Seasonal Forecast Model Research Articles

Quantifying the role of land–atmosphere feedbacks in mediating near‐surface temperature persistence

Weather persistence from month to month is an important source of predictability on the intraseasonal time‐scale. Persistence in the lower troposphere is thought to be mainly due to dynamical feedbacks between the atmosphere and physical variables at or below the surface. Many recent improvements of dynamical seasonal forecasting models have been due to better initialisations and representations of the lower boundary. While many physical mechanisms that mediate persistence are well‐understood, they have not previously been considered simultaneously in a causal framework. Here we use linear methods that are widely used in the field of causal inference in statistics to identify mediators between persistent near‐surface temperature anomalies from one month to the next. The variables considered are soil temperature and soil moisture at four levels, and snow depth. In Northern Europe, the high persistence of temperature in summer is shown to be mediated mainly by soil temperature anomalies. The high persistence in winter is mediated by snow depth and soil moisture in addition to soil temperature. The annual cycles of persistence are also investigated for the mid‐ and high latitudes of the Northern Hemisphere, confirming the important role of soil temperature in summer. Snow depth is an important mediator when, and where, its interannual variability is large: in winter at midlatitudes and in spring and autumn at high latitudes. The novel contribution of this study is the introduction of a linear causal methodology and a holistic approach that allows quantification and comparison of the different roles of soil variables and snow depth in mediating near‐surface temperature persistence.

Does a Strong El Ni\xf1o Imply a Higher Predictability of Extreme Drought?

The devastating North China drought in the summer of 2015 was roughly captured by a dynamical seasonal climate forecast model with a good prediction of the 2015/16 big El Niño. This raises a question of whether strong El Niños imply higher predictability of extreme droughts. Here we show that a strong El Niño does not necessarily result in an extreme drought, but it depends on whether the El Niño evolves synergistically with Eurasian spring snow cover reduction to trigger a positive summer Eurasian teleconnection (EU) pattern that favors anomalous northerly and air sinking over North China. The dynamical forecast model that only well represents the El Niño underpredicts the drought severity, while a dynamical-statistical forecasting approach that combines both the low- and high-latitudes precursors is more skillful at long lead. In a warming future, the vanishing cryosphere should be better understood to improve predictability of extreme droughts.

Monthly streamflow forecast for National Interconnected System (NIS) using Periodic Auto-regressive Endogenous Models (PAR) and Exogenous (PARX) with climate information

ABSTRACT This study aims to find a seasonal streamflow forecast model simultaneous to all stations of SIN using periodic autoregressive models with exogenous variables (PARX) using climate indexes. Comparing the results from PAR and PARX Models, this research analyzes the impact on forecasts by using climate information. The proposed models for streamflow forecast has been carried out using natural streamflow data from Operador Nacional do Sistema (ONS) and statistical techniques (such as multiple linear regression and stepwise method to choose explanatory variables). On 27 climate indexes utilized, 4 of them are suggested in this work. The performance analysis methodology is based on the ELECTRE method further the NASH coefficient, the mean absolute percentage error, the multi-criteria distance and correlation. Forecasts with one month lead, the PAR models present better results for most stations of SIN within seasons DJF, MAM, and JJA, while for SON season there is greater efficiency from PARX model. This kind of model shows better performance during dry season in the basins at Northern Brazil – Amazonas and Araguaia-Tocantins; Central-Eastern Brazil – Eastern Atlantic and the most rivers located in the Paraná basin.

A statistically based seasonal precipitation forecast model with automatic predictor selection and its application to central and south Asia

