Surface Temperature Forecasts Research Articles

A Novel Hybrid Data-Driven Model for Daily Land Surface Temperature Forecasting Using Long Short-Term Memory Neural Network Based on Ensemble Empirical Mode Decomposition.

Daily land surface temperature (LST) forecasting is of great significance for application in climate-related, agricultural, eco-environmental, or industrial studies. Hybrid data-driven prediction models using Ensemble Empirical Mode Composition (EEMD) coupled with Machine Learning (ML) algorithms are useful for achieving these purposes because they can reduce the difficulty of modeling, require less history data, are easy to develop, and are less complex than physical models. In this article, a computationally simple, less data-intensive, fast and efficient novel hybrid data-driven model called the EEMD Long Short-Term Memory (LSTM) neural network, namely EEMD-LSTM, is proposed to reduce the difficulty of modeling and to improve prediction accuracy. The daily LST data series from the Mapoling and Zhijiang stations in the Dongting Lake basin, central south China, from 1 January 2014 to 31 December 2016 is used as a case study. The EEMD is firstly employed to decompose the original daily LST data series into many Intrinsic Mode Functions (IMFs) and a single residue item. Then, the Partial Autocorrelation Function (PACF) is used to obtain the number of input data sample points for LSTM models. Next, the LSTM models are constructed to predict the decompositions. All the predicted results of the decompositions are aggregated as the final daily LST. Finally, the prediction performance of the hybrid EEMD-LSTM model is assessed in terms of the Mean Square Error (MSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), Pearson Correlation Coefficient (CC) and Nash-Sutcliffe Coefficient of Efficiency (NSCE). To validate the hybrid data-driven model, the hybrid EEMD-LSTM model is compared with the Recurrent Neural Network (RNN), LSTM and Empirical Mode Decomposition (EMD) coupled with RNN, EMD-LSTM and EEMD-RNN models, and their comparison results demonstrate that the hybrid EEMD-LSTM model performs better than the other five models. The scatterplots of the predicted results of the six models versus the original daily LST data series show that the hybrid EEMD-LSTM model is superior to the other five models. It is concluded that the proposed hybrid EEMD-LSTM model in this study is a suitable tool for temperature forecasting.

Parametric decadal climate forecast recalibration (DeFoReSt 1.0)

Abstract. Near-term climate predictions such as decadal climate forecasts are increasingly being used to guide adaptation measures. For near-term probabilistic predictions to be useful, systematic errors of the forecasting systems have to be corrected. While methods for the calibration of probabilistic forecasts are readily available, these have to be adapted to the specifics of decadal climate forecasts including the long time horizon of decadal climate forecasts, lead-time-dependent systematic errors (drift) and the errors in the representation of long-term changes and variability. These features are compounded by small ensemble sizes to describe forecast uncertainty and a relatively short period for which typically pairs of reforecasts and observations are available to estimate calibration parameters. We introduce the Decadal Climate Forecast Recalibration Strategy (DeFoReSt), a parametric approach to recalibrate decadal ensemble forecasts that takes the above specifics into account. DeFoReSt optimizes forecast quality as measured by the continuous ranked probability score (CRPS). Using a toy model to generate synthetic forecast observation pairs, we demonstrate the positive effect on forecast quality in situations with pronounced and limited predictability. Finally, we apply DeFoReSt to decadal surface temperature forecasts from the MiKlip prototype system and find consistent, and sometimes considerable, improvements in forecast quality compared with a simple calibration of the lead-time-dependent systematic errors.

