Integrating the impact of global change on the niche and physiology of marine nitrogen‐fixing cyanobacteria

Abstract Marine nitrogen fixation is a major source of new nitrogen to the ocean, which interacts with climate driven changes to physical nutrient supply to regulate the response of ocean primary production in the oligotrophic tropical ocean. Warming and changes in nutrient supply may alter the ecological niche of nitrogen‐fixing organisms, or ‘diazotrophs’, however, impacts of warming on diazotroph physiology may also be important. Lab‐based studies reveal that warming increases the nitrogen fixation‐specific elemental use efficiency (EUE) of two prevalent marine diazotrophs, Crocosphaera and Trichodesmium, thus reducing their requirements for the limiting nutrients iron and phosphorus. Here, we coupled a new diazotroph model based upon observed diazotroph energetics of growth and resource limitation to a state‐of‐the‐art global model of phytoplankton physiology and ocean biogeochemistry. Our model is able to address the integrated response of nitrogen fixation by Trichodesmium and Crocosphaera to warming under the IPCC high emission RCP8.5 scenario for the first time. Our results project a global decline in nitrogen fixation over the coming century. However, the regional response of nitrogen fixation to climate change is modulated by the diazotroph‐specific thermal performance curves and EUE, particularly in the Pacific Ocean, which shapes global trends. Spatially, the response of both diazotrophs is similar with expansion towards higher latitudes and reduced rates of nitrogen fixation in the lower latitudes. Overall, 95%–97% of the nitrogen fixation climate signal can be attributed to the combined effect of temperature on the niche and physiology of marine diazotrophs, with decreases being associated with a reduced niche and increases resulting due to a combination of expanding niche and temperature driven changes to EUE. Climate change impacts on both the niche and physiology of marine diazotrophs interact to shape patterns of marine nitrogen fixation, which will have important implications for ocean productivity in the future.


