Large-scale Climate Predictors Research Articles

Seasonal precipitation forecasting with large scale climate predictors: a hybrid ensemble empirical mode decomposition-NARX scheme

Seasonal precipitation forecasting with large scale climate predictors: a hybrid ensemble empirical mode decomposition-NARX scheme

Seasonal precipitation forecasting for water management in the Kosi Basin, India using large-scale climate predictors

AbstractA novel approach for qualitative seasonal forecast of precipitation at a basin scale is presented as significant enhancement in seasonal forecast at regional and country scales in India. The process utilizes empirical and typically lagged relationships between target variables of interest, namely precipitation at the basin level and various large-scale climate predictors (LSCPs). A total of 14 LSCPs have been considered for the seasonal forecast of precipitation with lead times of 1, 2, and 3 months in the Kosi Basin, India. Random split training and testing were conducted on seven machine-learning (ML) models using a potential predictor dataset for model selection. The Logistic Regression (LR) model was adopted since it had the highest mean accuracy score compared to the remaining six ML models. The LR model has been optimized by testing it on all possible combinations of potential predictors using Leave-One-Out Cross-Validation (CV) scheme. The resulting Seasonal Prediction Model (SPM) provides the probability of each tercile categorized as Above Normal (AN), Normal (N), and Below Normal (BN). The model has been evaluated using various metrics.

Future Projection of Mean Temperature of Post-Monsoon Season Over Bangladesh Using Statistical Downscaling of Global Climate Models

In statistical downscaling technique, regional or local information are derived by determining a statistical model which relates large-scale climate variables or predictors generated by Global Climate Models (GCMs) to regional and local variables or predictands. In this paper, the results of GCMs were statistically downscaled to produce future climate projections of mean temperature in the post-monsoon season (October and November), for the time periods 2021-2050 and 2071-2100 for Bangladesh. The future climate projections are based on the three emission scenarios RCP2.6, RCP4.5 and RCP8.5 provided by the fifth Coupled Model Intercomparison Project (CMIP5). This paper established a method to analyze GCMs for use in statistical downscaling and utilized fifteen GCMs. The GCMs were assessed based upon their performance in simulated past climate in Bangladesh and adjoining areas. Downscaling was undertaken by linking large scale climate variables, taken from the ERA-Interim (resolution 79 km) reanalysis temperature, a gridded data set incorporating observations and climate models, to local scale observations. Overall, all fifteen GCMs, via statistical downscaling, show that mean temperature of the post-monsoon season in Bangladesh will increase under future climate scenarios. Comparing the ensemble of future projections with the reference period (1981- 2010), the mean post-monsoon temperature in Bangladesh is projected for RCP8.5 showing warming by 0.310C in near future and 1.790C in far future. On the other hand, estimated warming is 0.390C in near future and 1.140C is far future for RCP4.5. Low emission scenarios RCP2.6, near future temperature is nearly same the far future temperature.
 Journal of Engineering Science 11(2), 2020, 27-35

Developing Climate Change Projections using Different Representative Concentration Pathways of Emission Scenario: In the Case Jimma, Ethiopia

Anthropogenic influence on the climate system is clear, and the recent emissions of greenhouse gases are the highest in history. This particular activity is aimed to project the future climate using different representative concentration pathways of emission scenarios for Jima station. Statistical downscaling approach was used to downscale rainfall and temperature. Predictors were synthesized based on correlation analysis between large scale climate predictors and observed station climate data. Monthly predictors were used to establish a regression model between the predictors and observed climate variables. The regression models were validated against observed station data and were used to generate downscaled future rainfall and temperature. Data for future scenario were collected from the Canadian Centre for Climate Modeling and Analysis (CCCma) of Canada Environment.

