A Novel Method for Improving Seasonal Atlantic Hurricane Activity Prediction Based on Convolutional Neural Networks and Visualization Maps

The influence of the Sea Surface Temperature (SST) measurements on tropical storm activity has been well studied in the research community. Both tropical Atlantic and tropical mean SSTs have been shown to be key predictors for the upcoming hurricane season strength. In this paper, we propose a novel approach to predicting the strength of the upcoming hurricane season, which exploits the spatial information contained in global SST maps. We design a Convolutional Neural Network (CNN) architecture that uses SST data maps as input images and predicts the Accumulated Cyclone Energy (ACE) index, classified into two classes: high intensity and medium-low intensity. Our experiments show that our novel design based on neural networks is 20% more accurate in predicting the strength of the incoming hurricane season using January SST data maps when compared to the current operational seasonal predictions issued in May by the National Oceanic and Atmospheric Administration (NOAA) for 2000-2019.

It is difficult to improve the seasonal prediction skill of winter temperature over North China, owing to the complex dynamics of East Asian winter and the relatively low prediction skill level of current climate models. Deep learning (DL) may be an informative and promising tool to enhance seasonal prediction, particularly in regions where the underlying mechanisms are not clear. Here, using a DL model based on the Convolutional Neural Network (CNN), we have found that the prediction skill for North China winter temperature (NCWT) can be extended up to five months by considering the remote impact of the Northeast Pacific sea-surface temperature (SST) on North China. Based on historical simulations of winter temperatures in North China, we selected six CMIP5 models with relatively small deviations for training the CNN, and the period chosen for training was 1852–1991. The ERA5 data during 1995–2017 were utilized to evaluate the performance of the CNN. Our CNN shows the best performance in a recent 10-year period (2008–2017), showing a significantly improved level of NCWT prediction skill with a correlation skill of 0.65 at a 5-month lead time, which is much better than the forecast skill of the state-of-the-art dynamic seasonal prediction system. Heat map analysis was used to explore the possible physical mechanisms associated with the NCWT anomaly from the perspective of the CNN; the results showed that the SST over the Northeast Pacific is highly relevant to NCWT prediction. The Northeast Pacific warming in the boreal summer is related to the development of the El Niño event in the coming winter, which may induce NCWT anomalies by atmospheric teleconnection. Climate model experiments support the role of Northeast Pacific warming in the boreal summer on NCWT. The improved capability for prediction from using the CNN may help to establish the energy policy for the coming winter and reduce the economic losses from extremely cold in North China.

In recent decades, satellites have played a pivotal role in observing ocean dynamics, providing diverse datasets with varying spatial resolutions. Notably, within these datasets, sea surface height (SSH) data typically exhibit low resolution, while sea surface temperature (SST) data have significantly higher resolution. This study introduces a Transfer Learning-enhanced Generative Adversarial Network (TLGAN) for reconstructing high-resolution SSH fields through the fusion of heterogeneous SST data. In contrast to alternative deep learning approaches that involve directly stacking SSH and SST data as input channels in neural networks, our methodology utilizes bifurcated blocks comprising Residual Dense Module and Residual Feature Distillation Module to extract features from SSH and SST data, respectively. A pixelshuffle module-based upscaling block is then concatenated to map these features into a common latent space. Employing a hybrid strategy involving adversarial training and transfer learning, we overcome the limitation that SST and SSH data should share the same time dimension and achieve significant resolution enhancement in SSH reconstruction. Experimental results demonstrate that, when compared to interpolation method, TLGAN effectively reduces reconstruction errors and fusing SST data could significantly enhance in generating more realistic and physically plausible results.