Individual contributions of anthropogenic physical processes associated to urban traffic in improving the road surface temperature forecast using TEB model

For years, some work has been undertaken on the traffic heat issue input in the Town Energy Balance (TEB). It has been the subject of many studies related to the summer period and urban heat islands topic. However, during winter conditions, the traffic energy input was marginally integrated into the modeling of the road surface parameters. This deficiency, may explain the differences between forecast and observations for road surface status (RSS) during winter season. Over the past decade, identification and quantification of traffic effects were undertaken. However, they have been studied independently, and non-numerical model integrates the energy contribution of traffic into the RSS. Based on the (TEB) model (v7.2), recent research provided a detailed integration of the traffic thermal contribution in the TEB. This study showed traffic increases the road surface temperature (RST) by 2–3°C, and its heat inputs improve significantly the RST modeling. This study consists in evaluating the thermal contribution of each traffic process to improve the RST modeling based on field experiments. Secondly, the most significant physical processes of traffic responsible for RST changes have been identified and their contributions discussed. Finally, we analyzed the effects of weather conditions onto the thermal contribution of traffic processes.

Comparison of spatial interpolation methods for gridded bias removal in surface temperature forecasts

All numerical weather prediction (NWP) models inherently have substantial biases, especially in the forecast of near-surface weather variables. Statistical methods can be used to remove the systematic error based on historical bias data at observation stations. However, many end users of weather forecasts need bias corrected forecasts at locations that scarcely have any historical bias data. To circumvent this limitation, the bias of surface temperature forecasts on a regular grid covering Iran is removed, by using the information available at observation stations in the vicinity of any given grid point. To this end, the running mean error method is first used to correct the forecasts at observation stations, then four interpolation methods including inverse distance squared weighting with constant lapse rate (IDSW-CLR), Kriging with constant lapse rate (Kriging-CLR), gradient inverse distance squared with linear lapse rate (GIDS-LR), and gradient inverse distance squared with lapse rate determined by classification and regression tree (GIDS-CART), are employed to interpolate the bias corrected forecasts at neighboring observation stations to any given location. The results show that all four interpolation methods used do reduce the model error significantly, but Kriging-CLR has better performance than the other methods. For Kriging-CLR, root mean square error (RMSE) and mean absolute error (MAE) were decreased by 26% and 29%, respectively, as compared to the raw forecasts. It is found also, that after applying any of the proposed methods, unlike the raw forecasts, the bias corrected forecasts do not show spatial or temporal dependency.

Multi-Annual Climate Predictions for Fisheries: An Assessment of Skill of Sea Surface Temperature Forecasts for Large Marine Ecosystems

Decisions made by fishers and fisheries managers are informed by climate and fisheries observations that now often span more than 50 years. Multi-annual climate forecasts could further inform such decisions if they were skillful in predicting future conditions relative to the 50-year scope of past variability. We demonstrate that an existing multi-annual prediction system skillfully forecasts the probability of next year, the next 1-3 years, and the next 1-10 years being warmer or cooler than the 50-year average at the surface in coastal ecosystems. Probabilistic forecasts of upper and lower seas surface temperature (SST) terciles over the next 3 or 10 years from the GFDL CM 2.1 10-member ensemble global prediction system showed significant improvements in skill over the use of a 50-year climatology for most Large Marine Ecosystems (LMEs) in the North Atlantic, the western Pacific, and Indian oceans. Through a comparison of the forecast skill of initialized and uninitialized hindcasts, we demonstrate that this skill is largely due to the predictable signature of radiative forcing changes over the 50-year timescale rather than prediction of evolving modes of climate variability. North Atlantic LMEs stood out as the only coastal regions where initialization significantly contributed to SST prediction skill at the 1 to 10 year scale.

Improving ENSO prediction in CFSv2 with an analogue‐based correction method

ABSTRACTThis study focuses on improving the El Niño–Southern Oscillation (ENSO) prediction in Climate Forecast System version 2 (CFSv2) of the National Centers for Environmental Prediction through the development and application of an analogue‐based correction method. We show that errors in sea surface temperature (SST) forecasts in CFSv2 in the tropical Pacific are strongly correlated with the observed SST anomaly index in the Niño3.4 region at the forecast initiation time, indicating that SST forecast errors in CFSv2 is similar in cases in which the corresponding initial SST states are also similar. Therefore, the analogue‐based correction method is developed, in which SST forecast errors in CFSv2 can be corrected empirically using historical forecast errors, which are calculated by the same model, initiated from states that are analogues of the present initial state. Results show that the corrected SST anomaly forecasts have improved skills, as measured by the temporal correlation coefficient and root‐mean‐square error, compared with uncorrected forecasts. This is true for several different Niño SST indices, at most of lead months, and for most of initiation calendar months. Particular improvement is found for forecasts of the SST anomaly indices that are specially used to depict the two different ENSO flavours/types. With regard to the Niño3.4 index, the analogue‐based correction method also predicted the 2014/2016 El Niño event relatively successfully. The results indicate that the analogue‐based correction method provides an effective means of empirically improving ENSO prediction in CFSv2.