Abstract. The study presents a statistically based seasonal precipitation forecast model, which automatically identifies suitable predictors from globally gridded sea surface temperature (SST) and climate variables by means of an extensive data-mining procedure and explicitly avoids the utilization of typical large-scale climate indices. This leads to an enhanced flexibility of the model and enables its automatic calibration for any target area without any prior assumption concerning adequate predictor variables. Potential predictor variables are derived by means of a cell-wise correlation analysis of precipitation anomalies with gridded global climate variables under consideration of varying lead times. Significantly correlated grid cells are subsequently aggregated to predictor regions by means of a variability-based cluster analysis. Finally, for every month and lead time, an individual random-forest-based forecast model is constructed, by means of the preliminary generated predictor variables. Monthly predictions are aggregated to running 3-month periods in order to generate a seasonal precipitation forecast. The model is applied and evaluated for selected target regions in central and south Asia. Particularly for winter and spring in westerly-dominated central Asia, correlation coefficients between forecasted and observed precipitation reach values up to 0.48, although the variability of precipitation rates is strongly underestimated. Likewise, for the monsoonal precipitation amounts in the south Asian target area, correlations of up to 0.5 were detected. The skill of the model for the dry winter season over south Asia is found to be low. A sensitivity analysis with well-known climate indices, such as the El Niño– Southern Oscillation (ENSO), the North Atlantic Oscillation (NAO) and the East Atlantic (EA) pattern, reveals the major large-scale controlling mechanisms of the seasonal precipitation climate for each target area. For the central Asian target areas, both ENSO and NAO are identified as important controlling factors for precipitation totals during moist winter and spring seasons. Drought conditions are found to be triggered by a cold ENSO phase in combination with a positive state of NAO in northern central Asia, and by cold ENSO conditions in combination with a negative NAO phase in southern central Asia. For the monsoonal summer precipitation amounts over southern Asia, the model suggests a distinct negative response to El Niño events.

The use of regression for assessing a seasonal forecast model experiment

Abstract. We show how factorial regression can be used to analyse numerical model experiments, testing the effect of different model settings. We analysed results from a coupled atmosphere–ocean model to explore how the different choices in the experimental set-up influence the seasonal predictions. These choices included a representation of the sea ice and the height of top of the atmosphere, and the results suggested that the simulated monthly mean air temperatures poleward of the mid-latitudes were highly sensitivity to the specification of the top of the atmosphere, interpreted as the presence or absence of a stratosphere. The seasonal forecasts for the mid-latitudes to high latitudes were also sensitive to whether the model set-up included a dynamic or non-dynamic sea-ice representation, although this effect was somewhat less important than the role of the stratosphere. The air temperature in the tropics was insensitive to these choices.

Predictive skill of climate indices compared to mean quantities in seasonal forecasts

Seasonal forecasting models are increasingly being used to forecast application‐relevant aspects of upcoming climatic conditions, often summarised by climate indices. Little is known, however, on how the predictive skill of such forecasts of climate indices relates to the predictive skill in forecasting seasonal mean conditions. Here we analyse forecasts of two generalised indices derived from daily minimum and maximum temperature: counts of events such as the number of frost days and accumulated threshold exceedances such as degree days. We find that the predictive skill of forecasts of these two types of indices is generally lower than the skill of seasonal mean daily minimum and maximum temperature. By use of a toy model we demonstrate that this reduction in skill is more pronounced for skilful forecasts and climate indices defined relative to a threshold at the tail of the distribution. Based on the toy model results we conclude that there is no indication of additional predictability in forecasts of these indices in excess of what is expected due to the predictability of the seasonal mean. To further support this hypothesis, we show that the skill in predicting climate indices can be statistically modelled successfully using the toy model and the skill in the seasonal mean.

Comparison of cross-validation and bootstrap aggregating for building a seasonal streamflow forecast model

Abstract. Based on a hindcast experiment for the period 1982–2013 in 66 sub-catchments of the Swiss Rhine, the present study compares two approaches of building a regression model for seasonal streamflow forecasting. The first approach selects a single "best guess" model, which is tested by leave-one-out cross-validation. The second approach implements the idea of bootstrap aggregating, where bootstrap replicates are employed to select several models, and out-of-bag predictions provide model testing. The target value is mean streamflow for durations of 30, 60 and 90 days, starting with the 1st and 16th day of every month. Compared to the best guess model, bootstrap aggregating reduces the mean squared error of the streamflow forecast by seven percent on average. Thus, if resampling is anyway part of the model building procedure, bootstrap aggregating seems to be a useful strategy in statistical seasonal streamflow forecasting. Since the improved accuracy comes at the cost of a less interpretable model, the approach might be best suited for pure prediction tasks, e.g. as in operational applications.

