By commenting, you agree to the Prohibited Content Policy
By commenting, you agree to the Prohibited Content Policy
News
See whats happening in Marketing & Advertising right now
Exclusive
Read and get insights from specially curated unique stories from editorial
Leaders Speak
Business leaders sharing their insights
Events
Explore and discuss challenges & trends in India’s leading B2B events
Awards
Recognise work that not only stood out but was also purposeful
Webinars
Join leaders & experts for roundtables, conferences, panels and discussions
Subscribe to our Daily Newsletter
By continuing you agree to our Privacy Policy & Terms & Conditions
Advertise With Us
We have various options to advertise with us including Events, Advertorials, Banners, Mailers, etc.
Download ETBrandEquity App
Save your favourite articles with seamless reading experience
Get updates on your preferred social platform
Follow us for the latest news, insider access to events and more. About Us Contact Us Newsletters
Our #1 AI Stock Pick is on a steep discount – 29$ instead of 99$! Click here to access exclusive investment research and ad free browsing! Our #1 AI Stock Pick is on a steep discount – 29$ instead of 99$! Click here to access exclusive research! We recently compiled a list of the 10 AI Stocks on Analysts’ Radars.In this article, we are going to take a look at where Mobileye Global Inc. (NASDAQ:MBLY) stands against the other AI stocks. Despite DeepSeek and Qwen 2.5 disrupting the global tech sector, investors welcomed the impressive Q4 AI revenues from leading US tech firms like International Business Machines Corp. (NYSE:IBM). The company’s stock price reached an all-time high of $261 on January 30th as it reported GenAI’s book of business surpassing $5 billion since the segment’s inception. “Three years ago, we laid out a vision for a faster-growing, more-profitable IBM. I’m proud of the work the IBM team has done to meet or exceed our commitments. With our focused strategy, enhanced portfolio, and culture of innovation, we’re well-positioned for 2025 and beyond and expect revenue growth of at least five percent and free cash flow of about $13.5 billion this year,” said IBM CEO Arvind Krishna. While the S&P 500 sharply fell from its record high of 6,118 on January 23rd to 6,012 on January 27th, the benchmark index appears to have momentum in the past few days, climbing to 6,071 on January 30th. However, several concerns remain about the adoption of AI, in terms of security and job losses. Code.org Founder and CEO Hadi Partovi said at the World Economic Forum that the risk isn’t people losing their jobs to AI, but losing work “to somebody else who knows how to use AI. That is going to be a much greater displacement.” “It’s not that the worker gets replaced by just a robot or a machine in most cases, especially for desk jobs, it’s that some better educated or more modernly educated worker can do that job because they can be twice as productive or three times as productive,” he explained. “The imperative is to teach how AI tools work to every citizen, and especially to our young people.” We selected AI stocks by reviewing news articles, stock analysis, and press releases. We listed the stocks in ascending order of their hedge fund sentiment taken from Insider Monkey’s database of 900 hedge funds. Why are we interested in the stocks that hedge funds pile into? The reason is simple: our research has shown that we can outperform the market by imitating the top stock picks of the best hedge funds. Our quarterly newsletter’s strategy selects 14 small-cap and large-cap stocks every quarter and has returned 275% since May 2014, beating its benchmark by 150 percentage points (see more details here). A driverless vehicle navigating city traffic, equipped with ADAS and safety features. Mobileye Global Inc. (NASDAQ:MBLY) develops and sells hardware and software solutions for driver-assistance systems (ADAS) and autonomous driving technologies to enhance mobility infrastructure and enable hands-off driving, facilitated through a variety of leading safety features like real-time detection of road users, geometry, and marking. On January 31st, Wells Fargo analyst Aaron Rakers trimmed Mobileye Global Inc.’s (NASDAQ:MBLY) stock price target to $18 from $25 but retained an “Overweight” rating. The brokerage is incrementally negative on the company due to the weak/conservative 2025 guide. Analysts say it is left to question the degree to which outlook reflects macro uncertainty versus competitive pressure. Wells Fargo added that Mobileye Global Inc.’s (NASDAQ:MBLY) remains a 2026-plus story. Overall MBLY ranks 7th on our list of the AI stocks on analysts’ radars. While we acknowledge the potential of MBLY as an investment, our conviction lies in the belief that AI stocks hold greater promise for delivering higher returns and doing so within a shorter timeframe. If you are looking for an AI stock that is more promising than MBLY but that trades at less than 5 times its earnings, check out our report about the cheapest AI stock. READ NEXT: 20 Best AI Stock To Buy Now and Complete List of All AI Companies Under $2 Billion Market Cap. Disclosure: None. This article is originally published at Insider Monkey. Yahoo Finance Returns since its inception in May 2014 (through November 14th, 2024) Warren Buffett Berkshire Hathaway $293,447,417,000 David Einhorn Greenlight Capital $1,491,303,000 George Soros Soros Fund Management $5,416,602,000 Jim Simons Renaissance Technologies $77,426,184,000 Leon Cooperman Omega Advisors $1,886,381,000 Carl Icahn Icahn Capital LP $22,521,664,000 Steve Cohen Point72 Asset Management $22,767,998,000 John Paulson Paulson & Co $3,510,256,000 David Tepper Appaloosa Management LP $4,198,712,000 Paul Tudor Jones Tudor Investment Corp $6,160,740,000 Artificial intelligence is the greatest investment opportunity of our lifetime. The time to invest in groundbreaking AI is now, and this stock is a steal! The whispers are turning into roars. Artificial intelligence isn’t science fiction anymore. It’s the revolution reshaping every industry on the planet. From driverless cars to medical breakthroughs, AI is on the cusp of a global explosion, and savvy investors stand to reap the rewards. Here’s why this is the prime moment to jump on the AI bandwagon: Exponential Growth on the Horizon: Forget linear growth – AI is poised for a hockey stick trajectory. Imagine every sector, from healthcare to finance, infused with superhuman intelligence. We’re talking disease prediction, hyper-personalized marketing, and automated logistics that streamline everything. This isn’t a maybe – it’s an inevitability. Early investors will be the ones positioned to ride the wave of this technological tsunami. Ground Floor Opportunity: Remember the early days of the internet? Those who saw the potential of tech giants back then are sitting pretty today. AI is at a similar inflection point. We’re not talking about established players – we’re talking about nimble startups with groundbreaking ideas and the potential to become the next Google or Amazon. This is your chance to get in before the rockets take off! Disruption is the New Name of the Game: Let’s face it, complacency breeds stagnation. AI is the ultimate disruptor, and it’s shaking the foundations of traditional industries. The companies that embrace AI will thrive, while the dinosaurs clinging to outdated methods will be left in the dust. As an investor, you want to be on the side of the winners, and AI is the winning ticket. The Talent Pool is Overflowing: The world’s brightest minds are flocking to AI. From computer scientists to mathematicians, the next generation of innovators is pouring its energy into this field. This influx of talent guarantees a constant stream of groundbreaking ideas and rapid advancements. By investing in AI, you’re essentially backing the future. The future is powered by artificial intelligence, and the time to invest is NOW. Don’t be a spectator in this technological revolution. Dive into the AI gold rush and watch your portfolio soar alongside the brightest minds of our generation. This isn’t just about making money – it’s about being part of the future. So, buckle up and get ready for the ride of your investment life! Act Now and Unlock a Potential 10,000% Return: This AI Stock is a Diamond in the Rough (But Our Help is Key!) The AI revolution is upon us, and savvy investors stand to make a fortune. But with so many choices, how do you find the hidden gem – the company poised for explosive growth? That’s where our expertise comes in. We’ve got the answer, but there’s a twist… Imagine an AI company so groundbreaking, so far ahead of the curve, that even if its stock price quadrupled today, it would still be considered ridiculously cheap. That’s the potential you’re looking at. This isn’t just about a decent return – we’re talking about a 10,000% gain over the next decade! Our research team has identified a hidden gem – an AI company with cutting-edge technology, massive potential, and a current stock price that screams opportunity. This company boasts the most advanced technology in the AI sector, putting them leagues ahead of competitors. It’s like having a race car on a go-kart track. They have a strong possibility of cornering entire markets, becoming the undisputed leader in their field. Here’s the catch (it’s a good one): To uncover this sleeping giant, you’ll need our exclusive intel. We want to make sure none of our valued readers miss out on this groundbreaking opportunity! That’s why we’re slashing the price of our Premium Readership Newsletter by a whopping 70%. For a ridiculously low price of just $29.99, you can unlock a year’s worth of in-depth investment research and exclusive insights – that’s less than a single restaurant meal! Here’s why this is a deal you can’t afford to pass up: • Access to our Detailed Report on this Game-Changing AI Stock: Our in-depth report dives deep into our #1 AI stock’s groundbreaking technology and massive growth potential. • 11 New Issues of Our Premium Readership Newsletter: You will also receive 11 new issues and at least one new stock pick per month from our monthly newsletter’s portfolio over the next 12 months. These stocks are handpicked by our research director, Dr. Inan Dogan. • One free upcoming issue of our 70+ page Quarterly Newsletter: A value of $149 • Bonus Reports: Premium access to members-only fund manager video interviews • Ad-Free Browsing: Enjoy a year of investment research free from distracting banner and pop-up ads, allowing you to focus on uncovering the next big opportunity. • 30-Day Money-Back Guarantee: If you’re not absolutely satisfied with our service, we’ll provide a full refund within 30 days, no questions asked.