Comparing the Performance of Two Road Weather Models in the Netherlands

Abstract High-quality road condition forecasts are a prerequisite for road authorities to ensure wintertime road safety. Harsh winter conditions can cause problems for traffic not only in countries where snowy winters are common but also in regions where the temperature drops below the freezing point occasionally. This study reports on the evaluation of the Royal Netherlands Meteorological Institute’s (KNMI) new road weather forecasting model by comparing it with the Finnish Meteorological Institute’s (FMI) road weather model, both run for 321 Dutch road weather stations, four times daily (0300, 0900, 1500, and 2100 UTC) during the test period, 15 January–28 February 2015. Road surface temperature forecasts by both models were evaluated against observations. The KNMI model produced slightly more accurate forecasts than the FMI model. The main reason for the difference is probably due to the optimization of the physical properties of the KNMI model for the Netherlands, whereas the FMI model is designed for quite different Finnish wintertime meteorological conditions. However, in general the road surface temperature forecasts were of quite comparable quality.

Effective forecast of Northeast Pacific sea surface temperature based on a complementary ensemble empirical mode decomposition–support vector machine method

The sea surface temperature (SST) has substantial impacts on the climate; however, due to its highly nonlinear nature, evidently non-periodic and strongly stochastic properties, it is rather diffic...

On the skill of seasonal sea surface temperature forecasts in the California Current System and its connection to ENSO variability

The California Current System (CCS) is a biologically productive Eastern Boundary Upwelling System that experiences considerable environmental variability on seasonal and interannual timescales. Given that this variability drives changes in ecologically and economically important living marine resources, predictive skill for regional oceanographic conditions is highly desirable. Here, we assess the skill of seasonal sea surface temperature (SST) forecasts in the CCS using output from Global Climate Forecast Systems in the North American Multi-Model Ensemble (NMME), and describe mechanisms that underlie SST predictability. A simple persistence forecast provides considerable skill for lead times up to ~4 months, while skill above persistence is mostly confined to forecasts of late winter/spring and derives primarily from predictable evolution of ENSO-related variability. Specifically, anomalously weak (strong) equatorward winds are skillfully forecast during El Niño (La Niña) events, and drive negative (positive) upwelling anomalies and consequently warm (cold) temperature anomalies. This mechanism prevails during moderate to strong ENSO events, while years of ENSO-neutral conditions are not associated with significant forecast skill in the wind or significant skill above persistence in SST. We find also a strong latitudinal gradient in predictability within the CCS; SST forecast skill is highest off the Washington/Oregon coast and lowest off southern California, consistent with variable wind forcing being the dominant driver of SST predictability. These findings have direct implications for regional downscaling of seasonal forecasts and for short-term management of living marine resources.

Impact of single-point GPS integrated water vapor estimates on short-range WRF model forecasts over southern India

Specifying physically consistent and accurate initial conditions is one of the major challenges of numerical weather prediction (NWP) models. In this study, ground-based global positioning system (GPS) integrated water vapor (IWV) measurements available from the International Global Navigation Satellite Systems (GNSS) Service (IGS) station in Bangalore, India, are used to assess the impact of GPS data on NWP model forecasts over southern India. Two experiments are performed with and without assimilation of GPS-retrieved IWV observations during the Indian winter monsoon period (November–December, 2012) using a four-dimensional variational (4D-Var) data assimilation method. Assimilation of GPS data improved the model IWV analysis as well as the subsequent forecasts. There is a positive impact of ∼10 % over Bangalore and nearby regions. The Weather Research and Forecasting (WRF) model-predicted 24-h surface temperature forecasts have also improved when compared with observations. Small but significant improvements were found in the rainfall forecasts compared to control experiments.

