pout Trisopterus esmarki (Nilsson, 1855) 25 Boreal 13·6 3·2 Cs Target A W S Bib Trisopterus luscus L. 46 Lusitanian 17·6 4·4 Wg Non-target LA LC N Poor cod Trisopterus minutus L. 40 Lusitanian 16·6 3·5 Ws Non-target A W N Species were categorized into a number of assemblages based on their thermal physiology, ecology, biogeography and exploitation status (Table 1). These assemblages are not mutually exclusive and each species appears in one or more assemblage categorization. This approach allows the identification of those traits most related to the climate change response with greater statistical power afforded by combining data from more than one species (Maxwell & Jennings 2005). The autumn thermal preference of each fish species was described using: (i) the most preferred temperature, and (ii) the range of the preferred temperatures (for details see Supplementary Appendix S1). The preferred temperatures of fishes were bimodally distributed: species preferring temperatures below 15·5 °C were classified as relatively cold-tolerant and those preferring temperatures above that level as warm-tolerant. Most species (n = 21) had narrow thermal ranges spanning less than 4 °C; a few species had slightly wider thermal ranges, such as dab, sole, solenette, lesser weaver, bib, plaice, four-bearded rockling and hake (Table 1). We used body size as a proxy measure of ecological performance. Body size is a good descriptor of life history and demography and also of production, consumption and metabolism (Reynolds et al. 2005; Jennings, De Oliveira & Warr 2007). Large-bodied species were defined as the 18 species with a maximum length ≥ 60 cm (Table 1). Species with numerical abundance lower than (or greater than) median numerical abundance were categorized as less abundant or abundant, respectively. The spatial extent of occurrence was measured as the mean number of ICES statistical rectangles occupied, and less common (or widespread) species had less than (or greater than) the median number of rectangles. The biogeographic affinities [boreal (northern) and Lusitanian (southern)] of each species were derived from the scientific literature (Wheeler 1969; Yang 1982). Exploitation status was based on stock assessment reports and regional atlases, and species were categorized as ‘target’, ‘bycatch’ and subject to some fishing mortality and ‘non-target’. Species’ geographic distributions were summarized using the centre of distribution estimated as the mean latitude weighted by the natural log of the mean abundance (survey catch) in each statistical rectangle (Rindorf & Lewy 2006). We used four measures of geographic distribution: the mean latitude, minimum latitude, maximum latitude and mean depth. Change in distribution was standardized by calculating anomalies of the departure from the mean over the 25-year study period. Latitude anomalies were converted from degrees to kilometres. Positive latitude anomalies represent northward change in species’ centre of distribution, whereas positive depth anomalies represent shallowing. Assemblage-scale distribution measures were calculated from the average of geographic anomalies across component species. All relationships between geographic response and time or a climate variable were tested using robust regression (Venables & Ripley 2002). A systematically changing survey distribution could confound the detection of climate-related geographic shifts. The number of survey stations has varied over time particularly during 1980–1984, but since then a relatively constant grid of > 70 stations have been surveyed each year. In spite of the changing number of stations in the early period, mean depth and mean latitude of the stations has remained stable and the interannual variation in survey distribution explained relatively little interannual variance in fish distributions, except for the redfish Sebastes viviparus Krøyer, 1845 which was excluded from the analysis. Long-term environmental data were provided by ICES (http://www.ices.dk). For the time series analysis, bottom temperatures (from the lower half of the water column) were averaged for winter (January–March) for 80 0·5° × 1° ICES statistical rectangles (there were insufficient replicate stations (< 15 rectangles) to perform an equivalent analysis for the summer period). Southern North Sea salinity data are collected from near-surface waters by ferries travelling between Harwich and Rotterdam at weekly intervals at approximately 52° N (Joyce 2006). The data were averaged by month and a winter mean taken for January–March. The North Atlantic Oscillation Index (NAOI) is the normalized sea level pressure difference between Gibraltar and Iceland. An annual index was calculated by averaging the winter (December–February) values and the data were sourced from http://www.cru.uea.ac.uk/cru/data/nao.htm (Jones, Jonsson & Wheeler 1997). The Gulf Stream Index (GSI) is a measure of the latitudinal height of the north wall of the Atlantic Gulf Stream and was sourced from web.pml.ac.uk/gulfstream/inetdat.htm (Taylor & Stephens 1980). GSI is not directly linked to North Sea conditions but is an indicator of regional North Atlantic climate. A composite index of North East Atlantic climate change was calculated as the first principal component axis of the 5-year running averages of five variables (winter bottom temperature, NAO, GSI, salinity and inflow). We used right-aligned 5-year running means calculated