Space is Limited! Only 1000 spots are available for this exclusive offer. Don’t let this chance slip away – subscribe to our Premium Readership Newsletter today and unlock the potential for a life-changing investment. Here’s what to do next: 1. Head over to our website and subscribe to our Premium Readership Newsletter for just $29.99. 2. Enjoy a year of ad-free browsing, exclusive access to our in-depth report on the revolutionary AI company, and the upcoming issues of our Premium Readership Newsletter over the next 12 months. 3. Sit back, relax, and know that you’re backed by our ironclad 30-day money-back guarantee. Don’t miss out on this incredible opportunity! Subscribe now and take control of your AI investment future!
No worries about auto-renewals! Our 30-Day Money-Back Guarantee applies whether you’re joining us for the first time or renewing your subscription a year later! I work for one of the largest independent financial publishers in the world – representing over 1 million people in 148 countries. We’re independently funding today’s broadcast to address something on the mind of every investor in America right now… Should I put my money in Artificial Intelligence? Here to answer that for us… and give away his No. 1 free AI recommendation… is 50-year Wall Street titan, Marc Chaikin. Marc’s been a trader, stockbroker, and analyst. He was the head of the options department at a major brokerage firm and is a sought-after expert for CNBC, Fox Business, Barron’s, and Yahoo! Finance… But what Marc’s most known for is his award-winning stock-rating system. Which determines whether a stock could shoot sky-high in the next three to six months… or come crashing down. That’s why Marc’s work appears in every Bloomberg and Reuters terminal on the planet… And is still used by hundreds of banks, hedge funds, and brokerages to track the billions of dollars flowing in and out of stocks each day. He’s used this system to survive nine bear markets… create three new indices for the Nasdaq… and even predict the brutal bear market of 2022, 90 days in advance. Click to continue reading… Get our editor’s daily picks straight in your inbox!
For every new rule, President Donald Trump plans to kill 10 old ones.
That’s the thrust of the president’s latest executive order, signed Friday, called “Unleashing Prosperity Through Deregulation.” Trump, ever the businessman, has often complained that government regulations stifle economic development.
The new order requires federal agencies to repeal “at least 10 existing rules, regulations or guidance documents” for every new rule or regulation created.
“It is the policy of my administration to significantly reduce the private expenditures required to comply with federal regulations to secure America’s economic prosperity and national security,” the order states.
The order also requires agency heads ensure the cost of any new regulations in fiscal 2025 “shall be significantly less than zero.”
It wasn’t immediately clear how federal agencies would go about choosing which 10 regulations to eliminate, or how their wider cost will be determined.
Trump issued a similar executive order in 2017, requiring government agencies to eliminate two rules for every new one created. Former President Joe Biden revoked the order in 2021, on his first day in office.
The order directs the Office of Management and Budget director to set guidance. The Senate Budget Committee voted on Thursday to advance Russell Vought, Trump’s pick to lead OMB, for a floor vote. Committee Democrats boycotted the vote. Vought was an author of the controversial Project 2025 agenda for a Republican administration.
DEL RIO, Texas — Kristi Noem will visit the Texas-Mexico border this weekend for the first time since being confirmed as the homeland security secretary under President Donald Trump, according to federal officials.
Noem, who served as South Dakota’s governor from 2019 to 2025, was confirmed Jan. 25 and will visit Del Rio on Saturday to meet with Customs and Border Protection, review operations and tour the border that’s been a focus of Trump ever since he entered the political arena in his first term.
Her visit comes as the Pentagon is also preparing to send around 500 additional service members, mostly from the U.S. Army’s 10th Mountain Division, to help secure the border, according to CBS News.
Noem, 53, is the eighth secretary of homeland security. She’s tasked with delivering on Trump’s favorite issue of border security. She has so far pledged to faithfully execute the president’s orders and copied his talk of an “invasion” at the U.S. border with Mexico.
She joined other Republican governors who sent National Guard troops to Texas to assist in Operation Lone Star. Her decision was especially criticized because she accepted a $1 million donation from a Tennessee billionaire to cover some of the deployment cost.
She said she opted to send National Guard troops “because of the invasion,” adding that “it is a war zone down there.”
Trump’s return to the White House has emboldened state leaders to ask for more help in security the border, something Gov. Greg Abbott repeatedly blasted the Biden administration for, in his eyes, failing to do. In the days after President Trump’s inauguration, Abbott asked Congress to reimburse Texas billions of dollars for Operation Lone Star, saying the money would cover the costs of stepping up border security, apprehending undocumented immigrants, staffing the border and transporting migrants to their countries of origin.
Next up in 5
Example video title will go here for this video
Next up in 5
Example video title will go here for this video
Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.
Advertisement Communications Earth & Environmentvolume 6, Article number: 75 (2025) Cite this article Metrics details
Marine net primary production supports critical ecosystem services and the carbon cycle. However, the lack of consensus in the direction and magnitude of projected change in net primary production from models undermines efforts to assess climate impacts on marine ecosystems with confidence. Here we use contemporary remote sensing net primary production trends (1998–2023) from six remote sensing algorithms to discriminate amongst fifteen divergent model projections. A model ranking scheme, based on the similarity of linear responses of net primary production to changes in sea surface temperature, chlorophyll-a and the mixed layer, finds that future declines in net primary production are more likely than presently predicted. Even the best ranking models still underestimate the sensitivity of declines in net primary production to ocean warming, suggesting shortcomings remain. Reproducing this greater temperature sensitivity may lead to even larger declines in future net primary production than presently considered for impact assessment.
Marine net primary production (NPP) by phytoplankton sustains biodiversity and is essential to global ocean ecosystems, but its future is uncertain1. Despite its importance for the assessment of climate change impacts on the marine system2, there is currently a lack of consensus regarding the sign of predicted change in NPP on regional and global scales under high emissions scenarios of the sixth coupled model intercomparison project (CMIP6)3,4. This leads to divergent global trends over the contemporary and future periods when comparing individual models (Fig. 1a) and a negligible projected change with a large standard deviation (−0.76 ± 3.44 Pg C year−1 by 21003; Fig. 1b) when these divergent trends are averaged for the multi-model ensemble mean. Importantly, this NPP trend uncertainty has increased by more than 50% since the previous IPCC assessment cycle3. Furthermore, upper trophic level models that assess future responses of fisheries typically subsample NPP projections from the ‘high’ and ‘low’ extremes of available projections5,6. This highlights the urgent need to interrogate Earth system models and discriminate amongst them in order to deliver increased confidence in projections of NPP in response to climate change. a Area-weighted mean-normalised net primary production (NPP) annual trends (% year−1) calculated using ordinary least squares for the historical (1850–2014), contemporary (1998–2023) and future (2015–2100) periods for the CMIP6 Earth system model ensemble. b Area-weighted ΔNPP (Pg C year−1), calculated as the difference between the end of the historical period (1995–2014) and the end of the century (2081–2100), for each of the Earth system models in the CMIP6 ensemble. Both panels are sorted by ΔNPP from low to high values.