| INTRODUC TI ON
Marine dinitrogen (N 2 ) fixation, or 'diazotrophy' is a key source of reactive nitrogen (N) to the global ocean supplying between 68 and 164 Tg N year −1 (Gruber & Sarmiento, 1997;Jickells et al., 2017;Luo et al., 2014;Tang et al., 2019;Wang et al., 2019) and fuels primary production in N limited regions of the ocean. Earth system models (ESM) project that N 2 fixation will decline over the coming century. As the climate driven signal in marine N 2 fixation emerges earlier than the trends in primary productivity, marine N 2 fixation may shape the response of primary producers to climate change (Wrightson & Tagliabue, 2020). The predicted increase in ocean temperature will affect multiple aspects of diazotrophy, with emphasis to date on the impact of warming on stratification and nutrient supply Sohm et al., 2011;Weber & Deutsch, 2014), with some work on how changing temperature will alter the physiology and thermal niche of diazotrophs (Fu et al., 2014;Jiang et al., 2018;Yang et al., 2021). Increasing sea surface temperature (SST) plays a primary role in controlling the thermal niche of diazotrophs. In the low latitudes, warming may surpass their thermal maximum leading to exclusion, whilst increasing temperatures below their thermal maximum allows poleward expansion (Boatman et al., 2020;Breitbarth et al., 2007;Fu et al., 2014).
Changing nutrient supply due to stratification can either open a competitive niche for diazotrophy if the supply of N declines such that it limits non-diazotrophs (Weber & Deutsch, 2010), or it can restrict rates of N 2 fixation if the supply of phosphorus (P) or iron (Fe) declines (Hutchins & Capone, 2022).
Alongside temperature and nutrient availability, another potentially important driver that may impact marine diazotrophy is carbon dioxide (CO 2 ). When Trichodesmium and Crocosphaera are exposed to increased concentrations of CO 2 , enhanced growth and N 2 fixation rates have been observed, and it has been suggested that like temperature, CO 2 may define an upper limit on N 2 fixation rates (Hutchins et al., 2007Walworth et al., 2021). Increased CO 2 concentrations have been proposed to reduce the diazotroph's requirement for carbon concentrating mechanisms (CCM), enabling more energetic investment into N 2 fixation, photosynthesis, and growth (Boatman et al., 2018). However, CO 2 only has a strong impact on diazotrophy under Fe replete conditions (Fu et al., 2008;Walworth et al., 2016). These results imply that increasing CO 2 in the future may benefit marine diazotrophs mostly in regions that are replete in Fe, such as the tropical North Atlantic Ocean.
Temperature can also indirectly impact diazotroph growth by influencing enzyme efficiency and altering diazotroph physiology.
Recent studies have used the concept of elemental use efficiencies (EUE) to account for the effect of temperature on enzyme efficiency and resource requirements of diazotrophy in an integrated manner (Jiang et al., 2018;Yang et al., 2021). Thermal shifts in N 2 fixation specific EUEs are calculated by measuring the rate of N 2 fixation normalized to the cellular element quotas of the diazotroph (e.g., using the Fe quota gives the iron use efficiency [IUE]), and observing how it changes across the diazotroph's thermal window. An increase in the EUE means that the diazotroph is performing more N 2 fixation per unit element considered, leading to a reduction in the nutrient demand of the diazotroph. These temperature driven changes to diazotroph physiology are mediated by changes in the biological utilization of the limiting nutrients Fe and P in response to warming. Thermal performance curves and N 2 fixation specific EUEs for P and Fe have been measured for two marine diazotrophs, Trichodesmium and Crocosphaera. Crocosphaera has a narrower thermal window for growth than Trichodesmium, as it grows between 20 and 35°C compared to 17 and 35°C for Trichodesmium (Boyd et al., 2013) with the thermal optimum for growth occurring at 28.7 and 27.9°C for Trichodesmium and Crocosphaera, respectively (Figure 1a;Jiang et al., 2018;Yang et al., 2021). The N 2 fixation EUEs also respond differently to temperature depending on the element and the diazotroph in question. The thermal optimum for Trichodesmium IUE and phosphorus use efficiency (PUE) occur at 31.8 and 30.5°C, respectively, while for Crocosphaera IUE and PUE the thermal optimums occur at 27.5 and 31.8°C, respectively (Figure 1b,c;Jiang et al., 2018;Yang et al., 2021). The different responses of both diazotrophs to temperature, including their growth rates, iron, and phosphorus use efficiencies, highlight the need for more information F I G U R E 1 Thermal performance curves of growth (a), iron use efficiency (b), and phosphorus use efficiency (c) for Trichodesmium (blue) and Crocosphaera (red). Curves were fitted to the data from Jiang et al. (2018) for Trichodesmium and from Yang et al. (2021) for Crocosphaera. Data points are shown by crosses. on how thermal fitness of each diazotroph shapes the response of diazotrophy to future ocean warming.
ESMs are the main tool to investigate how the future ocean will respond to climate change, and their results underpin important assessments by the IPCC (Eyring et al., 2016;van den Hurk et al., 2018).
However, current ESMs have an incomplete representation of N 2 fixation as focus is primarily upon the impacts of temperature on the niche of marine diazotrophs (Wrightson & Tagliabue, 2020). As temperature has the potential to modulate not only the extent of the thermal niche of diazotrophs but also their physiology via changing EUEs, diazotroph thermal fitness dynamics need to be incorporated into ESMs to assess the integrated climate change response (Boatman et al., 2020;Jiang et al., 2018;Yang et al., 2021). Such models should account for the temperature impacts on both the growth and niche of diazotrophs, as well as incorporating the effects of warming on diazotroph physiology via EUEs. Alongside these factors, changes in the physical environment (driven by warming, but also by changes in winds and salinity) will also alter the availability of nutrients. As growth rates and EUEs respond to temperature distinctly between diazotrophs, there is also a need to assess whether the ESM parameterizations based on Trichodesmium or Crocosphaera affect the response of diazotrophy to changes in climate. To date, the effects of temperature on growth and IUE for Crocosphaera and Trichodesmium have been assessed using an additive Michaelis-Menten based approach in response to annual average Fe concentration and SST from the NCAR CMIP5 model under the high emissions RCP8.5 scenario. The diagnostic modelling results suggest that N 2 fixation rates will increase globally by 22% and 91% for Trichodesmium and Crocosphaera, respectively (between two time slices at 2010 and 2100) due to increased IUEs and expansion of the diazotroph niche (Jiang et al., 2018;Yang et al., 2021). However, these diagnostic models focused on only temperature and Fe limitation, neglecting the role of other bottom-up and top-down drivers such as P limitation, light limitation, grazing, and competition with other phytoplankton in a fully prognostic sense. The susceptibility of diazotrophs to Fe limitation also varies as diazotrophs deploy different N 2 fixation strategies that can affect their Fe demand. For example, Trichodesmium performs N 2 fixation and photosynthesis simultaneously during the day, whilst Crocosphaera temporally segregates both processes by performing photosynthesis during the day and N 2 fixation at night (Berman-Frank et al., 2007). Trichodesmium is therefore required to satisfy the Fe demand of both processes simultaneously, while Crocosphaera can deploy a 'hot bunking' strategy that cycles the same cellular Fe pool between the two processes over the diel cycle. This has been suggested to reduce the Fe cost of Crocosphaera by 40%-50% compared to that required by Trichodesmium to fix the same amount of N 2 (Saito et al., 2011). Diazotrophs respond not only to temperature and Fe availability but to a suite of drivers such as grazing, light limitation, and fixed N, which can affect growth rates and alter the niche of diazotrophy. To assess the impact of climate change on marine diazotrophy, a holistic consideration of how temperature can affect diazotroph thermal fitness and N 2 fixation rates in the future is required (Hutchins & Capone, 2022).
The aim of this study was to investigate how diazotroph thermal fitness, both in terms of a changing thermal niche and EUEs, responds to climate change under the high emissions RCP8.5 scenario.
To do this, we developed a new state-of-the-art diazotroph compartment for the PISCES QUOTA model based upon observed diazotroph thermal performance curves of growth and EUEs to account for the thermal fitness of two marine diazotrophs, Trichodesmium and Crocosphaera, which are interchangeable within the model. Here, we describe the new model and experiments focused on investigating how the response of N 2 fixation to climate change differs between Trichodesmium and Crocosphaera, at regional scales.