Statistical tool for modeling of a daily precipitation process in the context of climate change

Abstract The present study proposes a climate change assessment tool based on a statistical downscaling (SD) approach for describing the linkage between large-scale climate predictors and observed daily rainfall characteristics at a local site. The proposed SD of the daily rainfall process (SDRain) model is based on a combination of a logistic regression model for representing the daily rainfall occurrences and a nonlinear regression model for describing the daily precipitation amounts. A scaling factor (SR) and correction coefficient (CR) are suggested to improve the accuracy of the SDRain model in representing the variance of the observed daily precipitation amounts in each month without affecting the monthly mean precipitation. SDRain facilitates the construction of daily precipitation models for the current and future climate conditions. The tool is tested using the National Center for Environmental Prediction re-analysis data and the observed daily precipitation data available for the 1961–2001 period at two study sites located in two completely different climatic regions: the Seoul station in subtropical-climate Korea and the Dorval Airport station in cold-climate Canada. Results of this illustrative application have indicated that the proposed functions (e.g. logistic regression, SR, and CR) contribute marked improvement in describing daily precipitation amounts and occurrences. Furthermore, the comparison analyses show that the proposed SD method could provide more accurate results than those given by the currently popular SDSM method.

Season-ahead forecasting of water storage and irrigation requirements – an application to the southwest monsoon in India

Abstract. Water risk management is a ubiquitous challenge faced by stakeholders in the water or agricultural sector. We present a methodological framework for forecasting water storage requirements and present an application of this methodology to risk assessment in India. The application focused on forecasting crop water stress for potatoes grown during the monsoon season in the Satara district of Maharashtra. Pre-season large-scale climate predictors used to forecast water stress were selected based on an exhaustive search method that evaluates for highest ranked probability skill score and lowest root-mean-squared error in a leave-one-out cross-validation mode. Adaptive forecasts were made in the years 2001 to 2013 using the identified predictors and a non-parametric k-nearest neighbors approach. The accuracy of the adaptive forecasts (2001–2013) was judged based on directional concordance and contingency metrics such as hit/miss rate and false alarms. Based on these criteria, our forecasts were correct 9 out of 13 times, with two misses and two false alarms. The results of these drought forecasts were compared with precipitation forecasts from the Indian Meteorological Department (IMD). We assert that it is necessary to couple informative water stress indices with an effective forecasting methodology to maximize the utility of such indices, thereby optimizing water management decisions.

Dynamical-statistical projections of the climate change impact on agricultural production in Benin by means of a cross-validated linear model combined with Bayesian statistics

West Africa is highly vulnerable to climate change and a robust quantification of the societal impacts of climate change is essential to guide the necessary adaptation efforts. Here, we project the potential impacts of climate change on nine important crops using climate change information from a gridded observational data set and a high-resolution regional climate model driven with and without land use changes. Probabilistic crop models are developed and forced with climate predictors until 2050. It is found that large-scale climate predictors are sufficiently robust for crop modelling in the absence of reliable local climate information. Pineapple, maize, groundnuts, cassava and cowpeas will face harmful effects with an average yield reduction in the range of 11%–33% by 2050, whereas sorghum, yam, cotton and rice will benefit from climate change with an average yield gain of 10–39%. Temperature increase rather than precipitation change is responsible for the projected yield changes. Our study also shows that land cover degradation in West Africa tends to reduce yield for most crops whilst favouring the production of yam and cotton.

Statistical–Dynamical Seasonal Forecast of North Atlantic and U.S. Landfalling Tropical Cyclones Using the High-Resolution GFDL FLOR Coupled Model