Emergent relationships between changes in remote sensing estimates of NPP and concomitant changes in ocean environmental conditions over the contemporary period can provide global constraints for Earth system models. Emergent constraints have been used to refine assessments of NPP trends in the tropical Pacific7,8 and ocean carbon uptake in the Arctic and Southern Ocean9,10 based on single parameter assessments but have yet to be exploited on a global scale or with multi-parameter relationships. Trends in marine NPP estimated from remote sensing however also vary considerably depending on the time period, algorithm implemented, and data product being used11,12,13,14. Some of the sensitivities to time period and data product are addressed by the generation of a coherent multi-sensor satellite record spanning 1998–2023 that merges all available single-sensor satellite missions with substantially reduced inter-sensor biases15. Nonetheless, intrinsic differences in remote sensing trends are still apparent in the range of algorithms available for quantifying NPP rates. Here we focus on six algorithms including: (1) the ‘vertically generalised production model’s (Eppley-VGPM16 and Behrenfeld-VGPM17), which define phytoplankton growth as a function of chlorophyll-a, light and temperature, the difference being that Eppley-VGPM is an exponential function of temperature, while Behrenfeld-VGPM is a 4th order polynomial; (2) the ‘carbon-based production models (Behrenfeld-CbPM18 and Westberry-CbPM19), which incorporate particulate backscatter as a proxy for phytoplankton carbon but differ in that Westberry-CbPM is both depth and wavelength resolved whilst Behrenfeld-CbPM is not; (3) the ‘absorption-based production model’ (Lee-AbPM20), which defines NPP as a function of phytoplankton absorption rather than chlorophyll; and (4) the ‘carbon, absorption, and fluorescence euphotic’ resolving model (Silsbe-CAFE21), which integrates the learning from all the above algorithms to define NPP as a function of energy absorption and efficiency (for more details please see Methods).
In this work, we rank fifteen CMIP6 Earth system models according to their ability to capture the emergent contemporary relationships observed between NPP and environmental variables (sea surface temperature, chlorophyll-a and the mixed layer depth) in the 6 remote sensing algorithms. Five of the remote sensing algorithms (Behrenfeld-VGPM, Behrenfeld-CbPM, Westberry-CbPM, Lee-AbPM and Silsbe-CAFE) concur that climate models projecting NPP declines rank higher, whilst the remaining algorithm (Eppley-VGPM) rank models that project positive NPP trends higher. These results suggest that future NPP decline is more likely than not, and this decline is currently underestimated by even the best ranked CMIP6 models, which predict the most intense NPP declines.
Remote sensing derived NPP trends are potentially susceptible to extreme climate events that align with either the start or end of the time series, e.g., El Niño-Southern Oscillation, which could act to increase or decrease estimated rates of change. This is more likely to impact the VGPM algorithms, which parameterise NPP as a function of temperature. To account for this, we performed a Monte Carlo jackknife resampling of ~80% of the time series (i.e., 20 of the 26 years – representing 7 different assessments) for all the analyses presented here. NPP trends from remote sensing are predominantly negative across the algorithms, apart from the two VGPM approaches which show positive trends, most notably at higher latitudes (>40°) (Fig. 2a, b). We note however that when globally averaged, the VGPM trends have standard deviations that are larger than the mean, which reduces confidence in the NPP trends that these two algorithms produce (−0.004 ± 0.10% year−1 & 0.04 ± 0.07% year−1 respectively; Fig. 2a, b). The remaining four algorithms all produce globally averaged trends of declining NPP that range from −0.27% year−1 to −1.45% year−1 (Fig. 2c–f). The predominantly negative global trends from these four algorithms are more robust than the VGPM algorithms given that the standard deviations are always lower than the mean (0.14–0.23% year−1). When trends across all algorithms are grouped into ocean provinces (Supplementary Fig. S1a22), the typical decline in remote sensing derived NPP across biomes and algorithms is emphasised (Supplementary Fig. S1b), apart from some high latitude regions in the Arctic and the Southern Ocean, where only the VGPM algorithms suggest an increase in NPP.
Global distribution of the mean jackknife resampled trends (1998–2023) of annual mean net primary production (NPP; Gg C year−1) from a Eppley-VGPM, b Behrenfeld-VGPM, c Behrenfeld-CbPM, d Westberry-CbPM, e Lee-AbPM and f Silsbe-CAFE. Inset text reports the area weighted mean ±1σ jackknife resampled (80% of the 26-year time series) NPP trend (% year−1) as displayed in Supplementary Fig. S1b.
Trends in NPP occur in response to concomitant modifications of the ocean environment that span ‘bottom up’ factors like resource limitation to ‘top down’ controls such as grazing. Contemporary evidence already exists for climate-driven adjustments in sea surface temperatures23 and the nutrient and light environment from altered stratification and mixed layer depths24. Such adjustments will impact phytoplankton physiology and their photosynthetic capacity, which is reflected in global trends in chlorophyll-a25. However, we lack insight into the relative roles played by different factors that shape contemporary trends in NPP across the different algorithms. To statistically assess what drives local trends in NPP, we use multiple linear regressions (MLR) that account for unequal variance and autocorrelation. We used MLRs to link contemporary trends in NPP to a suite of environmental and biological drivers across all algorithms and jackknife trend assessments (see Methods). These drivers are trends in annual mean sea surface temperature (SST; where warming increases phytoplankton metabolic rates and may retard nutrient supply due to greater ocean stratification), annual mean chlorophyll-a concentration (CHL; which reflects phytoplankton biomass and physiology), and annual mean mixed layer depth (MLD; which impacts adjustments in both light and nutrient supply). These variables are all available from satellite remote sensing or compiled data products and importantly are also accessible as standard outputs from CMIP6 Earth system models. A sensitivity analysis of the entire time series showed that including all three drivers into the MLR analysis led to an increase in the global coverage of significant (p < 0.05) regressions and a higher median adjusted R2 across all six remote sensing algorithms, as opposed to using either SST or CHL alone, or a combination of SST and CHL as predictors (Supplementary Fig. S2). The MLR applied to the Eppley-VGPM and Behrenfeld-VGPM NPP trends had the highest globally averaged mean adjusted R2 values (Supplementary Fig. S3a, b), decreasing (most notably at high latitudes) for Behrenfeld-CbPM and Westberry-CbPM (Supplementary Fig. S3c, d), whilst Lee-AbPM and Silsbe-CAFE had a more even global spread and lower global averages (Supplementary Fig. S3e, f). The higher global mean R2 values for the VGPM algorithms is perhaps not surprising as the MLR is constructed using two of the three algorithm inputs, SST and CHL, with photosynthetically active radiation the remaining input variable.
The MLR coefficients associated with each driver show a reduction in amplitude, roughly halving in strength from SST to CHL and again from CHL to MLD (Fig. 3a–c). This indicates that trends in SST and CHL are the most important predictors of trends in NPP, whilst MLD plays only a minor role. Zonally averaged, the distribution of MLR coefficients for SST reveals regional coherence at high latitudes across all algorithms (Fig. 3a), whereas at mid to low latitudes the Eppley-VGPM algorithm behaves anomalously by displaying a positive relationship between SST and NPP, while all other algorithms display negative SST coefficients (accentuated at equatorial latitudes). Negative coefficients are indicative of declining NPP as the surface ocean warms, potentially reflecting the role of nutrient limitation from a reduced surface reservoir as stratification intensifies. The atypical positive coefficients from Eppley-VGPM on the other hand may reflect a metabolic response that favours increased growth rates under warmer conditions26. The CHL coefficient is consistently positive across all six remote sensing NPP algorithms, corresponding to the regulation of NPP by phytoplankton standing stocks (Fig. 3b). Whilst both the VGPM and Silsbe-CAFE algorithms display minimal latitudinal gradients in the magnitude of the CHL coefficient, the three other algorithms display higher equatorward coefficients that decrease towards the poles. MLD (Fig. 3c) shows a broadly similar pattern across all algorithms with regression coefficients at low latitudes that are close to zero, while negative coefficients are more prominent at higher latitudes (exaggerated in the two CbPM algorithms relative to the others). Consistent across all algorithms the spatial standard deviations (Fig. 3a–c) were always substantially greater than the standard deviations of the jackknife assessments (Supplementary Fig. S4), confirming that regional variability plays a more significant role in the proposed ranking scheme.