Assessing the Skill and Value of Seasonal Thermal Stress Forecasts for Coral Bleaching Risk in the Western Pacific

AbstractOver the last 30 years, coral reefs around the world have been under considerable stress because of increasing anthropogenic pressures, overfishing, pollution, and climate change. A primary stress factor is anomalously warm water events, which can cause mass coral bleaching and widespread reef damage. Forecasts of sea surface temperature (SST) and the associated risk of coral bleaching can assist managers, researchers, and other stakeholders in monitoring and managing coral reef resources. At the Australian Bureau of Meteorology, monthly forecasts of SST and thermal stress metrics have been developed that are based on a dynamical seasonal prediction system known as the Predictive Ocean Atmosphere Model for Australia (POAMA). To support the effective use of these forecasts in risk-based decision-making frameworks in the western and central tropical Pacific Ocean, the skill of these forecast tools in this region was assessed using several categorical forecast skill scores. It was found that the model provides SST forecasts with statistically significant skill up to 8 months in advance (correlation coefficient > 0.4; p = 0.05) across the region. The highest skill (r > 0.9) was achieved over the central equatorial Pacific Ocean, likely as a result of this region’s strong relationship with the El Niño–Southern Oscillation (ENSO). Potential forecast value was assessed using a simplified cost–loss ratio decision model, which indicated that POAMA’s seasonal hot-spot thermal stress forecasts can provide valuable information to reef management and policy makers in the western Pacific region.

Improving road weather model forecasts by adjusting the radiation input

ABSTRACTConsiderable savings in winter road maintenance and accident costs can be achieved with accurate road weather forecasts. Forecasting road surface freezing time accurately enables the timely start of salting and thus ensures safety on roads. The optimal use of road weather observations is essential for the accuracy of short‐range road condition forecasts. Road weather models incorporate radiation and other atmospheric variables from numerical weather prediction models. In this study, observations were used to correct the forecast radiation and thus improve road weather forecasts for a set of specific sites. Eighteen different configurations of this methodology were tested for 20 road weather stations in Finland during the autumn–winter period 3 October 2013 to 13 January 2014. This study shows that the coupling method has potential to significantly improve road surface temperature forecasts. Two model configurations in particular turned out to be better than the others giving almost equally good road surface temperature forecasts. It was found that the length of the adjustment period using the corrected radiation had only a slight effect on the results. The outcome of this study can be used to improve road weather models and thus achieve more accurate forecasts.

Sea surface temperature simulation and forecast

Unconventional computing of sea surface temperature (SST) was once featured by NASA as a unique merger of science and art. Our approach led to a discovery that just one geographical point could be sufficient to track global anomalies of SST based on El Niño Southern Oscillation (ENSO). Such single point in the Pacific Ocean off of the island of Isabella in the Galapagos Islands was named the Galapagos indicator. Now we show that a single point in the Baltic Sea off of the coast of Göteborg could be also sufficient to track ENSO. We propose to name it the Baltic indicator. We also demonstrate that two crisis falls of oil price in 2008 and 2014 followed just after the local maximums of Baltic indicator. However, Baltic and Galapagos indicators do not show any evident trend in settling the global warming from the beginning of this century.