MJO prediction using the sub-seasonal to seasonal forecast model of Beijing Climate Center

By conducting several sets of hindcast experiments using the Beijing Climate Center Climate System Model, which participates in the Sub-seasonal to Seasonal (S2S) Prediction Project, we systematically evaluate the model’s capability in forecasting MJO and its main deficiencies. In the original S2S hindcast set, MJO forecast skill is about 16 days. Such a skill shows significant seasonal-to-interannual variations. It is found that the model-dependent MJO forecast skill is more correlated with the Indian Ocean Dipole (IOD) than with the El Niño–Southern Oscillation. The highest skill is achieved in autumn when the IOD attains its maturity. Extended skill is found when the IOD is in its positive phase. MJO forecast skill’s close association with the IOD is partially due to the quickly strengthening relationship between MJO amplitude and IOD intensity as lead time increases to about 15 days, beyond which a rapid weakening of the relationship is shown. This relationship transition may cause the forecast skill to decrease quickly with lead time, and is related to the unrealistic amplitude and phase evolutions of predicted MJO over or near the equatorial Indian Ocean during anomalous IOD phases, suggesting a possible influence of exaggerated IOD variability in the model. The results imply that the upper limit of intraseasonal predictability is modulated by large-scale external forcing background state in the tropical Indian Ocean. Two additional sets of hindcast experiments with improved atmosphere and ocean initial conditions (referred to as S2S_IEXP1 and S2S_IEXP2, respectively) are carried out, and the results show that the overall MJO forecast skill is increased to 21–22 days. It is found that the optimization of initial sea surface temperature condition largely accounts for the increase of the overall MJO forecast skill, even though the improved initial atmosphere conditions also play a role. For the DYNAMO/CINDY field campaign period, the forecast skill increases to 27 days in S2S_IEXP2. Nevertheless, even with improved initialization, it is still difficult for the model to predict MJO propagation across the western hemisphere–western Indian Ocean area and across the eastern Indian Ocean–Maritime Continent area. Especially, MJO prediction is apparently limited by various interrelated deficiencies (e.g., overestimated IOD, shorter-than-observed MJO life cycle, Maritime Continent prediction barrier), due possibly to the model bias in the background moisture field over the eastern Indian Ocean and Maritime Continent. Thus, more efforts are needed to correct the deficiency in model physics in this region, in order to overcome the well-known Maritime Continent predictability barrier.

Seasonal forecasting and health impact models: challenges and opportunities.

After several decades of intensive research, steady improvements in understanding and modeling the climate system have led to the development of the first generation of operational health early warning systems in the era of climate services. These schemes are based on collaborations across scientific disciplines, bringing together real-time climate and health data collection, state-of-the-art seasonal climate predictions, epidemiological impact models based on historical data, and an understanding of end user and stakeholder needs. In this review, we discuss the challenges and opportunities of this complex, multidisciplinary collaboration, with a focus on the factors limiting seasonal forecasting as a source of predictability for climate impact models.

Evaluating uncertainty in estimates of soil moisture memory with a reverse ensemble approach

Abstract. Soil moisture memory is a key component of seasonal predictability. However, uncertainty in current memory estimates is not clear and it is not obvious to what extent these are dependent on model uncertainties. To address this question, we perform a global sensitivity analysis of memory to key hydraulic parameters, using an uncoupled version of the H-TESSEL land surface model. Results show significant dependency of estimates of memory and its uncertainty on these parameters, suggesting that operational seasonal forecasting models using deterministic hydraulic parameter values are likely to display a narrower range of memory than exists in reality. Explicitly incorporating hydraulic parameter uncertainty into models may then give improvements in forecast skill and reliability, as has been shown elsewhere in the literature. Our results also show significant differences with previous estimates of memory uncertainty, warning against placing too much confidence in a single quantification of uncertainty.