Zonal averages ± standard deviations of the multiple linear regression coefficients for a, d sea surface temperature (SST), b, e chlorophyll-a concentrations (CHL) and c, f mixed layer depth (MLD) for the a–c Eppley-VGPM, Behrenfeld-VGPM, Behrenfeld-CbPM, Westberry-CbPM, Lee-AbPM and Silsbe-CAFE NPP algorithms and d–f the ensemble of CMIP6 Earth system models. Only pixels/grid points where the multiple linear regression analysis was significant are included in the zonal averages. The shaded region in panels d–f represents the range of coefficients as estimated from the remote sensing algorithms zonal averages (panels a–c).
Using an ensemble of fifteen Earth system models from CMIP6 we evaluate modelled trends in NPP (Fig. 1) in relation to the same set of drivers used in the remote sensing analysis to develop a ‘process based’ model ranking scheme. MLR analysis between modelled NPP and the three associated drivers (SST, CHL and MLD) were repeated globally for each Earth system model over the historical period (1850–2014) but without jackknife resampling. This approach is similar to previous model assessment exercises that focussed on a single driver variable for a specific region (e.g. refs. 7,9,10), whereas our assessment is global and incorporates three driver variables. Overall, the MLR analysis applied to the Earth system model data produced a similar range in skill across models as for the remote sensing algorithms (Supplementary Fig. S5).
Both the magnitudes and spatial distribution of the MLR coefficients across SST, CHL and MLD for each Earth system model reveal stark differences, relative to the remote sensing assessment (Fig. 3d–f). However, the general decline in their relative contribution to NPP trends from SST to CHL and lastly MLD largely remains, albeit to a lesser extent than the remote sensing algorithms. Whilst all CMIP6 models appear to place similar weights on SST at high latitudes, that are within the range of what was found for the six remote sensing NPP algorithms, at mid to low latitudes they vary markedly and are generally underestimated relative to remote sensing. This implies that CMIP6 NPP trends in the tropics and subtropics are not as sensitive to warming as remote sensing NPP trends to changes in SST (Fig. 3d). In addition, the CMIP6 models also tend to reflect a stronger positive influence of CHL on NPP, with coefficients that are almost double those found in the remote sensing algorithms, which suggests that some models may be oversensitive to CHL trends, most notably at high latitudes (Fig. 3e). MLD coefficients show a large discrepancy in their zonal distribution relative to remote sensing (Fig. 3f) with coefficients that are broadly positive and more similar to each other at high latitudes but larger and more diverse (spanning both positive and negative coefficients) at lower latitudes.
To constrain NPP projections we develop a regionally informed model ranking scheme by comparing the distribution of the MLR coefficients from each model with those from the jackknife analysis of each remote sensing algorithm. Models with a more similar distribution in their respective MLR coefficients (relative to a specific remote sensing algorithm) will rank higher than models with a very different distribution in their MLR coefficient values. This ranking is based on the dimensionless Earth mover’s distance (EMD) metric27, which quantifies the effort required to transform the distribution of the Earth system model MLR coefficients to match those obtained from each of the six remote sensing NPP algorithms. A low EMD value indicates that the Earth system model MLR coefficients closely match, i.e. are in good agreement, to those of the remote sensing algorithms. EMD metrics were calculated per jackknife assessment and per biome22, with MLR coefficients restricted using the interquartile range fence test (see Methods) to remove extreme outliers in each biome. The EMD metrics, for SST, CHL and MLD were then weighted according to each biome’s proportion of the globe and finally globally averaged to generate a single EMD value per MLR coefficient per model for each of the six remote sensing algorithm jackknife assessments (Supplementary Fig. S6). EMD metrics were generally higher for SST, followed by CHL and then MLD. The higher EMD values for SST imply a large disparity in the weighting of the MLR coefficients between remote sensing and models, which implies an inaccurate representation of the relationship between warming and NPP in the models.
Next, we average the EMD values across all three variables, SST, CHL and MLD, to generate a single EMD mean and standard deviation for each Earth system model per remote sensing algorithm jackknife assessment. For each remote sensing algorithm, we then rank the CMIP6 models using Z-scores that incorporate both the EMD mean and standard deviation. The Z-score is defined as the distance of a value to the group mean, such that high Z-scores indicate values that are atypical and much larger than the mean and vice versa. A low Z score thus indicates that the NPP driver relationship in the Earth system model more closely matched that of the remote sensing algorithm. We then combine both Z-scores (from the EMD mean and standard deviation) using equal weighting (i.e. we averaged the Z-scores), before sorting the combined Z-scores from smallest to largest to rank each Earth system model’s relative performance (Fig. 4). However, since each remote sensing algorithm is made up of 7 jackknife assessments, each of the 15 Earth system models is independently ranked 7 times (for each remote sensing algorithm) (Supplementary Fig. S7). Model ranking results are presented with respect to the mean ΔNPP (Fig. 4), which is averaged across the 7 model rankings. When the same model is ranked in the same position for all 7 of the jackknife assessments there is no standard deviation in ΔNPP, whereas a standard deviation reflects instances where ΔNPP was averaged across more than one model with the same ranking. Five algorithms concur that climate models projecting greater NPP declines rank higher, whilst the remaining algorithm (Eppley-VGPM) ranks models that project slightly positive NPP trends higher (Fig. 4).
Bar plots of mean ± standard deviation jackknife resampled ranked Earth system model ΔNPP (Pg C year−1) for a Eppley-VGPM, b Behrenfeld-VGPM, c Behrenfeld-CbPM, d Westberry-CbPM, e Lee-AbPM and f Silsbe-CAFE NPP algorithms. All bars are coloured by the mean Z-score across the jackknife resampling exercise. Please note that the absence of an errorbar is indicative of the same model being ranked in the same position for all 7 of the jackknife assessments.
The ranking scheme results in a split between five algorithms that rank models that have negative NPP projections higher and one (Eppley-VGPM) that favours models with positive NPP projections. Focussing first on the two VGPM algorithms, we find that trends in NPP from Eppley-VGPM and Behrenfeld-VGPM are primarily driven by SST and CHL (Supplementary Fig. S8a, b). The Eppley-VGPM algorithm parameterises phytoplankton growth using an exponential function of temperature, explaining why it ranks positive NPP projections higher (Fig. 4a). However, only the Behrenfeld-VGPM algorithm implements a penalty on growth when temperatures increase beyond a certain threshold, consistent with its favouring of models with negative NPP projections (Fig. 4b). The trends in NPP from the CbPM algorithms are predominantly driven by changes in CHL (Supplementary Fig. 8c, d), rather than changes in particulate backscatter, similar to previous studies28. Differences in the ranking between the two CbPM algorithms are likely because the Behrenfeld-CbPM algorithm does not account for changing light properties through the water column (Fig. 4b), whereas Westberry-CbPM does (Fig. 4c). Finally, NPP trends in the Lee-AbPM and Silsbe-CAFE algorithms, which more consistently rank negative projections highest (Fig. 4e, f), are driven by trends in phytoplankton absorption (Supplementary Fig. S8e, f). In so doing, when determining phytoplankton productivity, these algorithms respond to the efficiency of light absorption, rather than the absolute quantity of photosynthetic pigments, which makes them better suited to capturing NPP responses to environmental variability29.