| MODEL DE SCRIP TION
The new diazotroph model was developed for the PISCES QUOTA ESM, which allows for complete variable phytoplankton stoichiometry and applies optimal allocation of resources (Kwiatkowski et al., 2018). In the model, diazotroph growth and N 2 fixation are limited by temperature, light, and nutrient availability (P and Fe). N 2 fixation is facultative, allowing the diazotroph to use other forms of fixed N (nitrate and ammonium) (Holl & Montoya, 2005;Knapp, 2012;Mulholland et al., 2001). That said, diazotroph maximum growth rates are much lower than those ascribed to diatoms, nanophytoplankton and picoplankton, which results in their exclusion when only nitrate and ammonia are used as a N source. The full model description can be found in the supplementary material.
Within the model, diazotroph nutrient requirements are set by the prescribed minimum quotas, which restrict growth when nutrient concentrations do not satisfy the minimum quota. For N and P, the minimum quota is allometrically scaled, but the initial value of the minimum N and P quotas are predefined. For Fe however, the minimum quota (Q dz The Fe costs of photosynthesis, respiration, and nitrate reductase used in Equation (1) are taken from Flynn and Hipkin (1999) and follow the approach used for the other phytoplankton functional types (PFT) in PISCES QUOTA with an additional term for diazotrophs to account for the cost of N 2 fixation (Kwiatkowski et al., 2018). The x dz Nfix term represents the proportion of the diazotroph fixed N demand that comes from N 2 fixation. The Fe cost of N 2 fixation is based upon the work of Kustka, Sañudo-Wilhelmy, Carpenter, Capone, and Raven (2003), which suggested that the additional Fe requirement for growth by Trichodesmium using N 2 is ~30-50 × 10 −6 mol Fe mol −1 C, of which nitrogenase, the enzyme required for N 2 fixation, accounts for ~25%. This implies that the cost of nitrogenase is ~10 × 10 −6 mol Fe mol −1 C.
Diazotrophs also rely on the Mehler reaction which produces free oxygen radicals. In order to consume these free oxygen radicals, diazotrophs employ superoxide dismutase, which has an Fe cost of ~3 × 10 −6 mol Fe mol −1 C. The overall Fe cost for satisfying all the diazotrophs N demand from N 2 fixation is therefore 13 × 10 −6 mol Fe mol −1 C.
For this new version of the diazotroph model, the diazotroph PFT can switch between a Trichodesmium and Crocosphaera parameterization, which then alters the thermal performance curves of growth and EUEs appropriately ( Figure 1). The growth curves used in the model (Equation 3) were obtained by fitting a curve to observations of Trichodesmium (Jiang et al., 2018) and Crocosphaera (Yang et al., 2021) growth rates over a range of temperatures.
Observations of EUEs were also obtained and had curves fitted to produce the thermal performance curves for both Fe and P EUEs ( Figure 1). As the EUEs increase, the nutrient demand should decrease. Therefore, the IUE curve was then used as a simple scalar for the Fe cost of N 2 fixation. Similarly, the PUE curve was used as a scalar for the minimum P quota of the diazotroph. In this way, when the EUEs increased, the cellular Fe or P requirements decreased and when the EUEs decreased, the cellular Fe and P requirements increased. The EUEs used in this study were derived from experiments conducted under replete nutrient conditions (Jiang et al., 2018;Yang et al., 2021) to better isolate the direct and indirect drivers. The ensuing EUEs that emerge from the model integrate the effect of nutrient limitation. We used the observed thermal response curves for growth to set the maximum growth rate of each diazotroph, which is then controlled by temperature, light, and nutrient availability. This model was then run using either a fixed or temperature sensitive EUE for comparison. As our model only represents a single diazotroph for each experiment, it cannot account for any direct competition between both diazotrophs at this time. However, our model is able to highlight how different diazotroph assumptions influence the model responses to spatial and temporal variability. We model the thermal performance curve for diazotroph maximum growth rates via the following generic empirical equation: where dz max = maximum diazotroph growth rate (day −1 ); T = temperature (°C).
The temperature range for Trichodesmium growth was set from 17 to 35°C, while Crocosphaera has a narrower thermal window with growth permitted between 20 and 35°C. Values used to calculate the growth curves for both diazotrophs in Equation (3)