Explicitly-coupled cloud physics and radiation parameterizations and subsequent evaluation in WRF high-resolution convective forecasts

The impacts of various assumptions of cloud properties represented within a numerical weather prediction model's radiation scheme are demonstrated. In one approach, the model assumed the radiative effective radii of cloud water, cloud ice, and snow were represented by values assigned a priori, whereas a second, “coupled” approach utilized known cloud particle assumptions in the microphysics scheme that evolved during the simulations to diagnose the radii explicitly. This led to differences in simulated infrared (IR) brightness temperatures, radiative fluxes through clouds, and resulting surface temperatures that ultimately affect model-predicted diurnally-driven convection. The combined approach of evaluating simulated versus observed IR brightness temperatures, radiation reaching the ground, and surface temperature forecasts revealed the root model biases better than evaluating any single variable. This study found that the Weather Research and Forecasting (WRF) model predicted less overall clouds than was observed, particularly in the mid‑troposphere, but that properly connecting the assumptions of particle sizes in the microphysics scheme to the radiation scheme resulted in sensible cloud-radiation indirect effects and modest improvements in simulated IR brightness temperature, amount of solar radiation reaching the ground, and surface temperature.

Performance of weather research and forecasting model with variable horizontal resolution

In this paper, Weather Research and Forecasting (WRF) model is employed with three different horizontal grid spacings (45, 15 and 5 km) to assess the impact of horizontal resolution on short-range weather forecast. Simulations are carried out daily at 0000 UTC over the Indian region during the entire month of July 2011. A rigorous validation is performed against surface observations, radiosonde measurements and Tropical Rainfall Measuring Mission (TRMM) 3B42 merged rainfall product. Results show that horizontal resolution has a substantial impact on the WRF model forecast, particularly on the near surface temperature, moisture, winds and rainfall forecasts. Relative to 45-km horizontal grid spacing, 24-h forecasts of near surface temperature, moisture and winds are improved by ∼15, 9 and 4 %, respectively, when horizontal grid spacing is reduced to 5 km. Noteworthy improvement is also seen in the 24-h rainfall forecasts of the WRF model as the horizontal grid spacing decreased from 45 to 5 km. Larger improvements are observed over the Western Ghats and northeastern part of India compared to central India, which demonstrate the importance of finer resolution over the mountainous terrain compared to plains.

Impact of having realistic tropical cyclone frequency on ocean heat content and transport forecasts in a high‐resolution coupled model

AbstractThis study examines two sets of high‐resolution coupled model forecasts starting from no‐tropical cyclone (TC) and correct‐TC‐statistics initial conditions to understand the role of TC events on climate prediction. While the model with no‐TC initial conditions can quickly spin‐up TCs within a week, the initial conditions with a corrected TC distribution can produce more accurate forecast of sea surface temperature up to 1.5 months and maintain larger ocean heat content up to 6 months due to enhanced mixing from continuous interactions between initialized and forecasted TCs and the evolving ocean states. The TC‐enhanced tropical ocean mixing strengthens the meridional heat transport in the Southern Hemisphere driven primarily by Southern Ocean surface Ekman fluxes but weakens the Northern Hemisphere poleward transport in this model. This study suggests a future plausible initialization procedure for seamless weather‐climate prediction when individual convection‐permitting cyclone initialization is incorporated into this TC‐statistics‐permitting framework.

Predicting Western Pacific Subtropical High Using a Combined Tropical Indian Ocean Sea Surface Temperature Forecast

AbstractWeather and climate in East China are closely related to the variability of the western Pacific subtropical high (WPSH), which is an important part of the Asian monsoon system. The WPSH prediction in spring and summer is a critical component of rainfall forecasting during the summer flood season in China. Although many attempts have been made to predict WPSH variability, its predictability remains limited in practice due to the complexity of the WPSH evolution. Many studies have indicated that the sea surface temperature (SST) over the tropical Indian Ocean has a significant effect on WPSH variability. In this paper, a statistical model is developed to forecast the monthly variation in the WPSH during the spring and summer seasons on the basis of its relationship with SST over the tropical Indian Ocean. The forecasted SST over the tropical Indian Ocean is the predictor in this model, which differs significantly from other WPSH prediction methods. A 26-year independent hindcast experiment from 1983...