During Primary Production Algorithm Round Robin exercises30,31,32 no single algorithm has been found to perform best at all times and locations. However, there is a general reduction in the root mean square difference between remote sensing NPP estimates and direct field measurements for the Lee-AbPM and Silsbe-CAFE algorithms (relative to the VGPM and CbPM algorithms), suggesting that they perform best overall (Supplementary Fig. S921,32,33,34). Indeed, more recent studies that applied the Behrenfeld-VGPM, Westberry-CbPM and Lee-AbPM algorithms to OC-CCI data report similar findings where Lee-AbPM has the lowest RMSE34. In addition, the jackknife trend analysis we conducted on the time series (Supplementary Fig. S10) demonstrates that both the Eppley-VGPM and Behrenfeld-VGPM algorithms are strongly sensitive to the start or end dates of the time series (Supplementary Fig. S10a–d), with high coefficients of variation and even a switch in the dominant direction of NPP trends across the assessments. Although both CbPM algorithms had similarly high coefficients of variation across the globe (relative to the VGPM algorithms), they remain dominated by negative trends across all assessments, with some evidence of an increase in the magnitude of negative trends and the number of positive trends in response to a change in the start and end dates (Supplementary Fig. S10e–h). The Lee-AbPM and Silsbe-CAFE algorithms displayed the most robust response in NPP trends to the jackknife assessments, with much lower coefficients of variation and no tangible increase in the number of positive trends (with only a slight increase in the magnitude of negative trends, Supplementary Fig. S10i–l). Those areas of the globe that display relatively higher coefficients of variation (e.g. the Southern Ocean) thus represent regions with reduced confidence in the magnitude of the predominantly negative trends, but not in their direction. Overall, this indicates that there are larger uncertainties for global NPP trends from the VPGM and CbPM algorithms, relative to the trends estimated from Lee-AbPM and Silsbe-CAFE. Together these points of consideration around NPP algorithm validation and trend sensitivity to the jackknife assessments suggest that the Lee-AbPM and Silsbe-CAFE algorithms are the most robust and therefore best suited for the implementation of the model ranking scheme. Consequently, these results support a greater likelihood of global NPP declines into the future.
Using a multivariate model ranking scheme, based on the relationship between contemporary NPP and its associated drivers, we are able to discriminate amongst the Earth system model ensemble towards models that project future NPP declines. Despite advancements in model ranking through the application of our scheme, modelled trends of NPP over the contemporary period remain almost an order of magnitude too low (maximum range of ±0.2%; Fig. 1a) relative to remote sensing estimates (maximum range of −1.5%; Supplementary Fig. S1b). This reflects either too little sensitivity in the response of NPP to driver variability or incorrectly modelled trends in the drivers themselves. SST emerged as the most important driver of NPP, and the magnitude and direction of modelled contemporary trends in SST match the observational trends reasonably well, with a few model dependent regional exceptions (Supplementary Fig. S11a). The observed CHL trend on the other hand, which is the second most important driver of trends in NPP, is not well matched by Earth system models in magnitude, particularly at high latitudes, with some models also projecting regional differences in trend direction (Supplementary Fig. S11b). These differences in trend magnitude and oftentimes direction will translate into substantial differences in the weightings placed on SST and CHL for even the best ranked Earth system models. For MLD, nearly all models project trends that are opposite in direction to the observations (Supplementary Fig. S11c) and tend to underestimate the role of MLD (Fig. 3f) on trends in NPP. That said, MLD plays a relatively small role in governing trends in NPP, such that inaccuracies in a given model’s ability to capture trends in MLD will not strongly impact model performance. Accordingly, an improved reproduction of contemporary trends in NPP from Earth system models suggests NPP needs to become more sensitive to SST increases and less sensitive to CHL increases. Such a model would then become more sensitive to future levels of warming, suggesting that significant NPP declines may occur even under climate scenarios with strong degrees of mitigation that currently project stable NPP4, with important ramifications for the ocean carbon cycle, marine ecosystem change and management.
Remote sensing is a powerful tool for understanding changes in ocean properties over the contemporary period, with multi-decadal records commonly used to assess and constrain Earth system models’ ability to accurately represent spatial and temporal variability in ocean processes. The model ranking scheme applied here for NPP uses multiple metrics on a global scale to constrain trends and goes further than previous region-specific studies based on single driver metrics (e.g. ref. 7). Our model ranking suggests a strong likelihood of negative future projections in NPP, particularly when based on the most robust remote sensing algorithms (Lee-AbPM and Silsbe-CAFE). Furthermore, the magnitude of the observed decline in NPP across those remote sensing algorithms is much larger than predicted change across even the best ranked Earth system models, suggesting a large underestimation of ongoing change in NPP across the global ocean. Future assessments should not only consider the uncertainties inherent to remote sensing algorithms (Supplementary Fig. S8), despite the complexity in deriving them28 but should also expand on the Round Robin intercomparison exercises (Supplementary Fig. S9) as more in situ data becomes available. Furthermore, future model assessments should consider using additional parameters in combination with those proposed here, such as the resource limitation diagnostics in Earth system models (e.g. iron limitation, light limitation etc), which could be used to assess ongoing changes in the Southern Ocean35 and the equatorial Pacific36. Such approaches would deliver greater confidence in the mechanistic representation of NPP in Earth system models necessary to project associated impacts on marine ecosystems and biogeochemical cycles.
Net primary production (NPP; mg C m−2 d−1) was calculated using the following algorithms, the ‘vertically generalised production model’s (Eppley-VGPM16 and Behrenfeld-VGPM17; Eq. (1)), which relies on the relationship between chlorophyll a and temperature derived growth rates; the ‘carbon-based production model’s (Behrenfeld-CbPM18; Eq. (2) and Westberry-CbPM19; Eq. (3)), which uses backscatter derived phytoplankton carbon as a biomass indicator and physiology derived as variability in the chlorophyll a to carbon ratio; the ‘absorption-based production model’ (Lee-AbPM20; Eq. (4)), which does not make any assumptions on either chlorophyll a or backscatter as biomass proxies but relies on absorption characteristics to infer phytoplankton photosynthetic efficiency; and the ‘carbon, absorption, and fluorescence euphotic’ resolving model (Silsbe-CAFE21; Eq. (5)), which derives NPP as a function of energy absorption and efficiency.
where Chl is chlorophyll-a concentration (mg m−3), ({P}_{{opt}}^{B}) is a temperature (°C) based growth function (exponential for Eppley-VGPM and a 4th-order polynomial for Behrenfeld-VGPM), DL is day length (hours), f(PAR) uses PAR (mol photon m−2 d−1) the parameterized light term for the ratio of realised NPP to maximum potential NPP and Zeu is the depth of the euphotic zone (m).
where Cph is phytoplankton carbon (mg C m−3) derived from an empirical relationship with particulate backscatter at 443 nm (bbp (λ443), m−1), μ (μmax) is the growth rate, Chl:Cph is the chlorophyll-a to phytoplankton carbon ratio and Ig is the growth irradiance term.
where ZNO3 is the depth of nitracline (m), defined as the depth at which nitrate + nitrite exceed 0.5 μM, and dz is the depth (m).
where aph(λ443) is phytoplankton specific absorption at 443 nm (m−1), Kd(λ490) is the light attenuation coefficient at 490 nm (m−1) and PAR is the daily available photosynthetic radiation (mol photon m−2 d−1).
where QPAR is energy absorption, ({Phi }_{mu }^{max }) is the efficiency at which the absorbed energy is converted into carbon biomass and Ek is the light saturation parameter. For more details on all equations please refer to their specific publications.
The algorithms were applied to ocean colour remote sensing data from the European Space Agency Ocean Colour Climate Change Initiative (OC-CCI) data product (8-day, version 6.015) from 1998 to 2023, which was regridded to 25 km using bilinear interpolation. Photosynthetically active radiation (PAR; mol photons m−2 d−1) was taken from the merged GLOBColour product (http://globcolour.info) at 25 km 8-day resolution. For VGPM sea surface temperature (SST; °C) was taken from the Group for High Resolution Sea Surface Temperature (GHRSST; https://www.ghrsst.org/), which was regridded to 25 km as above. For CbPM and CAFE the mixed layer depth (MLD; m) was taken from the Hadley EN 4.2.2 gridded temperature and salinity profiles37, which were first regridded to 25 km as above and resampled to 8-days, then converted to density using the Gibbs Seawater TEOS-10 python package and the MLD derived from a density criterion of 0.03 kg m−3 and reference depth of 10 m38. Full explanation of the VGPM, CbPM and CAFE NPP calculations is provided by Ryan-Keogh et al.39, with data publicly available40. For AbPM we used the OC-CCI aph (λ443; m−1) and Kd(λ490) (m−1), in combination with the GLOBColour PAR, with data publicly available here41.
