The spread of COVID-19 in the U.S. prompted nonpharmaceutical interventions which caused a reduction in mobility everywhere, although with large disparities between different counties. Using a Bayesian spatial modeling framework, we investigated the association of county-level demographic and socioeconomic factors with changes in workplace mobility at two points in time: during the early stages of the epidemic (lockdown phase) and in the following phase (recovery phase) up to July 2020. While controlling for the perceived risk of infection, socioeconomic and demographic covariates explain about 40% of the variance in changes in workplace mobility during the lockdown phase, which reduces to about 10% during the recovery phase. During the lockdown phase, the results show larger drops in mobility in counties with richer families, that are less densely populated, with an older population living in dense neighborhoods, and with a lower proportion of Hispanic population. When also accounting for the residual spatial variability, the variance explained by the model increases to more than 70%, suggesting strong proximity effects potentially related to state- and county-wise regulations. These results provide community-level insights on the evolution of the U.S. mobility during the first wave of the epidemic that could directly benefit policy evaluation and interventions.

Recent advances in statistical and machine learning have opened the possibility of forecasting the behaviour of chaotic systems using recurrent neural networks. In this article we investigate the applicability of such a framework to geophysical flows, known to involve multiple scales in length, time and energy and to feature intermittency. We show that both multiscale dynamics and intermittency introduce severe limitations to the applicability of recurrent neural networks, both for short-term forecasts as well as for the reconstruction of the underlying attractor. We suggest that possible strategies to overcome such limitations should be based on separating the smooth large-scale dynamics from the intermittent/small-scale features. We test these ideas on global sea-level pressure data for the past 40 years, a proxy of the atmospheric circulation dynamics. Better short- and long-term forecasts of sea-level pressure data can be obtained with an optimal choice of spatial coarse graining and time filtering.

Fine resolution estimates of demographic and socioeconomic attributes are crucial for planning and policy development. While several efforts have been made to produce fine-scale gridded population estimates, socioeconomic features are typically not available at scales finer than Census units, which may hide local heterogeneity and disparity. In this paper we present a new statistical downscaling approach to derive fine-scale estimates of key socioeconomic attributes. The method leverages demographic and geographical extensive covariates available at multiple scales and additional Census covariates only available at coarse resolution, which are included in the model hierarchically within a "forward learning" approach. For each selected socioeconomic variable, a Random Forest model is trained on the source Census units and then used to generate fine-scale gridded predictions, which are then adjusted to ensure the best possible consistency with the coarser Census data. As a case study, we apply this method to Census data in the United States, downscaling the selected socioeconomic variables available at the block group level, to a grid of ~300 spatial resolution. The accuracy of the method is assessed at both spatial scales, first computing a pseudo cross-validation coefficient of determination for the predictions at the block group level and then, for extensive variables only, also for the (unadjusted) predicted counts summed by block group. Based on these scores and on the inspection of the downscaled maps, we conclude that our method is able to provide accurate, smoother, and more detailed socioeconomic estimates than the available Census data.

Recent advances in statistical learning have opened the possibility to forecast the behavior of chaotic systems using recurrent neural networks. In this letter we investigate the applicability of this framework to geophysical flows, known to be intermittent and turbulent. We show that both turbulence and intermittency introduce severe limitations on the applicability of recurrent neural networks, both for short term forecasts as well as for the reconstruction of the underlying attractor. We test these ideas on global sea-level pressure data for the past 40 years, issued from the NCEP reanalysis datase, a proxy of the atmospheric circulation dynamics. The performance of recurrent neural network in predicting both short and long term behaviors rapidly drops when the systems are perturbed with noise. However, we found that a good predictability is partially recovered when scale separation is performed via a moving average filter. We suggest that possible strategies to overcome limitations should be based on separating the smooth large-scale dynamics, from the intermittent/turbulent features.