Mitigating errors in surface temperature forecasts using approximate radiation updates

AbstractDue to computational expense, the radiation schemes in many weather and climate models are called infrequently in time and/or on a reduced spatial grid. The former can lead to a lag in the diurnal cycle of surface temperature, while the latter can lead to large surface temperature errors at coastal land points due to surface fluxes computed over the ocean being used where the skin temperature and surface albedo are very different. This paper describes a computationally efficient solution to these problems, in which the surface longwave and shortwave fluxes are updated every time step and grid point according to the local skin temperature and albedo. In order that energy is conserved, it is necessary to compute the change to the net flux profile consistent with the changed surface fluxes. The longwave radiation scheme has been modified to compute also the rate of change of the profile of upwelling longwave flux with respect to the value at the surface. Then at each grid point and time step, the upwelling flux and heating‐rate profiles are updated using the new value of skin temperature. The computational cost of performing approximate radiation updates in the ECMWF model is only 2% of the cost of the full radiation scheme, so increases the overall cost of the model by only of order 0.2%. Testing the new scheme by running daily 5 day forecasts over an 8 month period reveals significant improvement in 2 m temperature forecasts at coastal stations compared to observations.

Multiscale EnKF Assimilation of Radar and Conventional Observations and Ensemble Forecasting for a Tornadic Mesoscale Convective System

Abstract In recent studies, the authors have successfully demonstrated the ability of an ensemble Kalman filter (EnKF), assimilating real radar observations, to produce skillful analyses and subsequent ensemble-based probabilistic forecasts for a tornadic mesoscale convective system (MCS) that occurred over Oklahoma and Texas on 9 May 2007. The current study expands upon this prior work, performing experiments for this case on a larger domain using a nested-grid EnKF, which accounts for mesoscale uncertainties through the initial ensemble and lateral boundary condition perturbations. In these new experiments, conventional observations (including surface, wind profiler, and upper-air observations) are assimilated in addition to the WSR-88D and the Center for Collaborative Adaptive Sensing of the Atmosphere (CASA) radar data used in the previous studies, better representing meso- and convective-scale features. The relative impacts of conventional and radar data on analyses and forecasts are examined, and biases within the ensemble are investigated. The new experiments produce a substantially improved forecast, including better representation of the convective lines of the MCS. Assimilation of radar data substantially improves the ensemble precipitation forecast. Assimilation of conventional data together with radar observations substantially improves the forecast of near-surface mesovortices within the MCS, improves forecasts of surface temperature and dewpoint, and imparts a slight but noticeable improvement to short-term precipitation forecasts. Furthermore, ensemble analyses and forecasts are found to be sensitive to the localization radius applied to conventional data within the EnKF.

Impact of Atmosphere and Land Surface Initial Conditions on Seasonal Forecasts of Global Surface Temperature

Abstract The impact of land surface and atmosphere initialization on the forecast skill of a seasonal prediction system is investigated, and an effort to disentangle the role played by the individual components to the global predictability is done, via a hierarchy of seasonal forecast experiments performed under different initialization strategies. A realistic atmospheric initial state allows an improved equilibrium between the ocean and overlying atmosphere, increasing the model predictive skill in the ocean. In fact, in regions characterized by strong air–sea coupling, the atmosphere initial condition affects forecast skill for several months. In particular, the ENSO region, eastern tropical Atlantic, and North Pacific benefit significantly from the atmosphere initialization. On the mainland, the effect of atmospheric initial conditions is detected in the early phase of the forecast, while the quality of land surface initialization affects forecast skill in the following seasons. Winter forecasts in the high-latitude plains benefit from the snow initialization, while the impact of soil moisture initial state is particularly effective in the Mediterranean region and central Asia. However, the initialization strategy based on the full value technique may not be the best choice for land surface, since soil moisture is a strongly model-dependent variable: in fact, initialization through land surface reanalysis does not systematically guarantee a skill improvement. Nonetheless, using a different initialization strategy for land, as opposed to atmosphere and ocean, may generate inconsistencies. Overall, the introduction of a realistic initialization for land and atmosphere substantially increases skill and accuracy. However, further developments in the procedure for land surface initialization are required for more accurate seasonal forecasts.