CMIP6 data were obtained from the Earth System Grid Federation data server for the historical (1850–2014) and the high emission SSP5-8.5 (2015–2100) scenarios of the ACCESS (r1i1p1f1)42, CESM2-WACCM (r1i1p1f1)43, CESM2 (r4i1p1f1)43, CMCC-ESM2 (r1i1p1f1)44, CNRM-ESM2-1 (r1i1p1f2)45, CanESM5 (r1i1p2f1)46, EC-Earth3-CC (r1i1p1f1)47, GFDL-ESM4 (r1i1p1f1)48, IPSL-CM6A-LR (r1i1p1f1)49, MPI-ESM1-2-HR (r1i1p1f1)50, MPI-ESM1-2-LR (r1i1p1f1)50, MRI-ESM2-0 (r1i2p1f1)51, NorESM2-LM (r1i1p1f1)52, NorESM2-MM (r1i1p1f1)52 and UK-ESM1-0-LL (r1i1p1f2)53 models. Data variables downloaded include depth integrated NPP (‘intpp’; mol C m−2 d−1), SST (‘tos’; °C), sea surface chlorophyll-a concentration (‘chlos’; kg m−3) and MLD (‘mlotst’; m). All data variables were regridded on a regular 1° × 1° grid using the bilinear interpolation of Climate Data Operators54, and were resampled from a monthly resolution to annual means.
Trends of remote sensing annual mean NPP were calculated by first excluding any pixel whose time series had less than 50% of the data available. Before linear regressions were performed, the data were first tested for a normal distribution using the D’Agostino-Pearson test in the SciPy python package55. If the data were normally distributed, then linear regressions were performed using the Sci-Kit56 Huber-Regressor, where ε, the parameter to control the amount of robustness (i.e. the number of outliers), was set to value of 1.35. This value is to ensure maximum robustness whilst maintaining 95% statistical efficiency57. If a pixel had less than 50% of the time series following outlier removal, then no further tests were performed. If the data were not normally distributed, then linear regressions were performed using the non-parametric Mann–Kendall Test58. This same method was applied to calculate trends in annual mean SST, annual mean chlorophyll-a concentration (CHL) and annual mean MLD (MLD).
For spatially averaged biome annual mean NPP trends, either remote sensing or Earth system models, an ordinary least squares regression was applied to area weighted data normalised to the mean. Area weighting was determined as a function of latitude, such that remote sensing pixels/model grid points at higher latitudes are smaller than remote sensing pixels/model grid points at lower latitudes. ΔNPP (Pg C) for each Earth system model was calculated as the difference between the global averages from the historical reference period (1995–2014) and the SSP5-8.5 scenario reference period (2081–2100). Please note that the remote sensing annual means of high latitudes may be potentially underestimated due to the presence of cloud cover preventing the retrieval of data.
Annual means of NPP, SST, CHL and MLD were first jackknife resampled to 80% of the time series, representing 7 different possible assessments, and then mean-normalised, i.e. the resampled time series was divided by its mean. Multiple linear regressions (MLR) were then performed using the ordinary least squares function from the Statsmodel package59 using a Heteroskedasticity and Autocorrelation Consistent covariance estimator, where the time lags for autocorrelation were calculated following the Newey & West60 rule of thumb, defined in Eq. (6):
where T is the length of the time series, which in this case is 26 years for remote sensing and 165 years for the Earth system models. No MLR was performed for a remote sensing pixel or model grid point if any variable was missing data from any year of the time series or if the variance for any of the drivers was ~0. MLR coefficients for each prospective driver were excluded from further analysis if either the remote sensing pixel or model grid point were not significant (p > 0.05). Comparisons between the observational data products and model coefficients were performed using the Earth mover’s distance (EMD) metric27, also known as the Wasserstein distance in mathematics61 and Mallow’s distance in statistics62, defined here in Eq. (7):
where l1 is the first EMD, u and v are the respective distributions of the MLR coefficients from remote sensing and Earth system models and U and V are the respective cumulative distance functions of u and v. The MLR coefficient values for both the remote sensing and models were restricted using the interquartile range (IQR) fence test, IQR ± IQR×3, to remove any extreme outliers within each ocean biome. The EMDs were calculated on a per biome basis using the biome classification of Fay & McKinley22, with the EMD weighted by the biome’s areal proportion (%) of the global ocean. The EMDs for SST, CHL and MLD were then averaged to generate an EMD mean and standard deviation per Earth system model. To rank the models the Z-score (Eq. (2)), also known as standard score, was calculated using Eq. (8):
where x is either the model’s EMD mean (or standard deviation), μ is the model ensemble mean of either the EMD mean (or standard deviation) and σ is the model ensemble standard deviation of either the EMD mean (or standard deviation). The final Z-scores, determined from both the EMD mean and the EMD standard deviation, were then generated by combining with equal weighting (i.e. the Z-scores were averaged together).
Remote sensing NPP data for Behrenfeld-VGPM, Westberry-CbPM and Silsbe-CAFE are available at: https://doi.org/10.5281/zenodo.7849934. Remote sensing NPP data for Lee-AbPM are available at: https://doi.org/10.5281/zenodo.10014029. Ocean Colour Climate Change Initiative dataset, Version [6.0], European Space Agency, available online at http://www.esa-oceancolour-cci.org/. Ocean colour photosynthetically active radiation data retrieved from: https://www.globcolour.info/. Hadley EN4.2.2 temperature and salinity data were retrieved from: https://www.metoffice.gov.uk/hadobs/en4/download-en4-2-2.html. Sea surface temperatures were retrieved from: https://www.ghrsst.org/. CMIP6 data were obtained from: https://esgf.llnl.gov/. All data used in this study are available at https://zenodo.org/records/14185537.
The code for reproducing all figures and data in the manuscript is available at https://github.com/tjryankeogh/global_npp_trends.
Bindoff, N. L. et al. Changing Ocean, Marine Ecosystems, and Dependent Communities. In IPCC Special Report on the Ocean and Cryosphere in a Changing Climate (eds. Pörtner, H.-O. et al.) 477–587 (Intergovernmental Panel on Climate Change, 2019).
IPCC. Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. (Cambridge University Press, Cambridge, United Kingdom and New York, USA, 2021).
Tagliabue, A. et al. Persistent Uncertainties in Ocean Net Primary Production Climate Change Projections at Regional Scales Raise Challenges for Assessing Impacts on Ecosystem Services. Front. Clim.3, https://doi.org/10.3389/fclim.2021.738224 (2021).
Kwiatkowski, L. et al. Twenty-first century ocean warming, acidification, deoxygenation, and upper-ocean nutrient and primary production decline from CMIP6 model projections. Biogeosciences17, 3439–3470 (2020). ArticleCAS Google Scholar
Lotze, H. K. et al. Global ensemble projections reveal trophic amplification of ocean biomass declines with climate change. Proc. Natl. Acad. Sci.116, 12907–12912 (2019). ArticleCAS Google Scholar
Tittensor, D. P. et al. Next-generation ensemble projections reveal higher climate risks for marine ecosystems. Nat. Clim. Chang11, 973–981 (2021). Article Google Scholar
Kwiatkowski, L. et al. Emergent constraints on projections of declining primary production in the tropical oceans. Nat. Clim. Chang7, 355–358 (2017). ArticleCAS Google Scholar
Tagliabue, A. et al. An iron cycle cascade governs the response of equatorial Pacific ecosystems to climate change. Glob. Chang Biol.26, 6168–6179 (2020). Article Google Scholar
Terhaar, J., Kwiatkowski, L. & Bopp, L. Emergent constraint on Arctic Ocean acidification in the twenty-first century. Nature582, 379–383 (2020). ArticleCAS Google Scholar
Terhaar, J., Frölicher, T. L. & Joos, F. Southern Ocean anthropogenic carbon sink constrained by sea surface salinity. Sci. Adv.7, eabd5964 (2023). Article Google Scholar
Gregg, W. W., Conkright, M. E., Ginoux, P., O’Reilly, J. E. & Casey, N. W. Ocean primary production and climate: Global decadal changes. Geophys. Res. Lett.30, 1809 (2003).
