Search for: All records

Neural networks (NN) have become an important tool for prediction tasks—both regression and classification—in environmental science. Since many environmental-science problems involve life-or-death decisions and policy making, it is crucial to provide not only predictions but also an estimate of the uncertainty in the predictions. Until recently, very few tools were available to provide uncertainty quantification (UQ) for NN predictions. However, in recent years the computer-science field has developed numerous UQ approaches, and several research groups are exploring how to apply these approaches in environmental science. We provide an accessible introduction to six of these UQ approaches, then focus on tools for the next step, namely, to answer the question:Once we obtain an uncertainty estimate (using any approach), how do we know whether it is good or bad?To answer this question, we highlight four evaluation graphics and eight evaluation scores that are well suited for evaluating and comparing uncertainty estimates (NN based or otherwise) for environmental-science applications. We demonstrate the UQ approaches and UQ-evaluation methods for two real-world problems: 1) estimating vertical profiles of atmospheric dewpoint (a regression task) and 2) predicting convection over Taiwan based onHimawari-8satellite imagery (a classification task). We also provide Jupyter notebooks with Python code for implementing the UQ approaches and UQ-evaluation methods discussed herein. This article provides the environmental-science community with the knowledge and tools to start incorporating the large number of emerging UQ methods into their research.

Neural networks are used for many environmental-science applications, some involving life-or-death decision-making. In recent years new methods have been developed to provide much-needed uncertainty estimates for NN predictions. We seek to accelerate the adoption of these methods in the environmental-science community with an accessible introduction to 1) methods for computing uncertainty estimates in NN predictions and 2) methods for evaluating such estimates.

 

Sensitive responding to eye cues plays a key role in human social interactions. Pupil size provides subtle cues regarding a social interaction partner's arousal states. The current study assessed infants’ sensitivity to and preference for differences in pupil size. Specifically, we examined White 14‐month‐old infants’ pupillary responses when viewing own‐race and other‐race (Asian) eyes with dilating, constricting, and static medium‐sized pupils. Our results show that, independent of race, infants’ pupils dilated more when viewing eyes with dynamically changing (dilating and constricting) pupils than when viewing eyes with non‐changing, static, and medium‐sized pupils. We also measured infants’ looking preferences, showing that, independent of race, infants preferentially attended to eyes with dilated pupils. Moreover, our results show that infants orient more quickly to pupillary changes in own‐race eyes than in other‐race eyes. These findings demonstrate that infants detect, but do not mimic, changes in pupil size in others and show a preference for eyes with dilated pupils.

 

Abstract. The uptake of carbonyl sulfide (COS) by terrestrial plants is linked tophotosynthetic uptake of CO2 as these gases partly share the sameuptake pathway. Applying COS as a photosynthesis tracer in models requires anaccurate representation of biosphere COS fluxes, but these models have notbeen extensively evaluated against field observations of COS fluxes. In thispaper, the COS flux as simulated by the Simple Biosphere Model, version 4(SiB4), is updated with the latest mechanistic insights and evaluated with siteobservations from different biomes: one evergreen needleleaf forest, twodeciduous broadleaf forests, three grasslands, and two crop fields spread overEurope and North America. We improved SiB4 in several ways to improve itsrepresentation of COS. To account for the effect of atmospheric COS molefractions on COS biosphere uptake, we replaced the fixed atmospheric COS molefraction boundary condition originally used in SiB4 with spatially andtemporally varying COS mole fraction fields. Seasonal amplitudes of COS molefractions are ∼50–200 ppt at the investigated sites with aminimum mole fraction in the late growing season. Incorporating seasonalvariability into the model reduces COS uptake rates in the late growingseason, allowing better agreement with observations. We also replaced theempirical soil COS uptake model in SiB4 with a mechanistic model thatrepresents both uptake and production of COS in soils, which improves thematch with observations over agricultural fields and fertilized grasslandsoils. The improved version of SiB4 was capable of simulating the diurnal andseasonal variation in COS fluxes in the boreal, temperate, and Mediterraneanregion. Nonetheless, the daytime vegetation COS flux is underestimated onaverage by 8±27 %, albeit with large variability across sites. On aglobal scale, our model modifications decreased the modeled COS terrestrialbiosphere sink from 922 Gg S yr−1 in the original SiB4 to753 Gg S yr−1 in the updated version. The largest decrease influxes was driven by lower atmospheric COS mole fractions over regions withhigh productivity, which highlights the importance of accounting forvariations in atmospheric COS mole fractions. The change to a different soilmodel, on the other hand, had a relatively small effect on the globalbiosphere COS sink. The secondary role of the modeled soil component in theglobal COS budget supports the use of COS as a global photosynthesis tracer. Amore accurate representation of COS uptake in SiB4 should allow for improvedapplication of atmospheric COS as a tracer of local- to global-scaleterrestrial photosynthesis. 

Tropical South America plays a central role in global climate. Bowen ratio teleconnects to circulation and precipitation processes far afield, and the global CO₂growth rate is strongly influenced by carbon cycle processes in South America. However, quantification of basin‐wide seasonality of flux partitioning between latent and sensible heat, the response to anomalies around climatic norms, and understanding of the processes and mechanisms that control the carbon cycle remains elusive. Here, we investigate simulated surface‐atmosphere interaction at a single site in Brazil, using models with different representations of precipitation and cloud processes, as well as differences in scale of coupling between the surface and atmosphere. We find that the model with parameterized clouds/precipitation has a tendency toward unrealistic perpetual light precipitation, while models with explicit treatment of clouds produce more intense and less frequent rain. Models that couple the surface to the atmosphere on the scale of kilometers, as opposed to tens or hundreds of kilometers, produce even more realistic distributions of rainfall. Rainfall intensity has direct consequences for the “fate of water,” or the pathway that a hydrometeor follows once it interacts with the surface. We find that the model with explicit treatment of cloud processes, coupled to the surface at small scales, is the most realistic when compared to observations. These results have implications for simulations of global climate, as the use of models with explicit (as opposed to parameterized) cloud representations becomes more widespread.

 

In an era of rapid global change, our ability to understand and predict Earth's natural systems is lagging behind our ability to monitor and measure changes in the biosphere. Bottlenecks to informing models with observations have reduced our capacity to fully exploit the growing volume and variety of available data. Here, we take a critical look at the information infrastructure that connects ecosystem modeling and measurement efforts, and propose a roadmap to community cyberinfrastructure development that can reduce the divisions between empirical research and modeling and accelerate the pace of discovery. A new era of data‐model integration requires investment in accessible, scalable, and transparent tools that integrate the expertise of the whole community, including both modelers and empiricists. This roadmap focuses on five key opportunities for community tools: the underlying foundations of community cyberinfrastructure; data ingest; calibration of models to data; model‐data benchmarking; and data assimilation and ecological forecasting. This community‐driven approach is a key to meeting the pressing needs of science and society in the 21st century.