from the current year and the four previous years to approximate a fish's lifetime environmental experience. North Atlantic Current inflow into the North Sea is linked to regional climate variability, local biological productivity and fish recruitment success (Reid et al. 2003; Pingree 2005; ICES 2006a). Monthly predictions of net inflow across a section between Shetland and Orkney were derived from runs of a coupled physical, chemical and biological model system (NORWECOM) (Skogen et al. 1995). Water transport was measured in Sverdrups (106 m3 s−1) and increasing negative values represent greater southward inflow of Atlantic water. A demersal exploitation rate was calculated for each year between 1980 and 2003 as the catch-weighted sum of demersal fish fishing mortalities, as estimated in ICES North Sea stock assessments, for cod, haddock, saithe, whiting, plaice, and sole (Daan et al. 2005). Results climate change in the north east atlantic and north sea The North Atlantic Oscillation and the Gulf Stream indices have increased, peaking in 1995 with strong negative values in 1985 and 1996 (Fig. 1a,b). North Sea winter bottom temperatures have risen by 1·6 °C over 25 years, a 1 °C increase occurred in 1988–1989 alone (Fig. 1c). The mean annual temperature increase was 0·07 °C (± 0·02 SE; F1,23 = 10·9, P = 0·003). The warming bottom temperatures coincided with a long-term shift towards a positive NAO phase, a northward shift in the Gulf Stream and stronger Atlantic inflow into the northern North Sea (NAO: r = –0·81, P < 0·0001; GSI: r = –0·76, P < 0·0001: Fig. 1a,b). The inflow of Atlantic water into the North Sea also increased, peaking in 1990 followed by a slight weakening (Fig. 1d). Correspondingly, salinity in the southern North Sea was lower than average in the early 1980s and greater around 1990, coinciding with the peak inflow of saline Atlantic water (Fig. 1e). The first and second principal component axis capture 73% and 14%, respectively, of the variation in the 5-year running means of these five climate variables. The first axis score (dotted line) represents a longer-term trend in climate becoming negative by 1990, reaching a minimum in 1995 before rising to near zero in 2000 and stabilising thereafter (Fig. 1f). The second axis (solid line) represents shorter-term climate variability and was negative in the mid-1980s and mid-1990s and positive in the early 1990s and around 2000 (Fig. 1f). Figure 1Open in figure viewerPowerPoint Physical climate indices from the North Sea and North East Atlantic spanning 1980–2004; (a) North Atlantic Oscillation Index (December–February), (b) Gulf Stream Index, positive values represent northward displacement of the Gulf Stream wall, (c) mean winter bottom temperature (January–March), (d) net inflow between Orkney and Shetland, (e) southern North Sea salinity anomaly, and (f) the principal component axes (first axis, grey points; second axis, bold line) of the 5-year running mean of these five climate indices. Annual values are represented by the connected points with the 5-year right-aligned running mean represented by the bold line. deepening response of individual fish species over time Most species have deepened over time with 11 deepening significantly (at P < 0·01; Fig. 2). On average, the 22 deepening species have deepened by ~5·5 m decade−1 (range: 0·6–14 m decade−1). Coldwater species, like megrim and anglerfish, are deepening fastest with warm-water species shallowing over time (3, 5; for Latin names see Table 1). Sole and bib are southerly warm-water species and have been shallowing at rates of 7·6 m and 6 m decade−1, respectively (Fig. 3). Similar to the deepening pattern, these shallowing trends are also consistent with climate change and belie an initial deepening in the cool period around 1985 and subsequent shallowing in the warmer period around the mid- to late1990s (Fig. 3). Figure 2Open in figure viewerPowerPoint Trend in depth anomaly of individual fishes over time (m decade−1). Solid points are significant at P < 0·01. Figure 3Open in figure viewerPowerPoint Deepening of four coldwater boreal fishes and shallowing of two warmwater southern fishes over time. The trend line is a loess smoother (span = 0·75). Figure 5Open in figure viewerPowerPoint Trend in geographic response of different demersal fish assemblages over time; (a) mean depth, (b) mean latitude, (c) mean minimum latitude and (d) mean maximum latitude. Black and grey points indicate statistical significance at P ≤ 0·001 and P ≤ 0·01 respectively. The x-axis represents the direction and strength of geographic response over time – the slope of a regression of distribution measure on year. Positive values indicate shallower (panel a) or northerly distribution (panels b–d), with negative anomalies representing deepening or a more southerly distribution. deepening response of fish assemblages over time Overall, the 28-species North Sea demersal fish assemblage has deepened significantly at a rate of ~3·6 m decade−1 (F1,23 = 18·4, P < 0·0002; Fig. 4a). The mean depth varied from year to year but tracks temperature over the longer time-scale; the assemblage was shallowest in the cool mid-1980s and deepest during the peak warming in the mid-1990s and shallows slightly thereafter (Fig. 4a). Figure 4Open in figure viewerPowerPoint Annual variation in geographic response of the North Sea demersal fish assemblage. Anomalies of (a) mean depth, (b) mean