Behrenfeld, M. J. et al. Climate-driven trends in contemporary ocean productivity. Nature444, 752–755 (2006). ArticleCAS Google Scholar
Gregg, W. W. & Rousseaux, C. S. Global ocean primary production trends in the modern ocean color satellite record (1998–2015). Environ. Res. Lett.14, 124011 (2019). ArticleCAS Google Scholar
Kulk, G. et al. Primary Production, an Index of Climate Change in the Ocean: Satellite-Based Estimates over Two Decades. Remote Sens.12, https://doi.org/10.3390/rs12050826 (2020).
Sathyendranath, S. et al. An Ocean-Colour Time Series for Use in Climate Studies: The Experience of the Ocean-Colour Climate Change Initiative (OC-CCI). Sensors19, 4285 (2019). ArticleCAS Google Scholar
Eppley, R. W. Temperature and phytoplankton growth in the sea. Fish. Bull.70, 1063–1085 (1972). Google Scholar
Behrenfeld, M. J. & Falkowski, P. G. Photosynthetic rates derived from satellite-based chlorophyll concentration. Limnol. Oceanogr.42, 1–20 (1997). ArticleCAS Google Scholar
Behrenfeld, M. J., Boss, E., Siegel, D. A. & Shea, D. M. Carbon-based ocean productivity and phytoplankton physiology from space. Glob. Biogeochem. Cycles19, GB1006–GB1006 (2005). Article Google Scholar
Westberry, T., Behrenfeld, M. J., Siegel, D. A. & Boss, E. Carbon-based primary productivity modeling with vertically resolved photoacclimation. Global Biogeochem. Cycles22, GB2024 (2008).
Lee, Z. et al. An assessment of optical properties and primary production derived from remote sensing in the Southern Ocean (SO GasEx). J. Geophys. Res.116, C00F03 (2011).
Silsbe, G. M., Behrenfeld, M. J., Halsey, K. H., Milligan, A. J. & Westberry, T. K. The CAFE model: A net production model for global ocean phytoplankton. Glob. Biogeochem. Cycles30, 1756–1777 (2016). ArticleCAS Google Scholar
Fay, A. R. & McKinley, G. A. Global open-ocean biomes: mean and temporal variability. Earth Syst. Sci. Data6, 273–284 (2014). Article Google Scholar
Martínez-Moreno, J. et al. Global changes in oceanic mesoscale currents over the satellite altimetry record. Nat. Clim. Chang11, 397–403 (2021). Article Google Scholar
Sallée, J.-B. et al. Summertime increases in upper-ocean stratification and mixed-layer depth. Nature591, 592–598 (2021). Article Google Scholar
Yu, S., Bai, Y., He, X., Gong, F. & Li, T. A new merged dataset of global ocean chlorophyll-a concentration for better trend detection. Front. Mar. Sci.10, https://doi.org/10.3389/fmars.2023.1051619 (2023).
Boyd, P. W., Lennartz, S. T., Glover, D. M. & Doney, S. C. Biological ramifications of climate-change-mediated oceanic multi-stressors. Nat. Clim. Chang5, 71–79 (2015). Article Google Scholar
Rubner, Y., Tomasi, C. & Guibas, L. J. The Earth Mover’s Distance as a Metric for Image Retrieval. Int J. Comput Vis.40, 99–121 (2000). Article Google Scholar
Song, L., Lee, Z., Shang, S. & Wu, J. Sensitivity of a carbon-based primary production model on satellite ocean color products. Remote Sens Environ.311, 114304 (2024). Article Google Scholar
Marra, J., Trees, C. C. & O’Reilly, J. E. Phytoplankton pigment absorption: A strong predictor of primary productivity in the surface ocean. Deep Sea Res. Part I Oceanographic Res. Pap.54, 155–163 (2007). Article Google Scholar
Campbell, J. et al. Comparison of algorithms for estimating ocean primary production from surface chlorophyll, temperature, and irradiance. Glob. Biogeochem. Cycles16, 9-1–9–15 (2002). Article Google Scholar
Carr, M.-E. et al. A comparison of global estimates of marine primary production from ocean color. Deep Sea Res. Part II Topical Stud. Oceanogr.53, 741–770 (2006). Article Google Scholar
Saba, V. S. et al. An evaluation of ocean color model estimates of marine primary productivity in coastal and pelagic regions across the globe. Biogeosciences8, 489–503 (2011). ArticleCAS Google Scholar
Tao, Z., Wang, Y., Ma, S., Lv, T. & Zhou, X. A Phytoplankton Class-Specific Marine Primary Productivity Model Using MODIS Data. IEEE J. Sel. Top. Appl Earth Obs Remote Sens10, 5519–5528 (2017). Article Google Scholar
Wu, J. et al. Evaluation of three contrasting models in estimating primary production from ocean color remote sensing using long-term time-series data at oceanic and coastal sites. Remote Sens. Environ.302, 113983 (2024). Article Google Scholar
Ryan-Keogh, T. J., Thomalla, S. J., Monteiro, P. M. S. & Tagliabue, A. Multidecadal trend of increasing iron stress in Southern Ocean phytoplankton. Science379, 834–840 (2023).
Browning, T. J. et al. Persistent equatorial Pacific iron limitation under ENSO forcing. Nature621, 330–335 (2023). ArticleCAS Google Scholar
Good, S. A., Martin, M. J. & Rayner, N. A. EN4: Quality controlled ocean temperature and salinity profiles and monthly objective analyses with uncertainty estimates. J. Geophys. Res. Oceans118, 6704–6716 (2013). Article Google Scholar
de Boyer Montégut, C., Madec, G., Fischer, A. S., Lazar, A. & Iudicone, D. Mixed layer depth over the global ocean: An examination of profile data and a profile-based climatology. J. Geophys. Res.109, C12003 (2004).
Ryan-Keogh, T. J., Thomalla, S. J., Chang, N. & Moalusi, T. A new global oceanic multi-model net primary productivity data product. Earth Syst. Sci. Data15, 4829–4848 (2023). Article Google Scholar
Ryan-Keogh, T. Net primary production from the Eppley-VGPM, Behrenfeld-VGPM, Behrenfeld-CbPM, Westberry-CbPM and Silsbe-CAFE algorithms. https://doi.org/10.5281/ZENODO.7849935 (2023).
Ryan-Keogh, T. Net primary production from the Lee-AbPM algorithm. https://doi.org/10.5281/ZENODO.10014030 (2023).
Ziehn, T. et al. The Australian Earth System Model: ACCESS-ESM1.5. J. South. Hemisph. Earth Syst. Sci.70, 193–214 (2020). Article Google Scholar
Danabasoglu, G. et al. The Community Earth System Model Version 2 (CESM2). J. Adv. Model Earth Syst.12, e2019MS001916 (2020). Article Google Scholar
Lovato, T. et al. CMIP6 Simulations With the CMCC Earth System Model (CMCC-ESM2). J. Adv. Model Earth Syst.14, e2021MS002814 (2022). Article Google Scholar
Séférian, R. et al. Evaluation of CNRM Earth System Model, CNRM-ESM2-1: Role of Earth System Processes in Present-Day and Future Climate. J. Adv. Model Earth Syst.11, 4182–4227 (2019). Article Google Scholar
Christian, J. R. et al. Ocean biogeochemistry in the Canadian Earth System Model version 5.0.3: CanESM5 and CanESM5-CanOE. Geosci. Model Dev.15, 4393–4424 (2022). ArticleCAS Google Scholar
Döscher, R. et al. The EC-Earth3 Earth system model for the Coupled Model Intercomparison Project 6. Geosci. Model Dev.15, 2973–3020 (2022). Article Google Scholar
Dunne, J. P. et al. The GFDL Earth System Model Version 4.1 (GFDL-ESM 4.1): Overall Coupled Model Description and Simulation Characteristics. J. Adv. Model Earth Syst.12, e2019MS002015 (2020). Article Google Scholar
Boucher, O. et al. Presentation and Evaluation of the IPSL-CM6A-LR Climate Model. J. Adv. Model Earth Syst.12, e2019MS002010 (2020). Article Google Scholar
Mauritsen, T. et al. Developments in the MPI-M Earth System Model version 1.2 (MPI-ESM1.2) and Its Response to Increasing CO2. J. Adv. Model Earth Syst.11, 998–1038 (2019). Article Google Scholar
Yukimoto, S. et al. The Meteorological Research Institute Earth System Model Version 2.0, MRI-ESM2.0: Description and Basic Evaluation of the Physical Component. J. Meteorological Soc. Jpn. Ser. II97, 931–965 (2019). Article Google Scholar
Seland, Ø. et al. Overview of the Norwegian Earth System Model (NorESM2) and key climate response of CMIP6 DECK, historical, and scenario simulations. Geosci. Model Dev.13, 6165–6200 (2020). ArticleCAS Google Scholar
Yool, A. et al. Evaluating the physical and biogeochemical state of the global ocean component of UKESM1 in CMIP6 historical simulations. Geosci. Model Dev.14, 3437–3472 (2021). ArticleCAS Google Scholar
Schulzweida, U. CDO User Guide (2.3.0). Zenodo. https://doi.org/10.5281/zenodo.10020800 (2023).