Vote-Based Wind Speed Forecast Scheme Built Upon Different Artificial Neural Network Models

This paper designs and develops a wind speed forecast model based on artificial neural networks and voting strategies, aiming at efficiently integrating renewable energies into the power network. The hour-by-hour speed records collected in Jeju city during the past 15 years are classified and converted to create sequential, monthly, and seasonal forecast models, respectively. To predict the next hour speed, the speed records of the previous 5 hours are simultaneously fed to each model first. Then, the voting process picks and averages the two predictions having the best proximity out of 3. The evaluation procedure compares the predicted values and the actual speeds of 2014, which have not been used for training, and finds out that the maximum daily root mean square error for the proposed scheme is smaller than other stand-alone methods by 0.11mps. Moreover, the vote-based scheme avoids the worst case mis-prediction.

How seasonal forecast could help a decision maker: an example of climate service for water resource management

Abstract. The FP7 project EUPORIAS was a great opportunity for the climate community to co-design with stakeholders some original and innovative climate services at seasonal time scales. In this framework, Météo-France proposed a prototype that aimed to provide to water resource managers some tailored information to better anticipate the coming season. It is based on a forecasting system, built on a refined hydrological suite, forced by a coupled seasonal forecast model. It particularly delivers probabilistic river flow prediction on river basins all over the French territory. This paper presents the work we have done with "EPTB Seine Grands Lacs" (EPTB SGL), an institutional stakeholder in charge of the management of 4 great reservoirs on the upper Seine Basin. First, we present the co-design phase, which means the translation of classical climate outputs into several indices, relevant to influence the stakeholder's decision making process (DMP). And second, we detail the evaluation of the impact of the forecast on the DMP. This evaluation is based on an experiment realised in collaboration with the stakeholder. Concretely EPTB SGL has replayed some past decisions, in three different contexts: without any forecast, with a forecast A and with a forecast B. One of forecast A and B really contained seasonal forecast, the other only contained random forecasts taken from past climate. This placebo experiment, realised in a blind test, allowed us to calculate promising skill scores of the DMP based on seasonal forecast in comparison to a classical approach based on climatology, and to EPTG SGL current practice.

ENSO-based tropical cyclone forecasting using CF-ANFIS

According to literature, there are two aspects of a successful approach for seasonal forecasting of tropical cyclones, including factors relating to the formation and operation of the storms and regression methods. Dealing with the factors, El Nino---Southern Oscillation, and other global factors such as Quasi-Biennial Oscillation, Pacific Decanal Oscillation, etc. and local factors such as sea surface temperature, sea level pressures, etc. were examined for tropical cyclone forecasting. For regression, the most previous works used the linear regression-based model for seasonal tropical forecasting. However, the seasonal tropical forecasting requires high-dimensional data, so the forecasting ability using linear regression will have drawback. In this work, we analyse literatures of forecasting factors and regression methods for tropical cyclone forecasting. A CF-ANFIS algorithm integrating a conjunct space cluster and Cascade-forward neural network are proposed to forecast the number of tropical cyclone making landfall. This algorithm resolves the drawback by considering all forecast factors with high-dimensional data. The experimental results indicated that the CF-ANFIS for seasonal forecast of tropical cyclones is a significantly effective approach with high accuracy in comparison with traditional ANFIS.

Seasonal forecasting of tropical cyclogenesis over the North Indian Ocean

Over the North Indian Ocean (NIO) and particularly over the Bay of Bengal (BoB), the post-monsoon season from October to December (OND) are known to produce tropical cyclones, which cause damage to life and property over India and many neighbouring countries. The variability of frequency of cyclonic disturbances (CDs) during OND season is found to be associated with variability of previous large-scale features during monsoon season from June to September, which is used to develop seasonal forecast model of CDs frequency over the BoB and NIO based on principal component regression (PCR). Six dynamical/thermodynamical parameters during previous June–August, viz., (i) sea surface temperature (SST) over the equatorial central Pacific, (ii) sea level pressure (SLP) over the southeastern equatorial Indian Ocean, (iii) meridional wind over the eastern equatorial Indian Ocean at 850 hPa, (iv) strength of upper level easterly, (v) strength of monsoon westerly over North Indian Ocean at 850 hPa, and (vi) SST over the northwest Pacific having significant and stable relationship with CDs over BoB in subsequent OND season are used in PCR model for a training period of 40 years (1971–2010) and the latest four years (2011–2014) are used for validation. The PCR model indicates highly significant correlation coefficient of 0.77 (0.76) between forecast and observed frequency of CD over the BoB (NIO) for the whole period of 44 years and is associated with the root mean square error and mean absolute error ≤ 1 CD. With respect to the category forecast of CD frequency over BoB and NIO, the Hit score is found to be about 63% and the Relative Operating Curves (ROC) for above and below normal forecast is found to be having much better forecast skill than the climatology. The PCR model performs very well, particularly for the above and below normal CD year over the BoB and the NIO, during the test period from 2011 to 2014.