The responses of the large-scale circulation, clouds, and water vapor to an imposed sea surface temperature (SST) gradient are investigated. Simulations compare reasonably to averaged observations over the Pacific, considering the simplifications applied to the model. The model responses to sinusoidal SST patterns have distinct circulations in the upper and lower troposphere. The upper circulation is sensitive to the heating from deep convection over the warmest SST. Stronger SST gradients are associated with stronger longwave cooling above stratus clouds in the subsidence region, stronger lower-tropospheric large-scale circulation, a reduction of the rain area, and larger area coverage of low clouds. A similar SST gradient with a warmer mean temperature produces slightly weaker lower-tropospheric circulation, and slightly reduced low cloud coverage. The outgoing longwave radiation (OLR) is not sensitive to the mean SST or the range of the imposed sinusoidal SST gradient. The positive feedbacks of water vapor and decreasing high cloud OLR compensate for the increase in longwave emission with increasing mean temperature in these simulations. As the SST gradient is increased keeping the mean SST constant, the positive high cloud feedback is still active, but the air temperature increases in proportion to the maximum SST in the domain, increasing the clear-sky OLR value and keeping the average OLR constant. The net absorbed shortwave radiation (SWI) is found to be extremely sensitive to the SST gradient. The stronger lower-tropospheric large-scale circulation produces a higher water content in the high and low clouds, increasing the absolute magnitude of the shortwave cloud forcing. A 25% increase in the maximum zonal mass flux of the lower circulation of the 300-K mean, 4-K SST range simulation leads to a 7.4 W m−2 decrease in SWI. Increasing the mean SST creates a positive feedback in these simulations because of the decrease in the lower-tropospheric large-scale circulation and the resultant decrease in cloud optical depth.

Sea Surface Temperature (SST) prediction is a hot topic that has received tremendous popularity in recent years. Existing methods for SST prediction usually select one sea area of interest and conduct SST prediction by learning the spatial and temporal dependencies and patterns in historical SST data. However, global SST is a unified system of high regionality, and the SST in different sea areas shows different changing patterns due to the influence of various factors, e.g., geographic location, ocean currents and sea depth. Without a good understanding of such regionality of SST, we cannot quantitatively integrate the regionality information of SST into SST prediction models to make them adaptive to different SST patterns around the world and improve the prediction accuracy. To address this issue, we proposed the Multi-Stage Spatio–Temporal Clustering (MuSTC) method to quantitatively identify sea areas with similar SST patterns. First, MuSTC sequentially learns the representation of long-term SST with a deep temporal encoder and calculates the spatial correlation scores between grid ocean regions with self-attention. Then, MuSTC clusters grid ocean regions based on the original SST data, encoded long-term SST representation and spatial correlation scores, respectively, to obtain the sea areas with similar SST patterns from different perspectives. According to the experiments in three ocean areas, i.e., the North Pacific Ocean (NPO), the South Atlantic Ocean (SAO) and the North Atlantic Ocean (NAO), the clustering results generally match the distribution of ocean currents, which demonstrates the effectiveness of our MuSTC method. In addition, we integrate the clustering results into two representative spatio–temporal prediction models, i.e., Spatio–Temporal Graph Convolutional Networks (STGCN) and Adaptive Graph Convolutional Recurrent Network (AGCRN), to conduct SST prediction. According to the results of experiments, the integration of regionality information leads to the reduction of Root Mean Square Error (RMSE) by 1.95%, 1.39% and 1.28% in NPO, SAO and NAO, respectively, using the STGCN model, and the reduction of RMSE by 4.94%, 0.74% and 1.43% by using the AGCRN model. Such results indicate that the integration of regionality information could notably improve the prediction accuracy of SST.

South Peninsular India (SPI) benefits largely from the rainfall (RF) realised during the North East Monsoon (NEM) season that prevails from October to December spilling over to January in some of the years. Salient aspects of clouding / RF over SPI associated with 13 NEM seasons from 2000-01 to 2012-13 have been analysed using estimates of Outgoing Long wave Radiation (OLR) at 1° × 1° resolution derived from the radiance observations in the infra-red channel onboard the geostationary operational Indian satellite (INSAT) radiometers. OLR is considered as a proxy indicator for convective activity with the value of 230 Wm-2 as the threshold for RF occurrence. Year-to-year mean OLR patterns of the NEM season along with the latitudinal and longitudinal variabilities were analysed for dry, light and active phases of NEM. Based on rigorous analysis of INSAT OLR data for the above 13 years, it has been shown that during the active phase of NEM, Coastal Tamil Nadu (CTN) receives more RF while over BoB the RF is lower and decreases sharply over interior Tamil Nadu. This is a reiteration of a similar result from an earlier study based on 3 years (1996-98) OLR data from polar orbiting NOAA satellites. The spatial variation in OLR over the latitudes of 10.5, 12.5 and 14.5° N along the longitudes of 75.5-85.5° E has revealed the feature that north of 10.5° N, values of OLR are higher with decrease in RF from south to north. During dry phase of NEM in December and January, higher OLR is observed over northern latitudes of BoB than southern latitudes. It has been comprehended that in the southern latitudes of BoB, where higher sea surface temperatures (SST) are prevalent, more moisture is generated and pumped in to upper levels of the atmosphere leading to lower values of OLR compared to northern latitudes.