Virtanen, P. et al. SciPy 1.0: fundamental algorithms for scientific computing in Python. Nat. Methods17, 261–272 (2020). ArticleCAS Google Scholar
Pedregosa, F. et al. Scikit-learn: Machine Learning in Python. J. Mach. Learn. Res.85, 2825–2830 (2011). Google Scholar
Huber, P. J. Robust Statistics. https://doi.org/10.1002/0471725250 (1981).
Hussain, M. M. & Mahumd, I. pyMannKendall: a python package for non parametric Mann Kendall family of trends tests. J. Open Source Softw.4, 1556 (2019). Article Google Scholar
Seabold, S. & Perktold, J. Statsmodels: Econometric and Statistical Modeling with Python. 92–96, https://doi.org/10.25080/Majora-92bf1922-011 (2010).
Newey, W. & West, K. A simple, positive semi-definite, heteroscedasticity and autocorrelation consistent covariance matrix. Appl. Econ.33, 125–132 (2014). Google Scholar
Vaserstein, L. N. Markov processes over denumerable products of spaces, describing large systems of automata. Probl. Peredachi Informatsii5, 64–72 (1969). Google Scholar
Levina, E. & Bickel, P. The Earth Mover’s distance is the Mallows distance: some insights from statistics. Proc. Eighth IEEE Int. Conf. Computer Vision. ICCV 20012, 251–256 (2001). Google Scholar Download references
We acknowledge the European Space Agency and the Ocean Colour Climate Change Initiative teams for making the remote sensing ocean colour data available. GlobColour data (http://globcolour.info) used in this study has been developed, validated, and distributed by ACRI-ST, France. We acknowledge the World Climate Research Programme’s Working Group on Coupled Modelling, which is responsible for CMIP, and we thank the climate modelling groups (listed in the materials and methods of this work) for producing and making available their model output. We would like to thank Zhongping Lee for sharing the code for running the absorption-based production model. We would also like to thank Laurent Bopp and Chris Follett for advice. T.J.R.K. and S.J.T. were supported through the CSIR’s Southern Ocean Carbon-Climate Observatory (SOCCO) Programme (http://socco.org.za/) funded by the Department of Science and Innovation (DSI/CON C3184/2023), the CSIR’s Parliamentary Grant (0000005278) and the National Research Foundation (SANAP23042496681; MCR210429598142). A.T. received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program (grant agreement no. 724289) and the National Environment Research Council (NE/Y004531/1).
Southern Ocean Carbon-Climate Observatory, CSIR, Cape Town, South Africa
Thomas J. Ryan-Keogh & Sandy J. Thomalla
Department of Earth, Ocean and Ecological Sciences, School of Environmental Sciences, University of Liverpool, Liverpool, UK
Alessandro Tagliabue
Marine and Antarctic Research Centre for Innovation and Sustainability, Department of Oceanography, University of Cape Town, Cape Town, South Africa
Sandy J. Thomalla
You can also search for this author in PubMedGoogle Scholar
You can also search for this author in PubMedGoogle Scholar
You can also search for this author in PubMedGoogle Scholar
Thomas Ryan-Keogh conceived the study, performed the analysis, developed the methodology, wrote the code, produced the figures, interpreted the results and took the lead in writing the manuscript. Alessandro Tagliabue conceived the study, developed the methodology, interpreted the results and provided inputs and refinement for writing the manuscript. Sandy Thomalla conceived the study, interpreted the results and provided inputs and refinement for writing the manuscript.
Correspondence to Thomas J. Ryan-Keogh.
The authors declare no competing interests. Communications Earth & Environment thanks Etienne Pauthenet, Benoit Pasquier and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: José Luis Iriarte Machuca, Alice Drinkwater, Mengjie Wang. A peer review file is available. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. Reprints and permissions
Ryan-Keogh, T.J., Tagliabue, A. & Thomalla, S.J. Global decline in net primary production underestimated by climate models. Commun Earth Environ6, 75 (2025). https://doi.org/10.1038/s43247-025-02051-4 Download citation
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s43247-025-02051-4
Anyone you share the following link with will be able to read this content:
Sorry, a shareable link is not currently available for this article.
The Mega Millions jackpot has risen to $59 million for Friday’s drawing after no one won the top prize on Tuesday night.
If someone should match all six numbers in the Friday night drawing, they could choose a one-time cash payment of $26.6 million. The numbers are set to be drawn just after 11 p.m. ET and we’ll have the results below.
The most recent big Mega Millions winner was a lucky lottery ticketholder in Arizona, who struck gold on Jan. 18, winning a $113 million jackpot.
According to the lottery, only four tickets matched all six Mega Millions numbers in 2024 – the fewest jackpots won in a calendar year since the game began in 2002.
The winning numbers from the Friday, Jan. 31 Mega Millions drawing will be posted here once drawn. Need a break?Play the USA TODAY Daily Crossword Puzzle. Winning lottery numbers are sponsored by Jackpocket, the official digital lottery courier of the USA TODAY Network.
Any Mega Millions winners will be added here once announced by lottery officials. To view the list of past winners, visit the Mega Millions website.
To play the Mega Millions, you have to buy a ticket. You can do this at a variety of locations, including your local convenience store, gas station, or even grocery store. In some states, Mega Millions tickets can be bought online.
Once you have your ticket, you need to pick six numbers. Five of them will be white balls with numbers from 1 to 70. The gold Mega Ball ranges from 1 to 25.
If you’re feeling especially unlucky, or don’t want to go through the hassle of picking, you can ask for a “Quick Pick“ or an ”Easy Pick.” These options let the computer randomly generate numbers for you.
For an extra dollar, players can add the “Megaplier,” which increases non-jackpot prizes by two, three, four, or five times. The Megaplier is drawn before six Mega Millions numbers are drawn.
There are 15 Megapiler balls in all:
Tickets can be purchased in person at gas stations, convenience stores and grocery stores. Some airport terminals may also sell lottery tickets. You can also order tickets online through Jackpocket, the official digital lottery courier of the USA TODAY Network, in these U.S. states and territories: Arizona, Arkansas, Colorado, Idaho, Maine, Massachusetts, Minnesota, Montana, Nebraska, New Hampshire, New Jersey, New Mexico, New York, Ohio, Oregon, Puerto Rico, Texas, Washington, D.C., and West Virginia. The Jackpocket app allows you to pick your lottery game and numbers, place your order, see your ticket and collect your winnings all using your phone or home computer. Jackpocket is the official digital lottery courier of the USA TODAY Network. Gannett may earn revenue for audience referrals to Jackpocket services. Must be 18+, 21+ in AZ and 19+ in NE. Not affiliated with any State Lottery. Gambling Problem? Call 1-877-8-HOPE-NY or text HOPENY (467369) (NY); 1-800-327-5050 (MA); 1-877-MYLIMIT (OR); 1-800-981-0023 (PR); 1-800-GAMBLER (all others). Visit jackpocket.com/tos for full terms. Fernando Cervantes Jr. is a trending news reporter for USA TODAY. Reach him at fernando.cervantes@gannett.com and follow him on X @fern_cerv_.
Français 
English