AI/ML communities across NOAA
Artificial Intelligence (AI) and Machine Learning (ML) efforts are underway across NOAA. As NCAI develops, this will be a space for those Communities of Practice to provide a vehicle for discovery and networking.
Featured NOAA AI Research
Each month the NCAI Newsletter features AI-related NOAA research from our community members. The rotator below highlights research from the current and previous newsletters. Subscribe to the NCAI Newsletter offsite link.
Scientists turn to artificial intelligence to assess the warming effect of reduced pollution
Reduction in aerosol cooling unmasks greenhouse gas warming, exacerbating the rate of future warming. The strict sulfur regulation on shipping fuel implemented in 2020 (IMO2020) presents an opportunity to assess the potential impacts of such emission regulations and the detectability of deliberate aerosol perturbations for climate intervention.
Here we employ machine learning to capture cloud natural variability and estimate a radiative forcing of +0.074 ±0.005 W m−2 related to IMO2020 associated with changes in shortwave cloud radiative effect over three low-cloud regions where shipping routes prevail.
We find low detectability of the cloud radiative effect of this event, attributed to strong natural variability in cloud albedo and cloud cover. Regionally, detectability is higher for the southeastern Atlantic stratocumulus deck.
These results raise concerns that future reductions in aerosol emissions will accelerate warming and that proposed deliberate aerosol perturbations such as marine cloud brightening will need to be substantial in order to overcome the low detectability.
Samudra: An AI Global Ocean Emulator for Climate
AI tools are proving extremely effective in making fast and accurate predictions on weather to seasonal time scales. Capturing decadal to centennial changes, as those arising from ocean dynamics, remains an outstanding challenge for machine learning methods.
We built an advanced AI model called ”Samudra” to simulate global ocean behavior. Samudra is trained on simulated data from a state-of-the-art ocean climate model and predicts key ocean features such as sea surface height, currents, temperature, and salinity throughout the ocean’s depth. Samudra can accurately recreate patterns in ocean variables, including year-to-year changes.
It is stable over centuries and is 150 times faster than traditional ocean models. However, Samudra still faces challenges in balancing stability with accurately predicting the effects of external factors (like climate trends), and further improvements are needed to address this limitation.
Neural nets for sustainability conversations: modeling discussion disciplines and their impact
We live in the age polarization, where conversations on matters of sustainability more often produce acrimony or stalemate than productive action. Better understanding conversation features and their impacts may lead to better innovation, solution-design, and ongoing collaboration. We describe a study to test alternate machine learning models for classifying six ‘‘discussion disciplines’’, which are conversation features associated with rhetorical intent. The model providing the best outcome used the Bi-directional Encoder Representations from Transformers (BERT) layered with a Residual Network (ResNet).
The training data were 1135 utterances from Maine aquaculture town hall- like meetings and similar conversations, which had been hand-coded for the discussion disciplines. In addition, we generated 300 phrases corresponding to three conversation outcomes: Intent-to-Act, Options-Generation, and Rela- tionship-Building. We then used the trained model and information retrieval to classify a large corpus of 591 open- source transcripts, containing over 21,000 utterances.
A binary logistic regression analysis showed that two discussion disciplines, ‘‘Inclusion’’ and ‘‘Courtesy,’’ had positive, statistically significant, impacts on Intent-to-act: a 10 percentage point increase in the share of the Inclusion or Courtesy yielded a 45% or 34% increase, respectively, in the likelihood of Intent-to-Act.
This study shows the applicability of neural networks in modeling conversations and identifying the dialog acts that can provide measurable and predictable impact on conversation outcomes. Conver- sational intelligence can support a variety of human interactions, such as town halls, policy-deliberations, private– public partnerships, and sustainability teamwork.
Bryde’s whales produce Biotwang calls, which occur seasonally in long-term acoustic recordings from the central and western North Pacific
In 2014, a novel call was discovered in autonomous acoustic recordings from the Mariana Archipelago and designated a “Biotwang”. It was assumed to be produced by a baleen whale, but without visual verification it was impossible to assign a species. Using a combination of visual and acoustic survey data collected in the Mariana Archipelago, we determined that Biotwangs are produced by Bryde’s whales. Bryde’s whales occur worldwide in tropical and warm temperate waters, but their population structure and movements are not well understood. Genetic and morphological data recognize two populations in the western North Pacific (WNP), separate from those elsewhere in the Pacific.
We used a combination of manual and machine learning annotation methods to detect Biotwangs in our extensive historical passive acoustic monitoring datasets collected across the central and western North Pacific. We identified a consistent seasonal presence of Biotwangs in the Mariana Archipelago and to the east at Wake Island, with occasional occurrence as far away as the Northwestern Hawaiian Islands and near the equator (Howland Island).
The seasonal occurrence of Biotwangs is consistent with Bryde’s whales migrating between low and mid-latitudes, with a small peak in calling between February and April and a larger peak between August and November as the whales travel past the recording sites. Our results provide evidence for a pelagic WNP population of Bryde’s whales with broad distribution, but with seasonal and inter-annual variation in occurrence that imply a complex range most likely linked to changing oceanographic conditions in this region.
GOBAI: Gridded Ocean Biogeochemistry from Artificial Intelligence
GOBAI-O2 is a gridded data product that provides three-dimensional monthly fields of dissolved oxygen in the global ocean. It was constructed by training machine learning algorithms with observations of oxygen concentration ([O2]) from discrete shipboard measurements and autonomous sensors on biogeochemical Argo floats, then applying those machine learning algorithms to three-dimensional monthly gridded fields of temperature and salinity.
The algorithms used to produce GOBAI-O2 have been validated using real observations and synthetic data from model output, and the data product itself has been compared against the World Ocean Atlas and selected discrete measurements.
Results of these validation and comparison exercises are detailed in Sharp et al. (2023) offsite link.