This study examines space–time characteristics of seasonal rainfall predictability in a tropical region by analyzing observed data and model simulations over Senegal. Predictability is analyzed in terms of the spatial coherence of observed interannual variability at the station scale, and within-ensemble coherence of general circulation model (GCM) simulations with observed sea surface temperatures (SSTs) prescribed. Seasonal mean rainfall anomalies are decomposed in terms of daily rainfall frequency and daily mean intensity. The observed spatial coherence is computed from a 13-station network of daily rainfall during the July–September season 1961–98 in terms of (i) interannual variability of a standardized anomaly index (i.e., the average of the normalized anomalies of each station), (ii) the external variance (i.e., the fraction of common variance among stations), and (iii) the number of spatiotemporal degrees of freedom. Spatial coherence of interannual anomalies across stations is found to be much stronger for seasonal rainfall amount and daily occurrence frequency, compared with daily mean intensity of rainfall. Combinatorial analysis of the station observations suggests that, for occurrence and seasonal amount, the empirical number of spatial degrees of freedom is largely insensitive to the number of stations considered, and is between 3 and 4 for Senegal. For daily mean intensity, by contrast, each station is found to convey almost independent information, and the number of degrees of freedom would be expected to increase for a denser network of stations. The GCM estimates of potential predictability and skill associated with the SST forcing are found to be remarkably consistent with those inferred from the observed spatial coherence: there is a moderate-to-strong skill at reproducing the interannual variations of seasonal amounts and rainfall occurrence, whereas the skill is weak for the mean intensity of rainfall. Over Senegal during July–September, it is concluded that (i) regional-scale seasonal amount and rainfall occurrence frequency are predictable from SSTs, (ii) daily mean intensity of rainfall is spatially incoherent and largely unpredictable at the regional scale, and (iii) point-score estimates of seasonal rainfall predictability and skill are subject to large sampling variability.

The Spanish surface fishery operates mainly during the summer season in the waters of the Bay of Biscay. Sea surface temperature (SST) data recovered from satellite images are being used to improve the operational efficiency of fishing vessels (e.g. reduce search time and increase catch rate) and to improve the understanding of the variations in catch distribution and rate needed to properly manage fisheries. The images used for retrieval of SST often present gaps due to the existence of clouds or satellite malfunction periods. The data gaps can totally or partially affect the area of interest. Within this study, an application of a technique for the reconstruction of missing data called DINEOF (data interpolating empirical orthogonal functions) is analysed, with the aim of testing its applicability in operational SST retrieval during summer months. In this case study, the Bay of Biscay is used as the target area. Three months of SST Moderate Resolution Imaging Spectroradiometer (MODIS) images, ranging from 1 May 2006 to 31 July 2006, were used. The main objective of this work is to test the overall performance of this technique, under potential operational use for the support of the fleet during the summer fishing season. The study is designed to analyse the sensitivity of the results of this technique to several details of the methodology used in the reconstruction of SST, such as the number of empirical orthogonal functions (EOFs) retained, the handling of the seasonal cycle or the length (number of images) of the SST database used. The results are tested against independent SST data from International Comprehensive Ocean–Atmosphere Data Set (ICOADS) ship reports and standing buoys and estimations of the error of the reconstructed SST fields are given. Conclusions show that over this area three months of data are enough for efficient SST reconstruction, which yields four EOFs as the optimal number needed for this case study. An extended EOF experiment with SST and SST with a lag of one day was carried out to analyse whether the autocorrelation of the SST data allows better performance in the SST reconstruction, although the experiment did not improve the results. The validation studies show that the reconstructed SSTs can be trusted, even when the amount of missing data is very high. The mean absolute deviation maps show that the error is greatest near to the coast and mainly in the upwelling areas close to the French and north-western Spanish coasts.