Downscaling and extrapolating dynamic seasonal marine forecasts for coastal ocean users

Marine weather and climate forecasts are essential in planning strategies and activities on a range of temporal and spatial scales. However, seasonal dynamical forecast models, that provide forecasts in monthly scale, often have low offshore resolution and limited information for inshore coastal areas. Hence, there is increasing demand for methods capable of fine scale seasonal forecasts covering coastal waters. Here, we have developed a method to combine observational data with dynamical forecasts from POAMA (Predictive Ocean Atmosphere Model for Australia; Australian Bureau of Meteorology) in order to produce seasonal downscaled, corrected forecasts, extrapolated to include inshore regions that POAMA does not cover. We demonstrate the method in forecasting the monthly sea surface temperature anomalies in the Great Australian Bight (GAB) region. The resolution of POAMA in the GAB is approximately 2°×1° (lon.×lat.) and the resolution of our downscaled forecast is approximately 1°×0.25°. We use data and model hindcasts for the period 1994–2010 for forecast validation. The predictive performance of our statistical downscaling model improves on the original POAMA forecast. Additionally, this statistical downscaling model extrapolates forecasts to coastal regions not covered by POAMA and its forecasts are probabilistic which allows straightforward assessment of uncertainty in downscaling and prediction. A range of marine users will benefit from access to downscaled and nearshore forecasts at seasonal timescales.

Spatio-temporal variability of snow water equivalent in the extra-tropical Andes Cordillera from distributed energy balance modeling and remotely sensed snow cover

Abstract. Seasonal snow cover is the primary water source for human use and ecosystems along the extratropical Andes Cordillera. Despite its importance, relatively little research has been devoted to understanding the properties, distribution and variability of this natural resource. This research provides high-resolution (500 m), daily distributed estimates of end-of-winter and spring snow water equivalent over a 152 000 km2 domain that includes the mountainous reaches of central Chile and Argentina. Remotely sensed fractional snow-covered area and other relevant forcings are combined with extrapolated data from meteorological stations and a simplified physically based energy balance model in order to obtain melt-season melt fluxes that are then aggregated to estimate the end-of-winter (or peak) snow water equivalent (SWE). Peak SWE estimates show an overall coefficient of determination R2 of 0.68 and RMSE of 274 mm compared to observations at 12 automatic snow water equivalent sensors distributed across the model domain, with R2 values between 0.32 and 0.88. Regional estimates of peak SWE accumulation show differential patterns strongly modulated by elevation, latitude and position relative to the continental divide. The spatial distribution of peak SWE shows that the 4000–5000 m a.s.l. elevation band is significant for snow accumulation, despite having a smaller surface area than the 3000–4000 m a.s.l. band. On average, maximum snow accumulation is observed in early September in the western Andes, and in early October on the eastern side of the continental divide. The results presented here have the potential of informing applications such as seasonal forecast model assessment and improvement, regional climate model validation, as well as evaluation of observational networks and water resource infrastructure development.

A Track Pattern–Based Seasonal Prediction of Tropical Cyclone Activity over the North Atlantic