Background: The spread of COVID-19 in the US prompted non-pharmaceutical interventions which caused a sudden reduction in mobility everywhere, although with large local disparities between different counties. Methods: Using a Bayesian spatial modelling framework, we investigated the association of county-level demographic and socioeconomic factors with changes in workplaces mobility at two points in time: during the early stages of the epidemic (lockdown phase) and in the following phase (recovery phase). Findings: While controlling for the epidemiological situation, we found that the county-level socioeconomic and demographic covariates explain about 40% of the variance in changes in workplaces mobility in the lockdown phase, which reduces to about 10% in the recovery phase. During the lockdown phase, larger drops in workplaces mobility were observed in counties with a higher income, an older population, a lower density of Hispanic population, that are less-densely populated but with a larger density of workforce. Additionally, when also accounting for the residual spatial variability, the variance explained by the model in both phases increases up to 80%, suggesting strong proximity effects. Interpretation: This study suggests a strong association in the early stages of the epidemic between county-level changes in workplaces mobility and demographic and socioeconomic inequalities. Similar behaviours in nearby counties are present across the whole period of study, indicating a potential link to state- and county-wise regulations. These results provide community-level insights on the evolution of the US mobility during the COVID-19 epidemic that could directly benefit policy evaluation and interventions. Funding: None.Declaration of Interests: The authors declare no conflict of interests.

Multi-scale interactions between the main players of the atmospheric water cycle are poorly understood, even in the present-day climate, and represent one of the main sources of uncertainty among future climate projections. Here, we present a method to downscale observations of relative humidity available from the Sondeur Atmosphérique du Profil d'Humidité Intertropical par Radiométrie (SAPHIR) passive microwave sounder at a nominal horizontal resolution of 10 km to the finer resolution of 90 m using scattering ratio profiles from the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) lidar. With the scattering ratio profiles as covariates, an iterative approach applied to a non-parametric regression model based on a quantile random forest is used. This allows us to effectively incorporate into the predicted relative humidity structure the high-resolution variability from cloud profiles. The finer-scale water vapour structure is hereby deduced from the indirect physical correlation between relative humidity and the lidar observations. Results are presented for tropical ice clouds over the ocean: based on the coefficient of determination (with respect to the observed relative humidity) and the continuous rank probability skill score (with respect to the climatology), we conclude that we are able to successfully predict, at the resolution of cloud measurements, the relative humidity along the whole troposphere, yet ensure the best possible coherence with the values observed by SAPHIR. By providing a method to generate pseudo-observations of relative humidity (at high spatial resolution) from simultaneous co-located cloud profiles, this work will help revisit some of the current key barriers in atmospheric science. A sample dataset of simultaneous co-located scattering ratio profiles of tropical ice clouds and observations of relative humidity downscaled at the resolution of cloud measurements is available at https://doi.org/10.14768/20181022001.1 (Carella et al., 2019).

Lack of reliable observational metadata represents a key barrier to understanding sea surface temperature (SST) measurement biases, a large contributor to uncertainty in the global surface record. We present a method to identify SST measurement practice by comparing the observed SST diurnal cycle from individual ships with a reference from drifting buoys under similar conditions of wind and solar radiation. Compared to existing estimates, we found a larger number of engine room‐intake (ERI) reports post–World War II and in the period 1960–1980. Differences in the inferred mixture of observations lead to a systematic warmer shift of the bias adjusted SST anomalies from 1980 compared to previous estimates, while reducing the ensemble spread. Changes in mean field differences between bucket and ERI SST anomalies in the Northern Hemisphere over the period 1955–1995 could be as large as 0.5°C and are not well reproduced by current bias adjustment models.

Uncertainty in the bias adjustments applied to historical sea‐surface temperature (SST) measurements made using buckets are thought to make the largest contribution to uncertainty in global surface temperature trends. Measurements of the change in temperature of water samples in wooden and canvas buckets are compared with the predictions of models that have been used to estimate bias adjustments applied in widely used gridded analyses of SST. The results show that the models are broadly able to predict the dependence of the temperature change of the water over time on the thermal forcing and the bucket characteristics: volume and geometry; structure and material. Both the models and the observations indicate that the most important environmental parameter driving temperature biases in historical bucket measurements is the difference between the water and wet‐bulb temperatures. However, assumptions inherent in the derivation of the models are likely to affect their applicability. We observed that the water sample needed to be vigorously stirred to agree with results from the model, which assumes well‐mixed conditions. There were inconsistencies between the model results and previous measurements made in a wind tunnel in 1951. The model assumes non‐turbulent incident flow and consequently predicts an approximately square‐root dependence on airflow speed. The wind tunnel measurements, taken over a wide range of airflows, showed a much stronger dependence. In the presence of turbulence the heat transfer will increase with the turbulent intensity; for measurements made on ships the incident airflow is likely to be turbulent and the intensity of the turbulence is always unknown. Taken together, uncertainties due to the effects of turbulence and the assumption of well‐mixed water samples are expected to be substantial and may represent the limiting factor for the direct application of these models to adjust historical SST observations.