Coral reefs are facing severe threats and are at risk of accelerated decline due to climate change‐induced changes in their environment. Ongoing efforts to understand the mechanisms of coral response to warming rely on multiple sources of temperature data. Yet, it remains uncertain whether the Sea Surface Temperature (SST) data used for coral reef studies are consistent among different data products, despite potential implications for conservation. A better understanding of the consistency among the different SST data applied to coral reefs may facilitate the fusion of data into a standard product. This will improve monitoring and understanding of the impact of global warming on coral reefs. Four types of SST data across North‐Western and South‐Western Australia are compared to assess their differences and ability to observe high thermal stress during historical coral bleaching events. The four SST data sources included those derived from Global Circulation Models, NOAA CoralTemp SST product, ESA CCI SST product, and coral core derived SST. Coral bleaching risk indicators, Degree Heating Week (DHW), and Degree Heating Month (DHM) were calculated using these sources and compared for consistency. DHW and DHM were inconsistent among data sets and did not accurately reflect high thermal stress metrics during moderate and severe bleaching events. Some reefs did not experience bleaching in spite of high DHWs and DHMs, suggesting a mismatch in data scales, or perhaps other oceanographic factors and coral adaptation. By exploring the differences and similarities among these four data sources, this study highlights the need to compare existing indicators of thermal stress from different data sets.

The sea surface temperature (SST) mapping could be performed with a wide spatial and temporal extent in a reasonable time limit. The space-borne sensor of AVHRR was widely used for the purpose. However, the current SST retrieval techniques for infrared channels were limited only for the cloud-free area, because the electromagnetic waves in the infrared wavelengths could not penetrate the cloud. Therefore, the SST availability was low for the single image. To overcome this problem, we studied to produce the composite of three day's SST map. The diurnal changes of SST data are quite stable through a short period of time if no abrupt natural disaster occurrence. Therefore, the SST data of three consecutive days with nearly coincident daily time were merged in order to create a three day's composite SST data. The composite image could increase the SST availability. In this study, we acquired the level 1b AVHRR (Advanced Very High Resolution Radiometer) images from Malaysia Center of Remote Sensing (MACRES). The images were first preprocessed and the cloud and land areas were masked. We made some modifications on the technique of obtaining the threshold value for cloud masking. The SST was estimated by using the day split MCSST algorithm. The cloud free water pixels availability were computed and compared. The mean of SST for three day's composite data were calculated and a SST map was generated. The cloud free water pixels availability were computed and compared. The SST data availability was increased by merging the SST data.

The authors have empirically examined the dependence of the outgoing longwave radiation (OLR) on sea surface temperature (Ts), precipitable water (W), and height-mean relative humidity (RH¯). The OLR is obtained from 4 yr of data from the Earth Radiation Budget Experiment (ERBE), while Ts, W, and RH¯ are obtained from objective analyses of rawinsonde and ship data. It is found that in the midlatitudes, the surface temperature explains over 80% of the variability in the clear-sky OLR (Fcs) and almost half of the variability in the total OLR (Ftot). It fails badly in the tropics and subtropics, however, where (Ts explains only about 20% of the variability in (Fcs, and is largely decoupled from Ftot. The two-dimensional contour plot of the OLR binned with respect to Ts and RH¯ is marked by distinct changes in the gradient that are consistent with inferences from earlier investigations. For low values of Ts (<10°C), the OLR depends mainly on Ts. For values of Ts above 10°C, the OLR depends increasing...

