Uncovering the Over-Smoothing Challenge in Image Super-Resolution: Entropy-Based Quantification and Contrastive Optimization.

PSNR-oriented models are a critical class of super-resolution models with applications across various fields. However, these models tend to generate over-smoothed images, a problem that has been analyzed previously from the perspectives of models or loss functions, but without taking into account the impact of data properties. In this paper, we present a novel phenomenon that we term the center-oriented optimization (COO) problem, where a model's output converges towards the center point of similar high-resolution images, rather than towards the ground truth. We demonstrate that the strength of this problem is related to the uncertainty of data, which we quantify using entropy. We prove that as the entropy of high-resolution images increases, their center point will move further away from the clean image distribution, and the model will generate over-smoothed images. Implicitly optimizing the COO problem, perceptual-driven approaches such as perceptual loss, model structure optimization, or GAN-based methods can be viewed. We propose an explicit solution to the COO problem, called Detail Enhanced Contrastive Loss (DECLoss). DECLoss utilizes the clustering property of contrastive learning to directly reduce the variance of the potential high-resolution distribution and thereby decrease the entropy. We evaluate DECLoss on multiple super-resolution benchmarks and demonstrate that it improves the perceptual quality of PSNR-oriented models. Moreover, when applied to GAN-based methods, such as RaGAN, DECLoss helps to achieve state-of-the-art performance, such as 0.093 LPIPS with 24.51 PSNR on 4× downsampled Urban100, validating the effectiveness and generalization of our approach.

Dynamics of Gut Bacteria Across Different Zooplankton Genera in the Baltic Sea.

In aquatic ecosystems, zooplankton-associated bacteria potentially have a great impact on the structure of ecosystems and trophic networks by providing various metabolic pathways and altering the ecological niche of host species. To understand the composition and drivers of zooplankton gut microbiota, we investigated the associated microbial communities of four zooplankton genera from different seasons in the Baltic Sea using the 16S rRNA gene. Among the 143 ASVs (amplified sequence variants) observed belonging to heterotrophic bacteria, 28 ASVs were shared across all zooplankton hosts over the season, and these shared core ASVs represented more than 25% and up to 60% of relative abundance in zooplankton hosts but were present at low relative abundance in the filtered water. Zooplankton host identity had stronger effects on bacterial composition than seasonal variation, with the composition of gut bacterial communities showing host-specific clustering patterns. Although bacterial compositions and dominating core bacteria were different between zooplankton hosts, higher gut bacteria diversity and more bacteria contributing to the temporal variation were found in Temora and Pseudocalanus, compared to Acartia and Synchaeta. Diet diatom and filamentous cyanobacteria negatively correlated with gut bacteria diversity, but the difference in diet composition did not explain the dissimilarity of gut bacteria composition, suggesting a general effect of diet on the inner conditions in the zooplankton gut. Synchaeta maintained high stability of gut bacterial communities with unexpectedly low bacteria-bacteria interactions as compared to the copepods, indicating host-specific regulation traits. Our results suggest that the patterns of gut bacteria dynamics are host-specific and the variability of gut bacteria is not only related to host taxonomy but also related to host behavior and life history traits.

Sensitivity to hydrogen peroxide of the bloom-forming cyanobacterium Microcystis PCC 7806 depends on nutrient availability.

Application of low concentrations of hydrogen peroxide (H2O2) is a relatively new and promising method to selectively suppress harmful cyanobacterial blooms, while minimizing effects on eukaryotic organisms. However, it is still unknown how nutrient limitation affects the sensitivity of cyanobacteria to H2O2. In this study, we compare effects of H2O2 on the microcystin-producing cyanobacterium Microcystis PCC 7806 under light-limited but nutrient-replete conditions, nitrogen (N) limitation and phosphorus (P) limitation. Microcystis was first grown in chemostats to acclimate to these different experimental conditions, and subsequently transferred to batch cultures where they were treated with a range of H2O2 concentrations (0-10 mg L-1) while exposed to high light (100 µmol photons m-2 s-1) or low light (15 µmol photons m-2 s-1). Our results show that, at low light, N- and P-limited Microcystis were less sensitive to H2O2 than light-limited but nutrient-replete Microcystis. A significantly higher expression of the genes encoding for anti-oxidative stress enzymes (2-cys-peroxiredoxin, thioredoxin A and type II peroxiredoxin) was observed prior to and after the H2O2 treatment for both N- and P-limited Microcystis, which may explain their increased resistance against H2O2. At high light, Microcystis was more sensitive to H2O2 than at low light, and differences in the decline of the photosynthetic yield between nutrient-replete and nutrient-limited Microcystis exposed to H2O2 were less pronounced. Leakage of microcystin was stronger and faster from nutrient-replete than from N- and P-limited Microcystis. Overall, this study provides insight in the sensitivity of harmful cyanobacteria to H2O2 under various environmental conditions.

Changes in water color shift competition between phytoplankton species with contrasting light-harvesting strategies.

The color of many lakes and seas is changing, which is likely to affect the species composition of freshwater and marine phytoplankton communities. For example, cyanobacteria with phycobilisomes as light-harvesting antennae can effectively utilize green or orange-red light. However, recent studies show that they use blue light much less efficiently than phytoplankton species with chlorophyll-based light-harvesting complexes, even though both phytoplankton groups may absorb blue light to a similar extent. Can we advance ecological theory to predict how these differences in light-harvesting strategy affect competition between phytoplankton species? Here, we develop a new resource competition model in which the absorption and utilization efficiency of different colors of light are varied independently. The model was parameterized using monoculture experiments with a freshwater cyanobacterium and green alga, as representatives of phytoplankton with phycobilisome-based vs. chlorophyll-based light-harvesting antennae. The parameterized model was subsequently tested in a series of competition experiments. In agreement with the model predictions, the green alga won the competition in blue light whereas the cyanobacterium won in red light, irrespective of the initial relative abundances of the species. These results are in line with observed changes in phytoplankton community structure in response to lake brownification. Similarly, in marine waters, the model predicts dominance of Prochlorococcus with chlorophyll-based light-harvesting complexes in blue light but dominance of Synechococcus with phycobilisomes in green light, with a broad range of coexistence in between. These predictions agree well with the known biogeographical distributions of these two highly abundant marine taxa. Our results offer a novel trait-based approach to understand and predict competition between phytoplankton species with different photosynthetic pigments and light-harvesting strategies.

Suppressing Cyanobacteria with Hydrogen Peroxide Is More Effective at High Light Intensities.

Hydrogen peroxide (H2O2) can be used as an emergency method to selectively suppress cyanobacterial blooms in lakes and drinking water reservoirs. However, it is largely unknown how environmental parameters alter the effectiveness of H2O2 treatments. In this study, the toxic cyanobacterial strain Microcystis aeruginosa PCC 7806 was treated with a range of H2O2 concentrations (0 to 10 mg/L), while being exposed to different light intensities and light colors. H2O2 treatments caused a stronger decline of the photosynthetic yield in high light than in low light or in the dark, and also a stronger decline in orange than in blue light. Our results are consistent with the hypothesis that H2O2 causes major damage at photosystem II (PSII) and interferes with PSII repair, which makes cells more sensitive to photoinhibition. Furthermore, H2O2 treatments caused a decrease in cell size and an increase in extracellular microcystin concentrations, indicative of leakage from disrupted cells. Our findings imply that even low H2O2 concentrations of 1-2 mg/L can be highly effective, if cyanobacteria are exposed to high light intensities. We therefore recommend performing lake treatments during sunny days, when a low H2O2 dosage is sufficient to suppress cyanobacteria, and may help to minimize impacts on non-target organisms.