Abstract A seasonal prediction model of tropical cyclone (TC) activities for the period August–October over the North Atlantic (NA) has been developed on the basis of TC track patterns. Using the fuzzy c-means method, a total of 432 TCs in the period 1965–2012 are categorized into the following four groups: 1) TCs off the U.S. East Coast, 2) TCs over the Gulf of Mexico, 3) TCs that recurve into the open ocean of the central NA, and 4) TCs that move westward in the southern NA. The model is applied to predict the four TC groups separately in conjunction with global climate forecasts from the National Centers for Environmental Prediction (NCEP) Climate Forecast System, version 2 (CFSv2). By adding the distributions of the four TC tracks with precalculated weighting factors, this seasonal TC forecast model provides the spatial distribution of TC activities over the entire NA basin. Multiple forecasts initialized in six consecutive months from February to July are generated at monthly intervals to examine the applicability of this model in operational TC forecasting. Cross validations of individual forecasts show that the model can reasonably predict the observed TC frequencies over NA at the 99% confidence level. The model shows a stable spatial prediction skill, proving its advantage for forecasting regional TC activities several months in advance. In particular, the model can generate reliable information on regional TC counts in the near-coastal regions as well as in the entire NA basin.

The West African Monsoon Onset: A Concise Comparison of Definitions

Abstract The onset of the West African monsoon (WAM) marks a vital time for local and regional stakeholders. While the seasonal progression of monsoon winds and the related migration of precipitation from the Guinea Coast toward the Sudan/Sahel is apparent, there exist contrasting man-made definitions of what the WAM onset means. Broadly speaking, onset can be analyzed regionally, locally, or over a designated intermediate scale. There are at least 18 distinct definitions of the WAM onset in publication, with little work done on comparing observed onset from different definitions or comparing onset realizations across different datasets and resolutions. Here, nine definitions have been calculated using multiple datasets of different metrics at different resolutions. It is found that mean regional onset dates are consistent across multiple datasets and different definitions. There is low interannual variability in regional onset, suggesting that regional seasonal forecasting of the onset provides few benefits over climatology. In contrast, local onsets show high spatial, interannual, and interdefinition variability. Furthermore, it is found that there is little correlation between local onset dates and regional onset dates across West Africa, implying a disharmony between regional measures of onset and the experience on a local scale. The results of this study show that evaluation of seasonal monsoon onset forecasts is far from straightforward. Given a seasonal forecasting model, it is possible to simultaneously have a good and a bad prediction of monsoon onset simply through selection of the onset definition and observational dataset used for comparison.

S<sup>4</sup>CAST v2.0: sea surface temperature based statistical seasonal forecast model

Abstract. Sea surface temperature is the key variable when tackling seasonal to decadal climate forecasts. Dynamical models are unable to properly reproduce tropical climate variability, introducing biases that prevent a skillful predictability. Statistical methodologies emerge as an alternative to improve the predictability and reduce these biases. In addition, recent studies have put forward the non-stationary behavior of the teleconnections between tropical oceans, showing how the same tropical mode has different impacts depending on the considered sequence of decades. To improve the predictability and investigate possible teleconnections, the sea surface temperature based statistical seasonal foreCAST model (S4CAST) introduces the novelty of considering the non-stationary links between the predictor and predictand fields. This paper describes the development of the S4CAST model whose operation is focused on studying the impacts of sea surface temperature on any climate-related variable. Two applications focused on analyzing the predictability of different climatic events have been implemented as benchmark examples.

Predictors of precipitation for improved water resources management in the Tarim River basin: Creating a seasonal forecast model

In recent years, an expansion of irrigated agriculture and rapid population growth have threatened the Tarim River basin's natural ecosystems and caused water shortages. Improving the water resources management in the basin requires accurate seasonal precipitation forecasts. Based on previous research, possible predictors of precipitation were selected and either downloaded directly or calculated using NCEP/NCAR Reanalysis 1 or NOAA Extended Reconstructed Sea Surface Temperature (SST) V3b data. Predictors were correlated with precipitation data, provided by the National Climate Centre of the China Meteorological Administration for the period 1961 to 2010 and averaged over the subbasins of the Tarim River. The Spearman rank correlation analyses with lead times of up to six months (or two seasons) revealed significant (at the 1% level) and strong (ρ ≤ −0.6 or ρ ≥ 0.6) correlations of precipitation in all subbasins with the SST and monsoon indices as well as with the Siberian High Intensity (SHI) and the Westerly Circulation Index (WCI). Lastly, we demonstrate the setup of a forecast model based on a multiple linear regression on the example of the Hotan River subbasin. This model predicts precipitation 5 months in advance with reasonable accuracy in two out of three configurations.