| Calculation of the nitrogen fixation EUEs for both diazotrophs
To incorporate the thermal performance curves of the EUEs into the model, we fitted a curve to the observations from Jiang et al. (2018) and Yang et al. (2021) (Figure 1b,c; Equations 4 and 5) and converted them into a scaling term where the scaling was set to 1 when the diazotroph growth rate was 0.1 day −1 , which was the reference growth rate for the calculation of the Fe cost of N 2 fixation (Kustka, Sañudo-Wilhelmy, Carpenter, Capone, & Raven, 2003). At a growth rate of 0.1 day −1 , the Fe cost of N 2 fixation is 13 × 10 −6 mol Fe mol −1 C, and, since the EUE modulates the nutrient demand, the IUE scaling relationship was then used to scale the Fe cost of N 2 fixation, and the PUE scaling relationship was used to scale the minimum P quota.
The generic scaling equations are Values used to calculate the thermal performance curves of the EUEs of both diazotrophs are provided in Table 1. Within the model, the minimum Fe quota of the diazotroph is set by the sum of several Fe costs (Equation 1). To incorporate the IUEs of N 2 fixation into the model, a scaling approach was used. Following Kustka, Sañudo-Wilhelmy, Carpenter, Capone, and Raven (2003), the Fe cost of N 2 fixation is 13 × 10 −6 mol Fe mol −1 C for a 0.1 day −1 growth rate, so the scaling needs to be set to 1 where growth is equal to 0.1 day −1 , as this was the growth rate at which the reference Fe cost of N 2 fixation was calculated. The IUE curve was divided by the IUE of the diazotroph when growth was 0.1 day −1 (Equation 6). When the IUE increases, the Fe cost of N 2 fixation would decrease so the reciprocal of the IUE scaling was required (Equation 7). The Fe cost scaling was then used to modulate the Fe cost of N 2 fixation depending on temperature (Equation 8). Following the approach used for the IUEs, the PUE scaling was performed in a similar manner (Equations 9 and 10). However, to account for the change in the P (2) Facultative term proportion of N supply from N 2 fixation , demand of the diazotroph, the PUE scaling was used to modulate the minimum P quota of the diazotroph (Equation 11), IUE dz 0.1 = IUE at 0.1 day −1 growth rate (Trichodesmium = 33.49 mol N h −1 mol −1 Fe, Crocosphaera = 20.64 mol N h −1 mol −1 Fe), PUE dz 0.1 = PUE at 0.1 day −1 growth rate (Trichodesmium = 0.25 mol N h −1 mol −1 P, Crocosphaera = 0.2628 mol N h −1 mol −1 P).

| Model experiments
Several simulations were performed to investigate how climate change affects the different diazotrophs. Our reference simulations include specific thermal performance curves and temperature dependent EUEs for either Trichodesmium or Crocosphaera. To test for the influence of a lower Fe cost of N 2 fixation for Crocosphaera, we also conducted an additional experiment where the Fe cost of N 2 fixation was reduced by 40% (7.8 × 10 −6 mol Fe mol −1 C) following Saito

| Model nutrient limitation
Before discussing the results, it is important to highlight that  (Dutkiewicz et al., 2012(Dutkiewicz et al., , 2014Sohm et al., 2011;Zehr & Capone, 2020). In the Atlantic Ocean, episodic Fe input controls patterns of N 2 fixation with increased Fe concentrations

| RE SULTS AND D ISCUSS I ON
We first focus on the reference simulations for both Trichodesmium and Crocosphaera, using the state-of-the-art version of the model with both temperature dependent EUEs active (Tricho REF, Croco REF , Table 2).