Sea Surface Temperature (SST) represents the marine component of surface global temperature, the indicator underpinning the Paris Agreement. This thesis presents major advances in the understanding of the systematic biases and their uncertainty associated with changes in the observing protocol in the ship-only SST record since about 1850. First, by developing a method that probabilistically groups the observations in plausible ship tracks (and therefore potentially associates observations made with the same measurement method), the length of the tracks and the percentage of reports associated with individual platforms increased substantially. Following this analysis, the consistency of the SST was also found to have improved. Secondly, by comparing the SST diurnal variations observed by individual ships with a reference derived from drifting buoys, the SST measurement method was verified or estimated. Following this new classification of the changing ratio of bucket to engine-room inlet (ERI) observations, the difference between bucket and ERI SST anomalies in the period 1955 - 70 increased more rapidly when compared to existing estimates. Better and well validated physical models of SST biases in observations made with buckets were developed by comparing measurements made in the laboratory to predictions of models used in common gridded analyses to bias adjust SST observations made with buckets. Uncertainties due to the effects of turbulence and the assumption of well-mixed water samples were identified as a substantial limiting factor for the direct application of these models to the historical record. Building on the improved platform and observational metadata, SST observations from ships in the period 1992 - 2007 were bias adjusted by modelling their differences from climate-quality satellite data within a Bayesian hierarchical spatial model and as a function of the leading drivers characteristic to the observational biases for each measurement type. A comparison with existing bias adjustments, showed that current SST estimates for the past two decades might be characterized by undetected biases, especially in the ERI record, that could affect the estimates of global and regional surface temperature trends.

AbstractGlobal surface temperature changes are a fundamental expression of climate change. Recent, much-debated variations in the observed rate of surface temperature change have highlighted the importance of uncertainty in adjustments applied to sea surface temperature (SST) measurements. These adjustments are applied to compensate for systematic biases and changes in observing protocol. Better quantification of the adjustments and their uncertainties would increase confidence in estimated surface temperature change and provide higher-quality gridded SST fields for use in many applications.Bias adjustments have been based on either physical models of the observing processes or the assumption of an unchanging relationship between SST and a reference dataset, such as night marine air temperature. These approaches produce similar estimates of SST bias on the largest space and time scales, but regional differences can exceed the estimated uncertainty. We describe challenges to improving our understanding of ...