Abstract Retrospective seasonal forecasts of North Atlantic tropical cyclone (TC) activity over the period 1980–2014 are conducted using a GFDL high-resolution coupled climate model [Forecast-Oriented Low Ocean Resolution (FLOR)]. The focus is on basin-total TC and U.S. landfall frequency. The correlations between observed and model predicted basin-total TC counts range from 0.4 to 0.6 depending on the month of the initial forecast. The correlation values for U.S. landfalling activity based on individual TCs tracked from the model are smaller and between 0.1 and 0.4. Given the limited skill from the model, statistical methods are used to complement the dynamical seasonal TC prediction from the FLOR model. Observed and predicted TC tracks were classified into four groups using fuzzy c-mean clustering to evaluate the model’s predictability in observed classification of TC tracks. Analyses revealed that the FLOR model has the highest skill in predicting TC frequency for the cluster of TCs that tracks through the Caribbean and the Gulf of Mexico. New hybrid models are developed to improve the prediction of observed basin-total TC and landfall TC frequencies. These models use large-scale climate predictors from the FLOR model as predictors for generalized linear models. The hybrid models show considerable improvements in the skill in predicting the basin-total TC frequencies relative to the dynamical model. The new hybrid model shows correlation coefficients as high as 0.75 for basinwide TC counts from the first two lead months and retains values around 0.50 even at the 6-month lead forecast. The hybrid model also shows comparable or higher skill in forecasting U.S. landfalling TCs relative to the dynamical predictions. The correlation coefficient is about 0.5 for the 2–5-month lead times.

Investigating the influence of Remote Climate Drivers as the Predictors in Forecasting South Australian spring rainfall

Australian rainfall is related with numerous key climate predictors namely El-Nino Southern Oscillation (ENSO), Indian Ocean Dipole (IOD) and Southern Annular Mode (SAM). Some studies have tried to discover the effects of these climate predictors on rainfall variability of different parts of Australia, particularly Western Australia, Queensland and Victoria. Nonetheless, clear association between separate or combined large-scale climate predictors and South Australian spring rainfall is yet to be established. Past studies showed that maximum rainfall predictability was only 20% considering isolated/individual effects of ENSO and SAM predictors in this region. The present study further explored these hypotheses by investigating two additional important aspects: investigating the relationship between lagged individual climate predictors with spring rainfall and linked (multiple combinations of ENSO and SAM) influences of significant lagged-climate indicators on spring rainfall forecasting using multiple regression (MR) modeling. Three stations were chosen as case studies for this region. MR models with combined-lagged climate predictors (SOI-SAM based models) showed better forecast ability in both model calibration and validation periods for all the stations. Results demonstrated that rainfall predictability significantly increased using combined climate predictor’s influence compared to their individual effect. It was discovered that rainfall predictability increased up-to 63% using combined climate predictors compared to their single influences. The maximum attained rainfall predictability for the SOI-SAM based models was 47% for calibration period that significantly enhanced with combined predictors influence to 97% during validation period. Therefore, MR analyses delineated the capabilities and influences of remote climate drivers in forecasting South Australian spring rainfall.

Long-Lead Water Supply Forecast Using Large-Scale Climate Predictors and Independent Component Analysis

Interest in water supply forecasting has grown prominently due to population growth and increasing demands for water. One important aspect of successfully managing the supply of water is accurate and reliable forecasts of seasonal streamflow volumes. Much of the streamflow in mountainous regions is a result of the collection of seasonal snowpack over the winter months and the melting of this snowpack over the spring and summer. However, there has been increasing pressure on operational agencies to issue longer-lead water supply forecasts that would be released in late fall or early winter preceding the runoff season. Longer-lead forecasts are difficult to make due to the uncertainty in future winter and spring climate conditions and the lack of snowpack information. During the late fall and early winter, large-scale oceanic and atmospheric information can provide insight into future climate conditions and spring runoff and have shown to be useful in developing long-lead forecast. In this study, statistical models are developed that incorporate large-scale climate signals into seasonal streamflow forecasting scheme. The objective of this study is to increase the skill of longer-lead water supply forecasts and also to increase the skill of current shorter-lead forecasts that rely on season-to-date hydroclimate information (e.g., precipitation since October, forecast month accumulated snowpack, and antecedent streamflow as a soil-moisture proxy). We employ a principal component regression model, with the addition of independent component analysis (ICA) climate predictors, for forecasting in two Pacific Northwest basins. It was found that incorporating large-scale climate predictors selected by ICA into a forecast model allowed a longer-lead-time forecast (beginning in September of previous year) and also contributed a significant amount of skill during the spring runoff season through April.