| Regional response of Diazotrophy to climate change
Globally, total N 2 fixation is projected to decrease over the next century for both Trichodesmium and Crocosphaera. The decline in N 2 fixation is stronger for Crocosphaera than for Trichodesmium with integrated N 2 fixation decreasing from 69.1 to 58.9 Tg N year −1 (−15% or −10.2 Tg N year −1 ) and from 70.6 to 65.8 Tg N year −1 (−7% or decline was reduced and delayed relative to Crocosphaera with only a 3% decline occurring at the end of the century (Figure 3c)

| Spatial patterns of marine nitrogen fixation
The spatial distribution of the N 2 fixation climate signal (  The direct impact of changes in temperature on diazotroph thermal performance through changes in diazotroph growth rates was able to explain 55%-59% of the N 2 fixation climate signal. Globally, SST increases by between 1 to 12°C by 2091-2100 under the high emissions RCP8.5 scenario. If temperature surpasses the thermal optimum for growth, the diazotroph will experience thermal stress (red regions, Figure 5), which would decrease maximum growth and N 2 fixation rates. If the temperature is below the thermal optimum for growth (blue regions, Figure 5b), the diazotroph would not be thermally stressed and so growth and N 2 fixation rates would increase with warming leading to an expanded thermal niche. We evaluated the role of temperature using the monthly maximum SST during 2091-2100. Combining the spatial maps of the change in N 2 fixation ( Figure 5a) and thermal stress (quantified using the difference between SST and Topt) associated with diazotroph growth (Figure 5b), two regimes could be identified. The first regime was assigned to the regions where the diazotroph was thermally stressed (i.e., SST > Topt) and, as expected, N 2 fixation was restricted (Blue regions, low latitudes). This regime represented 19.3% and 22.5% of the niche of Trichodesmium and Crocosphaera, respectively ( Figure 5c; Table 3). The second regime was associated with regions where the diazotroph was not thermally stressed (i.e., SST < Topt) and N 2 fixation increased as expected due to warming (red region, high latitudes = expanding thermal niche) and accounted for 35.9%