Sea surface temperature (SST) is typically used as the marine component of the global surface temperature record, a primary metric of climate change. SST observations from ships form one of the longest instrumental records of surface marine climate. However, over the years different methods of measuring SST have been used, each with different bias characteristics. The estimation of systematic biases in the SST record is critical for climatic decadal predictions, and uncertainties in long-term trends are expected to be dominated by uncertainties in biases introduced by changes of instrumentation and measurement practices. Although the largest systematic errors in SST observations are typically assumed to relate to the period before about 1940, where SST measurements were mostly made using uninsulated buckets, there are also issues with modern data, in particular when the SST reported is the temperature of the engine-room cooling water intake (ERI). On the other hand, biases associated with modern buckets observations are assumed to be typically smaller but yet not negligible. Even in the well sampled modern era, existing studies on SST biases only provide broad estimates based on subsamples of the data and ignoring ship-byship differences. Moreover, although most of the observations in this period have known measurement type, this information can still be missing or wrongly reported. Here we take advantage of a new, high spatial resolution, gap-filled, daily SST for the period 1992-2010 from the European Space Agency Climate Change Initiative (ESA CCI) for SST dataset version 1.1. In this study, we use a Bayesian statistical model to characterise the uncertainty in the SST reports for individual ships using the ESA CCI SST as a reference. A Bayesian spatial analysis is used to model the differences of the observed SST from the ESA CCI SST for each ship. Known metadata and differences in the diurnal cycle are used to classify ships according to their measurement method and different models are then applied to characterise buckets and ERI biases. For bucket observations the difference between the observed SST and the ESA CCI SST is modelled as a function of the climatological air-sea temperature difference, while for ERI reports a constant offset plus a function of the climatological SST is used. By explicitly modelling the spatial correlation present in the data, this method allows us to better estimate the seasonal mean bias and the related uncertainty for each ship, down weighting observations taken at the same site as well as whole regions which where found artificially warm or cold relative to other areas sampled by the ship. Future work will use the results from this well-characterised period to understand how to extend the analysis back in time to periods where such high quality reference SST is not available, leading to a full characterisation of the SST ship biases and their uncertainty and a better estimate of the SST trend.

Drifting buoy sea‐surface temperature ( SST ) records have been used to characterize the diurnal variability of ocean temperature at a depth of order 20 cm. We use measurements covering the period 1986–2012 from the International Comprehensive Ocean‐Atmosphere Data Set ( ICOADS ) version 2.5, which is a collection of marine surface observations that includes individual SST records from drifting buoys. Appropriately transformed, this dataset is well suited for estimation of the diurnal cycle, since many drifting buoys have high temporal coverage (many reports per day), and are globally distributed. For each drifter for each day, we compute the local‐time daily SST variation relative to the local‐time daily mean SST . Climatological estimates of subdaily SST variability are found by averaging across various strata of the data: in 10° latitudinal bands as well as globally; and stratified with respect to season, wind speed and cloud cover. A parameterization of the diurnal variability is fitted as a function of the variables used to stratify the data, and the coefficients for this fit are also provided with the data. Results are consistent with expectations based on the previous work: the diurnal temperature cycle peaks in early afternoon (circa 2 pm local time); there is an increase in amplitude and a decrease in seasonality towards the equator. Generally, the ocean at this depth cools on windy days and warms on calm days, so that a component of subdaily variability is the SST tendency on slower timescales. By not ‘closing’ the diurnal cycle when stratified by environmental conditions, this dataset differs from previously published diurnal‐cycle parameterizations. This thorough characterization of the SST diurnal cycle will assist in interpreting SST observations made at different local times of day for climatological purposes, and in testing and constraining models of the diurnal‐cycle and air‐sea interaction at high temporal resolution.

ABSTRACTThe International Comprehensive Ocean‐Atmosphere Data Set (ICOADS) provides the main archive for surface marine observations for the past approximately 150 years. ICOADS ship identifier (ID) information is often missing or unusable, preventing the linking of reports to an individual ship. A method for the reconstruction of ship voyages in ICOADS is presented, by which groups of reports can be associated with an individual ship or ship track. The method defines a function representing the probability density function (pdf) of any particular report being associated with a group of reports. The parameters of the pdf are calculated from the ship data themselves, giving the likely variation of a ship report perpendicular to its overall direction of travel. For groups of reports with ID information, the PDF is used to associate reports without ID information with the known‐ID track. Reports without ID information are then clustered together to form the most probable track. Results are shown for the period 1855–1969. Both the percentage of reports associated with tracks and the length of those tracks increase substantially following tracking. Initial validation of the results was performed by visual inspection: the model implementation was then refined to improve the results. Confidence in the tracking is increased by a demonstration that the method clusters together reports with similar sea surface temperature characteristics. Issues in the data were found to be one of the main challenges in implementing the tracking technique. Particular problems encountered included the coarse resolution of some position information; reports that were mispositioned in either space or time; unidentified duplicate reports; and the fragmentation of voyages between different ICOADS acquisition sources. Some of these effects could be ameliorated by pre‐processing of ICOADS reports, however a full reprocessing of the historical input sources to ICOADS would be required to make further improvements.