latitude, (c) minimum latitude and (d) maximum latitude. Positive anomalies indicate shallower (panel a) or northerly distribution (panels b–d), and negative anomalies representing deepening or a more southerly distribution. Only the mean depth anomaly exhibits a significant trend over time (P < 0·001). The solid line is the 5-year running mean of the first principal component axis representing climate change (see Fig. 1f). The deepening response over time was consistent across all but one assemblage and significant for 14 out of 19, at P < 0·01 (Fig. 5a). The average rate of deepening for these assemblages was 4·3 m decade−1 (range: 3–6 m decade−1). Those assemblages not exhibiting a significant depth response are comprised of species that are warm-tolerant, small-bodied, less common with relatively low occupancy, have a northern range boundary in the North Sea and are unexploited; these species include scaldfish, solenette and bib (Table 1). latitudinal response of fish assemblages over time In contrast to the relatively coherent deepening, the demersal fish assemblage exhibited heterogeneous latitudinal range changes with no overall trend north or south (Fig. 4b–d). Mean latitude, and to a lesser degree maximum latitude, was more southerly in the cool 1980s and farther north during the warmer 1990s (Fig. 4b,d). Minimum latitude was more northerly in the early cooler years, moving southward in the warmer years before retracting northward in the late 1990s (Fig. 4c). Two broad geographic responses to climate change patterns are discernable: (i) a northward shift both in mean latitude and maximum latitude, and (ii) a mixed or southward shift in minimum latitude. The northward shift in mean latitude over time was exhibited by assemblages comprised of abundant, widespread, warm-tolerant species with narrow thermal ranges, such as grey gurnard and poor cod (Fig. 5b). Some assemblages did not deepen and instead revealed a southward shift of the minimum latitude. These assemblages were comprised of unexploited, warm-tolerant, small-bodied, abundant, less common, low-occupancy species, with a northern range boundary in the North Sea, such as scaldfish, solenette and bib. sensitivity of the depth and latitude response to temperature, climate and exploitation Depth and latitude anomalies responded most significantly to 5-year running means of winter bottom temperature while accounting for year effect (Fig. 6) and composite climate index when aggregated at the assemblage-level. This pattern only holds when smoothed by running means and holds best for the 5-year running mean; the zero-lagged data did not yield significant relationships. The deepening of the whole North Sea fish assemblage was related to winter bottom temperature and composite climate index: the assemblage was relatively shallow in cooler years and deeper in warmer years (winter bottom temperature 5-year running means: F1,23 = 25·9, P < 0·001; PCA 1 of 5-year running means: F1,23 = 24·4, P < 0·001). All 19 assemblages deepened with warmer climate at a rate of 2–7 m °C−1 and the mean and maximum latitudes of most assemblages moved northward with warming climate at a rate of 10–70 km °C−1. However, the minimum latitude moved southward with warming climate by up to 80 km °C−1 for many assemblages, except for the northward movement of the minimum latitude of warm-specialist species at a rate of 40 km °C−1. Figure 6Open in figure viewerPowerPoint Temperature sensitivity of the geographic response of fish assemblages measured as the (a) deepening in metres per degree of warming averaged over the current and previous 4 years (m °C−1) or (b–d) range shift in kilometres moved per degree of warming (km °C−1). Negative values represent a southward shift in response to warming and positive values a northward shift. Solid points indicate statistical significance of the overall model at P < 0·001 and grey points P < 0·01. Demersal exploitation rate explained relatively little variance in the interannual variation of geographic distribution of the demersal fish assemblage. When all combinations of three explanatory variables (composite climate index or bottom temperature, year and exploitation rate) are considered together, only bottom temperature or climate index significantly explain most variance in depth anomaly of the demersal fish assemblage (Appendix S2, Supplementary material). This pattern also holds for assemblages of target or bycatch species. Discussion We present evidence for a coherent deepening of the North Sea fish assemblage in response to climate change. The rate of deepening of the whole assemblage was 3·6 m decade−1 and for individual species ranges up to 10 m decade−1. This rate of deepening is analogous and comparable to upward altitudinal response of terrestrial organisms, which averages 6·1 m decade−1 (Parmesan & Yohe 2003). Before considering the indicator properties of the deepening of the demersal fish assemblage, we consider these three questions: (i) What is the ecological significance of deepening fishes? (ii) Why is the deepening response more coherent than the latitudinal response? (iii) Are changes in fish distribution largely a consequence of fisheries exploitation? The ecological significance of upward-shifting alpine fauna is readily apparent. These species face shrinking habitats and greater likelihood of extinction (Grabherr, Gottfried & Paull 1994). However, the ecological significance of the deepening