| Identifying the drivers controlling the change in nitrogen fixation
and 28.7% of the niche of Trichodesmium and Crocosphaera respectively ( Figure 5c; Table 3). This assessment of the effect of temperature on diazotroph thermal performance in regard to growth left almost half of the ocean (black region, 41% and 45% of the niche of Trichodesmium and Crocosphaera respectively) in which the change could not be explained (Figure 5c). In these regions, despite being thermally stressed (SST > Topt), N 2 fixation increased. This was due to temperature driven changes in diazotroph physiology mediated through altered EUEs in response to warming, explaining the climate trend in N 2 fixation for around a quarter of the diazotroph's niche, with the remainder being attributed to the emergence of a new N-limited niche, which promoted diazotrophy. As discussed above, temperature can also affect rates of N 2 fixation by altering the efficiency of enzymes, and the EUEs can be used to explore this. If temperature surpasses the thermal optimum of the EUEs (i.e., SST > Topt EUE), the declining EUEs (e.g., due to enzymes denaturing) lead to increased nutrient demand and enhanced nutrient limitation (red areas, Figure 5d,e). Alternatively, if the temperature remains below the thermal optimum of the EUEs (i.e., SST < Topt EUE), then EUEs increase with ocean warming, alleviating nutrient limitation, and promoting both growth and N 2 fixation despite reduced maximum growth rates (blue areas, Figure 5d,e).
This concept of temperature adjusted EUEs can be used to further explain the N 2 fixation trend in regions not explained by the temperature effects on the thermal niche of diazotrophy (black region, Figure 5c). First, a regime can be identified where one or both EUEs for each diazotroph have increased due to warming, and N 2 fixation rates increased due to reduced nutrient limitation despite diazotroph growth being thermally stressed (shades of orange/yellow), this regime accounted for 22.5% and 27.8% of the niche of Trichodesmium and Crocosphaera, respectively (Figure 5f; Table 3). A second regime displayed reduced EUEs in response to warming alongside declining N 2 fixation, and despite no thermal stress on diazotroph growth, N 2 fixation declined likely due to enhanced nutrient demand (shade of blue areas): this regime represented 3.4% and 4.4% of the niche of Trichodesmium and Crocosphaera respectively (Figure 5e; Table 3).
For Trichodesmium, both Fe and P use efficiencies increase, but for Crocosphaera only P use efficiency increases within the black region (Figure 5d,e). Thus, changing EUEs due to warming explain the response of N 2 fixation in around a quarter of their niche. Finally, temperature changes due to climate can also indirectly impact diazotrophy through the decline in the upper 100 m N inventory due to enhanced vertical stratification creating a niche for diazotroph in regions with excess P relative to N. The decrease in the N inventory leads to increased N limitation of fast growing non-diazotroph PFTs, providing the slower growing diazotrophs with a competitive advantage. This new niche for diazotrophy emerged largely in the Pacific Ocean (pink area, Figure 5f). This new competitive niche explained 13.8% of the niche for both diazotrophs (Figure 5f; Table 3).
Reducing the Fe cost of N 2 fixation for Crocosphaera produced very similar results to those of the standard Crocosphaera model (Table 3).
Overall, by applying this environmental grouping approach, 95% and 97% of the spatial N 2 fixation signal can be attributed to drivers for Trichodesmium and Crocosphaera, respectively. Around half of the signal is attributed to the effect of temperature on diazotroph growth defining a thermal niche for diazotrophy, a quarter due to the effect of warming via changing EUEs and the remainder due to competition with non-N 2 -fixing plankton in N limited regions. The small fraction of the ocean (at most 5%) that cannot be attributed to these factors are being controlled by other factors such as grazing, light availability or community shifts (Table 3).
In our model, we can further examine how the changing diazotroph physiology due to the effect of temperature on EUEs was reflected in their overall nutrient limitation. Here, we focus on regions where the climate trend in N 2 fixation rates was not simply due to the temperature effect on the thermal niche of diazotrophy (i.e., the black region in Figure 5c). We isolated this area and compared the climate trend of diazotroph nutrient limitation for both Trichodesmium and Crocosphaera to the model runs where no temperature-EUE parameterisation was present (Figure 6a,b).
In general, for the majority of this region, nutrient limitation was

| Responses to warming
The current version of the model assumes that the thermal performance of the modelled diazotroph is fixed and neglects any thermal evolution. This means that once their maximum thermal threshold is surpassed by rising SST, they are excluded, which drives a large decline in both their niche and absolute N 2 fixation rates. However, biology is highly dynamic, with both evolution and adaptation likely to occur.
A recent experimental evolution study comparing Trichodesmium with Crocosphaera under sustained thermal selection suggested that the former showed little capacity to adapt to warming, but instead relied on non-genetic plasticity to meet temperature challenges (Qu et al., 2022). Crocosphaera however exhibited a limited ability to adapt to supraoptimal warming supported by a suite of specific genetic changes, suggesting that evolutionary capacity may need to be considered at least for this diazotroph (Qu et al., 2022). This may imply that in the future, Crocosphaera may more readily adapt to warming compared to Trichodesmium, enabling Crocosphaera to potentially occupy the niche that Trichodesmium has been thermally excluded from.
However, more experimental work is required to better understand how both diazotroph groups adapt to warming before this evolutionary capacity can be incorporated into ESMs. Our results provide a gauge as to the rate at which temperature will exceed the thermal optimum of growth for Trichodesmium and Crocosphaera and how this compares to experimental studies of thermal adaptation of both diazotrophs. Our results imply that, based on the monthly maximum SST, the thermal optimum for growth has already been surpassed for much of the low latitude ocean by the end of the historical period for both diazotrophs (Figure 8). At the end of the historical period (1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005) the area of the diazotroph's niche where they are experiencing thermal stress (i.e. where temperature exceeds the thermal optimum) was 6%-31% for Trichodesmium and 15%-43% for Crocosphaera (Figure 8; Table 4).
By the end of the century, the area of thermal stress roughly doubles for both diazotrophs, and for most of the low latitudes,

F I G U R E 7 Contribution of each environmental group to the climate signal of nitrogen fixation for Tricho REF (solid), Croco REF (hatched)
and CrocoLowFe REF (dotted). Temperature impacts on the niche of diazotrophy: Blue (restricting growth/N 2 fix), red (promoting growth/N 2 fix), and pink (decreasing N inventory due to stratification opening a competitive niche for diazotrophy). Temperature impacts on physiology: Orange (1 or both elemental use efficiencies [EUEs] more efficient) and cyan (1 or both EUEs decreasing leading to nutrient limitation).
Black bars indicate the environmental group which cannot be explained by temperature. Percentages show how much of the area of the diazotroph's geographic niche each group occupies.