&amp;lt;p&amp;gt;We investigate the impact of large climatological biases in the tropical Atlantic on reanalysis and seasonal prediction performance using the Norwegian Climate Prediction Model (NorCPM) in a standard and an anomaly coupled configuration. Anomaly coupling corrects the climatological surface wind and sea surface temperature (SST) fields exchanged between oceanic and atmospheric models, and thereby significantly reduces the climatological model biases of precipitation and SST. NorCPM combines the Norwegian Earth system model (NorESM) with the Ensemble Kalman Filter and assimilates SST and hydrographic profiles. We perform a reanalysis for the period 1980-2010 and a set of seasonal predictions for the period 1985-2010 with both model configurations. Anomaly coupling improves the accuracy and the reliability of the reanalysis in the tropical Atlantic, because the corrected model enables a dynamical reconstruction that satisfies better the observations and their uncertainty.&amp;lt;span&amp;gt;&amp;amp;#160; &amp;lt;/span&amp;gt;Anomaly coupling also enhances seasonal prediction skill in the equatorial Atlantic to the level of the best models of the North American multi-model ensemble, while the standard model is among the worst. However, anomaly coupling slightly damps the amplitude of Atlantic Ni&amp;amp;#241;o and Ni&amp;amp;#241;a events. The skill enhancements achieved by anomaly coupling are largest for forecast started from August and February. There is strong spring predictability barrier, with little skill in predicting conditions in June. The anomaly coupled system show some skill in predicting the secondary Atlantic Ni&amp;amp;#241;o-II SST variability that peaks in November-December from August 1st.&amp;lt;/p&amp;gt;

Dynamical prediction systems have shown potential to meet the emerging need for seasonal forecasts of regional Arctic sea ice. Observationally constrained initial conditions are a key source of skill for these predictions, but the direct influence of different observation types on prediction skill has not yet been systematically investigated. In this work, we perform a hierarchy of observing system experiments with a coupled global data assimilation and prediction system to assess the value of different classes of oceanic and atmospheric observations for seasonal sea ice predictions in the Barents Sea. We find notable skill improvements due to the inclusion of both sea surface temperature (SST) satellite observations and subsurface conductivity–temperature–depth (CTD) measurements. The SST data are found to provide the crucial source of interannual variability, whereas the CTD data primarily provide climatological and trend improvements. Analysis of the Barents Sea ocean heat budget suggests that ocean heat content anomalies in this region are driven by surface heat fluxes on seasonal time scales.

Recognition of human activity based on convolutional neural network has received the interest of researchers in recent years due to its significant improvement in accuracy. A large number of algorithms based on the deep learning approach have been proposed for activity recognition purpose. However, with the increasing advancements in technologies having limited computational resources, it needs to design an efficient deep learning-based approaches with improved utilization of computational resources. This paper presents a simple and efficient 2-Dimensional convolutional neural network (2-D CNN) architecture with very small size convolutional kernel for human activity recognition. The merit of the proposed CNN architecture over standard deep learning architectures is fewer trainable parameters and lesser memory requirement which enables it to train the proposed CNN architecture on low GPU memory-based devices and also works well with smaller as well as larger size datasets. The proposed approach consists of mainly four stages: namely (1) creation of dataset and data augmentation, (2) designing 2-D convolutional neural network (CNN) architecture, (3) the proposed 2-D CNN architecture trained from scratch up to optimum stage, and (4) evaluation of the trained 2-D CNN architecture. To illustrate the effectiveness of the proposed architecture several extensive experiments are conducted on three publicly available datasets, namely IXMAS, YouTube, and UCF101 dataset. The results of the proposed method and its comparison with other state-of-the-art methods [8-12,14,18-26,29-33] demonstrate the usefulness of the proposed method.

Multi-sensor data from different satellites are used to identify an upwelling area in the sea off northeast Taiwan. Sea surface temperature (SST) data derived from infrared and microwave, as well as sea surface height anomaly (SSHA) data derived from satellite altimeters are used for this study. An integration filtering algorithm based on SST data is developed for detecting the cold patch induced by the upwelling. The center of the cold patch is identified by the maximum negative deviation relative to the spatial mean of a SST image within the study area and its climatological mean of each pixel. The boundary of the cold patch is found by the largest SST gradient. The along track SSHA data derived from satellite altimeters are then used to verify the detected cold patch. Applying the detecting algorithm, spatial and temporal characteristics and variations of the cold patch are revealed. The cold patch has an average area of 1.92 × 104 km2. Its occurrence frequencies are high from June to October and reach a peak in July. The mean SST of the cold patch is 23.8 °C. In addition to the annual and the intraseasonal fluctuation with main peak centered at 60 days, the cold patch also has a variation period of about 4.7 years in the interannual timescale. This implies that the Kuroshio variations and long-term and large scale processes playing roles in modifying the cold patch occurrence frequency.