A statistical approach to downscaling of sub-daily extreme rainfall processes for climate-related impact studies in urban areas

This paper presents a spatial-temporal downscaling approach to describe the linkage between large-scale climate variables for daily scale to annual maximum (AM) precipitations for daily and sub-daily scales at a local site. More specifically, the proposed approach is based on a combination of a spatial downscaling method to link large-scale climate variables as provided by General Circulation Model (GCM) simulations with daily extreme precipitations at a local site and a temporal downscaling procedure to describe the relationships between daily extreme precipitations with sub-daily extreme precipitations using the scaling General Extreme Value (GEV) distribution. The feasibility of the proposed downscaling method has been tested based on climate simulation outputs from two GCMs under the A2 scenario (HadCM3A2 and CGCM2A2) and using available AM precipitation data for durations ranging from 5 minutes to 1 day at 15 raingage stations in Quebec (Canada) for the 1961–1990 period. Results of this numerical application has indicated that it is feasible to link large-scale climate predictors for daily scale given by GCM simulation outputs with daily and sub-daily AM precipitations at a local site. Furthermore, it was found that AM precipitations at a local site downscaled from the HadCM3A2 displayed a small change in the future, while those values estimated from the CGCM2A2 indicated a large increasing trend for future periods.

Generating streamflow forecasts for the Yakima River Basin using large-scale climate predictors

Multifaceted demands on western water supply, such as irrigation and mandated biological flows, coupled with climate variability are increasing the importance of supply forecasting to water managers. In this study, we develop and examine the accuracy of a seasonal ensemble streamflow forecasting model for the Yakima River Basin. The model incorporates large-scale climate information, related to the Pacific North American (PNA) pattern, with the objective of increasing the skill of the forecasts for water managers and stakeholders in the basin. Our study has found that spring runoff in the Yakima Basin is strongly correlated (correlation significant at the 5% significance level) with two of the large-scale circulation patterns associated with the PNA pattern from the preceding fall and winter seasons. Incorporating such climate information into our forecasts allowed a longer lead-time (a season in advance) planning period for water managers and stakeholders from the current practice of an April 1st forecast. The ensemble nature of our streamflow forecasts provides continuous probability distributions that will help the decision maker to objectively quantify the associated risk with selected streamflow values.

Multi-site downscaling of heavy daily precipitation occurrence and amounts

This paper compares three statistical models for downscaling heavy daily precipitation occurrence and amounts at multiple sites given lagged and contemporaneous large-scale climate predictors (such as atmospheric circulation, thickness, and moisture content at the surface, 850 and 500 hPa). Three models (a Radial Basis Function (RBF) Artificial Neural Network (ANN), Multi Layer Perceptron (MLP) ANN and a Conditional Resampling Method (SDSM)) were applied to area-average and station daily precipitation amounts in northwest (NWE) and southeast (SEE) England. Predictor selection via both stepwise multiple linear regression and compositing confirmed vorticity and humidity as important downscaling variables. Model skill was evaluated using indices of heavy precipitation for area averages, individual sites and inter-site behaviour. When tested against independent data (1979–1993), multi-site ANN models correctly simulated precipitation occurrence 80% of the time. The ANNs tended to over-estimate inter-site correlations for amounts due to their fully deterministic forcing, but performance was marginally better than SDSM for most seasonal-series of heavy precipitation indices. Conversely, SDSM yielded better inter-site correlation and representation of daily precipitation quantiles than the ANNs. All models had greatest skill for indices reflecting persistence of large-scale winter precipitation (such as maximum 5-day totals) or dry-spell duration in summer. Overall, predictability of daily precipitation was greater in NWE than SEE.