F I G U R E 8
Percentage of the area of the diazotroph's niche where temperature has surpassed the thermal optimum for growth for both Trichodesmium (blue) and Crocosphaera (red) using the monthly maximum sea surface temperature (SST) (solid lines) and the annual mean SST (dashed lines). Black vertical line indicates the end of the historical period and the start of the RCP8.5 forcing.
the diazotrophs are thermally stressed within 10 years of the high emissions RCP8.5 scenario (Figure 8; Table 4). This indicates that if diazotrophs cannot adapt to warming in the future they may be excluded from broad regions of the low latitudes. If the annual mean temperature is used the area of thermal stress is ~25% less than if the monthly maximum temperature is used but the outlook is the same ( Table 4). The thermal niche of each diazotroph is determined by the specific thermal performance curves for growth that define the thermal thresholds for diazotroph growth. Due to the colder temperatures at high latitudes in the historical period (1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005), Trichodesmium and Crocosphaera were excluded from 31% to 40% and 36% to 49% of the ocean, respectively. With future warming, this area decreases over the coming century by up to 8% for Trichodesmium or by up to 10% for Crocosphaera, as diazotrophs expand their niche into higher latitudes (Table S1). In this study, the monthly maximum temperature was used as these are the actual conditions the diazotrophs will experience in the model.
Our estimates of diazotroph thermal stress are based on a strict temperature criterion that states if the specific thermal optimum of diazotroph growth is surpassed then the diazotroph is thermally stressed. However, a recent modelling study investigating optimal growth of Trichodesmium defined optimal growth conditions as those that allow growth rates of >0.25 day −1 and suggested that when considering the combined impact of temperature, light, and Fe availability, the area of optimal conditions experienced by Trichodesmium may increase by up to 173% by 2100 (Boatman et al., 2020). The study by Boatman et al. (2020) also indicated that the thermal niche of Trichodesmium will likely expand at high latitudes and reduce in equatorial regions, agreeing with the findings of this study. It is worth noting however, that under lower emissions scenarios, the associated reduction in warming would reduce the extent of thermal stress for both diazotrophs by 12%-17% compared to the high emissions RCP8.5 scenario (Table 2; Figure S3).
Any reduction in warming and thermal stress under alternative emissions scenarios would lead to a lesser degree of thermal exclusion and enable diazotrophs to remain at low latitudes, promoting N 2 fixation in these regions (Table 4). At high latitudes however, any reduction in warming under lower emissions trajectories would restrict the thermal expansion and greater N 2 fixation seen under the high emissions scenario.

| WIDER IMPLI C ATI ON S AND FURTHER WO RK
Currently, the model can only represent one diazotroph (either Trichodesmium or Crocosphaera) at a time, which does not allow for competition between the two diazotrophs to occur. It would therefore be interesting to implement both diazotrophs into the model as co-existing PFTs to investigate how competition between the two organisms in the model impacts rates of N 2 fixation. In the ocean, Trichodesmium and Crocosphaera would compete for resources, particularly Fe and P. Both microbes are adapted to low P environments and are able to access dissolved organic phosphate (DOP) alleviating P limitation (Dyhrman et al., 2002;. Unlike Crocosphaera, Trichodesmium is also able to deploy high affinity P strategies, enabling growth on polyphosphate and phosphonates providing a competitive advantage and potentially reducing competition Orchard et al., 2010). Both organisms also occur at different depths with Crocosphaera generally present deeper in the water column to avoid photoinhibition, while Trichodesmium is better able to cope with high irradiance levels and prefers the high light surface waters, and so spatial separation may also prevent competition (Andresen et al., 2010;Inomura et al., 2019).
As with all global ocean biogeochemical models, nutrient limitation in our model is determined by the most limiting nutrient (either Fe or P for diazotrophs). However, throughout regions of the Atlantic and Pacific Oceans, diazotrophs have been observed to be exposed to simultaneous Fe and P co-limitation (Cerdan- Garcia et al., 2021;Mills et al., 2004;Wen et al., 2022), with a recent metaproteomic study suggesting that Fe-P co-stress may be considered the normal conditions that Trichodesmium is exposed to in the North Atlantic (Held et al., 2020). Under laboratory conditions, enhanced growth and N 2 fixation rates were observed for both Trichodesmium and Crocosphaera when each diazotroph was exposed to Fe and TA B L E 4 Area of the thermal niche of diazotrophy where thermal stress is occurring for Trichodesmium and Crocosphaera for both the monthly maximum temperature and annual mean temperature for the historical period (1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005) and for several RCP climate forcing scenarios (RCP4.5, RCP6.0 and RCP8.5 (2091-2100)) P co-limitation (Garcia et al., 2015;Walworth et al., 2016). These results suggest that both diazotrophs have adapted for growth in Fe and P co-limited conditions, highlighting the need to incorporate nutrient co-limitation in future model studies.

