Estimates of the diffuse attenuation coefficient ($K_d$) at two different wavelengths and band-integrated (PAR) were obtained using different published algorithms developed for open ocean waters spanning in type from explicit-empirical, semi-analytical and implicit-empirical and applied to data from spectral radiometers on board six different satellites (MODIS-Aqua, MODIS-Terra, VIIRS–SNPP, VIIRS-JPSS, OLCI-Sentinel 3A and OLCI-Sentinel 3B). The resultant $K_d$s were compared to those inferred from measurements of radiometry from sensors on board autonomous profiling floats (BGC-Argo). Advantages of BGC-Argo measurements compared to ship-based ones include: 1. uniform sampling in time throughout the year, 2. large spatial coverage, and 3. lack of shading by platform. Over 5000 quality-controlled matchups between $K_d$s derived from float and from satellite sensors were found with values ranging from 0.01 to 0.67 m${}^{-1}$. Our results show that although all three algorithm types provided similarly ranging values of $K_d$ to those of the floats, for most sensors, a given algorithm produced statistically different $K_d$ distributions from the two others. Algorithm results diverged the most for low $K_d$ (clearest waters). Algorithm biases were traced to the limitations of the datasets the algorithms were developed and trained with, as well as the neglect of sun angle in some algorithms. This study highlights: 1. the importance of using comprehensive field-based datasets (such as BGC-Argo) for algorithm development, 2. the limitation of using radiative-transfer model simulations only for algorithm development, and 3. the potential for improvement if sun angle is taken into account explicitly to improve empirical $K_d$ algorithms. Recent augmentation of profiling floats with hyper-spectral radiometers should be encouraged as they will provide additional constraints to develop algorithms for upcoming missions such as NASA’s PACE and SBG and ESA’s CHIME, all of which will include a hyper-spectral radiometer.

Phytoplankton photoacclimation is a well‐documented response to changes in light and nutrient availability, with the Chlorophyll a to phytoplankton Carbon ratio ( = Chl: ) increasing at low light and decreasing under high light to optimize growth rate. Accurate estimation of phytoplankton growth rates and Net Primary Production (NPP) from space requires knowledge of , but cloud cover creates gaps. Current NPP models fill in the gaps by interpolating Chl (and other inputs) from clear‐sky pixels, ignoring the possibility of photoacclimation underneath clouds. Using data from 9,000 matchups between BioGeoChemical‐Argo floats and cloud cover from the Moderate Resolution Imaging Spectroradiometer, we compared the response of to irradiances under cloudy and clear skies. We found that phytoplankton photoacclimate similarly regardless of sky conditions at the global scale. This study highlights an incorrect assumption in current NPP estimates and suggests ways to improve global assessments of both chlorophyll and NPP.

Sudden shifts in marine plankton communities in response to environmental changes are of special concern because of their low predictability and high potential impacts on ocean ecosystems. We explored how anthropogenic climate change influences the spatial extent and frequency of changepoints in plankton populations by comparing the behavior of a plankton community in a coupled Earth system model under pre‐industrial, historical 20th century, and projected 21st century forcing. The ocean areas where surface ocean temperature, nutrient concentrations, and different plankton types exhibited changepoints expanded over time. In contrast, regional hotspots where changepoints occur frequently largely disappeared. Heterotrophy and larger organism sizes were associated with more changepoints. In the pre‐industrial and 20th century, plankton changepoints were associated with shifts in physical fronts, and more often with changepoints for iron and silicate than for nitrate and phosphate. In the 21st century, climate change disrupts these interannual‐variability‐driven changepoint patterns. Together, our results suggest anthropogenic climate change may drive less frequent but more widespread changepoints simultaneously affecting several components of pelagic food webs.

We recently found a significant bias between spectral diffuse attenuation coefficient (K _d (λ)) retrievals by common ocean color algorithms and measurements from profiling floats [ Remote. Sens. 14, 4500 (2022)10.3390/rs14184500 ]. Here we show, using a multi-satellite match-up dataset, that the bias is markedly reduced by simple "tuning" of the algorithm’s empirical coefficients. However, while the float dataset encompasses a larger proportion of the ocean’s variability than previously used datasets, it does not cover the whole range of variability of observed remote sensing reflectance (R _rs ). Thus, using algorithms tuned to this more comprehensive dataset may still result in a temporal and/or geographical bias in global application. To address this generalization issue, we evaluated a variety of analytical algorithms based on radiative transfer theory and settled on a specific one. This algorithm computes K _d (λ) from inherent optical properties (IOPs) obtained from an R _rs inversion and information about the angular distribution of the radiance transmitted through the air/ocean interface. The resulting K _d (λ) estimates at 412 and 490 nm were not appreciably biased against the float measurements. Evaluation using other in-situ datasets and radiative transfer simulations was also satisfactory. Statistical performance was good in both clear and turbid waters. Further work should be conducted to examine whether the tuned algorithms and/or the new analytical algorithm demonstrate adequate hyperspectral performance.

Accurate estimation of the diffuse attenuation coefficient of photosynthetically active radiation, K_d(PAR), is critical for understanding and modeling key physical, chemical, and biological processes in waters. In this study, a deep learning model (DLKPAR) was developed for remotely estimating K_d(PAR). Compared to the traditional empirical algorithms and semi-analytical algorithm, DLKPAR demonstrated an improvement in the model’s stability and accuracy. By using in situ NOMAD data to evaluate the model’s performance, DLKPAR had lower root mean square difference (RMSD; 0.028 vs. 0.030-0.048 m^-1) and mean absolute relative difference (MARD; 0.14 vs. 0.17-0.25) and higher R² (0.94 vs. 0.82-0.94). The statistical results of the matchup NOMAD and Argo data to the MODIS also indicated DLKPAR improves the inversion accuracy of K_d(PAR) and could be applied to remotely estimate K_d(PAR) in the global oceans. Therefore, we anticipate that DLKPAR could yield reliable K_d(PAR) values from ocean color remote sensing, providing an accurate estimation of visible light attenuation in the upper ocean and facilitating biogeochemical cycle research.

There was an error in the original publication