Although our model is unusual in representing both
Trichodesmium and Crocosphaera responses to climate change, they are not the only diazotrophs in the ocean, and molecular techniques have identified a wide diversity of diazotrophic organisms co-existing in the ocean including both autotrophic and heterotrophic diazotrophs (Zehr & Capone, 2020 however, other processes such as photosynthesis and respiration involve enzymes that may respond differently to warming compared to those associated with N 2 fixation. In the studies that measured the EUEs of N 2 fixation, the carbon fixation EUEs were also measured and show slight differences compared with the N 2 fixation EUEs, which may lead to regional shifts in the diazotrophy niche for both Trichodesmium and Crocosphaera (Jiang et al., 2018;Yang et al., 2021). A similar approach using EUEs could be applied to other PFTs within the model. The thermal windows of non-diazotroph phytoplankton have been found to range from temperatures as cold as <5°C to warmer temperatures of 35°C and each species has a specific thermal optimum for growth (Boyd et al., 2013). This implies that the thermal performance of phytoplankton will be highly variable throughout the ocean with each phytoplankton experiencing different levels of thermal stress based upon their adaptation to temperature. Warming will therefore impact upon growth rate and EUEs of different phytoplankton, shaping patterns of nutrient limitation and ultimately defining their environmental niche. The response of different PFT to warming will alter patterns of resource availability and competition influencing ocean biogeochemistry. Therefore, to gain a more complete understanding of how warming will impact ocean biogeochemistry it is essential to include temperature adjusted EUEs for both other PFTs and for other processes such as carbon fixation.

| CON CLUS IONS
In this study, we have developed a new state-of-the-art explicit diazotroph model for PISCES QUOTA to investigate how diazotroph thermal fitness shapes patterns of marine N 2 fixation. The model can switch between two prevalent marine diazotrophs, Trichodesmium and Crocosphaera, and uses observed thermal performance curves of growth and N 2 fixation EUEs to represent the thermal fitness of both diazotrophs. This enables the integrated effects of warming on both the niche and physiology of both diazotrophs to be assessed and identify how this shapes the response marine N 2 fixation to climate change. We have shown that both diazotroph-specific thermal performance curves and EUEs impact the response of N 2 fixation to climate change. N 2 fixation is predicted to decrease globally for both diazotrophs, but regional differences occur particularly in the Pacific Ocean, which acts to shape the global response of Trichodesmium and Crocosphaera to climate change and the knock-on effects for NPP. Both diazotrophs exhibit broadly similar spatial patterns of N 2 fixation with increases in the high latitudes driven by thermal expansion and decreases in the low latitudes due to thermal exclusion. The integrated impact of temperature on marine diazotrophy explained 95%-97% of the N 2 fixation climate change signal, with two groups of drivers emerging, those associated with a change in the diazotroph's niche and those associated with a change in diazotroph physiology. Decreases in N 2 fixation were dominated by a change in the diazotroph niche, while increases were driven by a combination of both a changing niche and changing physiology. With temperatures rising diazotroph thermal stress will increase, and it is predicted that by the end of the century, the area of thermal stress will double. This implies that if diazotrophs cannot adapt rapidly enough to increasing temperatures they may be excluded from large regions of the low latitude ocean. Overall, we have performed a holistic consideration of the impact of warming on diazotrophy, highlighting that the effects of temperature on diazotroph thermal fitness will interact to shape the response of N 2 fixation to climate change, which will have important implications for marine primary productivity in the future.

CO N FLI C T O F I NTE R E S T
The authors have declared no conflict of interest.