Empirical-statistical downscaling and error correction of regional climate models and its impact on the climate change signal

Realizing the error characteristics of regional climate models (RCMs) and the consequent limitations in their direct utilization in climate change impact research, this study analyzes a quantile-based empirical-statistical error correction method (quantile mapping, QM) for RCMs in the context of climate change. In particular the success of QM in mitigating systematic RCM errors, its ability to generate “new extremes” (values outside the calibration range), and its impact on the climate change signal (CCS) are investigated. In a cross-validation framework based on a RCM control simulation over Europe, QM reduces the bias of daily mean, minimum, and maximum temperature, precipitation amount, and derived indices of extremes by about one order of magnitude and strongly improves the shapes of the related frequency distributions. In addition, a simple extrapolation of the error correction function enables QM to reproduce “new extremes” without deterioration and mostly with improvement of the original RCM quality. QM only moderately modifies the CCS of the corrected parameters. The changes are related to trends in the scenarios and magnitude-dependent error characteristics. Additionally, QM has a large impact on CCSs of non-linearly derived indices of extremes, such as threshold indices.     

Distributional changes in rainfall and river flow in Sarawak,Malaysia

Climate change may not change the rainfall mean, but the variability and extremes. Therefore, it is required to explore the possible distributional changes of rainfall characteristics over time. The objective of present study is to assess the distributional changes in annual and northeast monsoon rainfall (November-January) and river flow in Sarawak where small changes in rainfall or river flow variability/distribution may have severe implications on ecology and agriculture. A quantile regression-based approach was used to assess the changes of scale and location of empirical probability density function over the period 1980-2014 at 31 observational stations. The results indicate that diverse variation patterns exist at all stations for annual rainfall but mainly increasing quantile trend at the lowers, and higher quantiles for the month of January and December. The significant increase in annual rainfall is found mostly in the north and central-coastal region and monsoon month rainfalls in the interior and north of Sarawak. Trends in river flow data show that changes in rainfall distribution have affected higher quantiles of river flow in monsoon months at some of the basins and therefore more flooding. The study reveals that quantile trend can provide more information of rainfall change which may be useful for climate change mitigation and adaptation planning.     

A new daily weather generator to preserve extremes and low-frequency variability

This paper addresses deficiencies of stochastic Weather Generators (WGs) in terms of reproduction of low-frequency variability and extremes, as well as the unanticipated effects of changes to precipitation occurrence under climate change scenarios on secondary variables. A new weather generator (named IWG) is developed in order to resolve such deficiencies and improve WGs performance. The proposed WG is composed of three major components, including a stochastic rainfall model able to reproduce realistic rainfall series containing extremes and inter-annual monthly variability, a multivariate daily temperature model conditioned to the rainfall occurrence, and a suitable multi-variate monthly generator to fit the low-frequency variability of daily maximum and minimum temperature series. The performance of IWG was tested by comparing statistical characteristics of the simulated and observed weather data, and by comparing statistical characteristics of the simulated runoff outputs by a daily rainfall-runoff model fed by the generated and observed weather data. Furthermore, IWG outputs are compared with those of the well-known LARS-WG weather generator. The tested characteristics are a variety of different daily statistics, low-frequency variability, and distribution of extremes. It is concluded that the performance of the IWG is acceptable, better than LARS-WG in the majority of tests, especially in reproduction of extremes and low-frequency variability of weather and runoff series.     

Asymmetric impact of rainfall on India’s food grain production: evidence from quantile autoregressive distributed lag model

This study used a quantile autoregressive distributed lag (QARDL) model to capture asymmetric impact of rainfall on food production in India. It was found that the coefficient corresponding to the rainfall in the QARDL increased till the 75th quantile and started decreasing thereafter, though it remained in the positive territory. Another interesting finding is that at the 90th quantile and above the coefficients of rainfall though remained positive was not statistically significant and therefore, the benefit of high rainfall on crop production was not conclusive. However, the impact of other determinants, such as fertilizer and pesticide consumption, is quite uniform over the whole range of the distribution of food grain production.     

Assessment of a climate model to reproduce rainfall variability and extremes over Southern Africa

It is increasingly accepted that any possible climate change will not only have an influence on mean climate but may also significantly alter climatic variability. A change in the distribution and magnitude of extreme rainfall events (associated with changing variability), such as droughts or flooding, may have a far greater impact on human and natural systems than a changing mean. This issue is of particular importance for environmentally vulnerable regions such as southern Africa. The sub-continent is considered especially vulnerable to and ill-equipped (in terms of adaptation) for extreme events, due to a number of factors including extensive poverty, famine, disease and political instability. Rainfall variability and the identification of rainfall extremes is a function of scale, so high spatial and temporal resolution data are preferred to identify extreme events and accurately predict future variability. The majority of previous climate model verification studies have compared model output with observational data at monthly timescales. In this research, the assessment of ability of a state of the art climate model to simulate climate at daily timescales is carried out using satellite-derived rainfall data from the Microwave Infrared Rainfall Algorithm (MIRA). This dataset covers the period from 1993 to 2002 and the whole of southern Africa at a spatial resolution of 0.1° longitude/latitude. This paper concentrates primarily on the ability of the model to simulate the spatial and temporal patterns of present-day rainfall variability over southern Africa and is not intended to discuss possible future changes in climate as these have been documented elsewhere. Simulations of current climate from the UK Meteorological Office Hadley Centre’s climate model, in both regional and global mode, are firstly compared to the MIRA dataset at daily timescales. Secondly, the ability of the model to reproduce daily rainfall extremes is assessed, again by a comparison with extremes from the MIRA dataset. The results suggest that the model reproduces the number and spatial distribution of rainfall extremes with some accuracy, but that mean rainfall and rainfall variability is under-estimated (over-estimated) over wet (dry) regions of southern Africa.     

Validation of TRMM Precipitation Radar satellite data over Indonesian region

Research has been conducted to validate monthly and seasonal rain rates derived from the Tropical Rainfall Measuring Mission Precipitation Radar (PR) using rain gauge data analysis from 2004 to 2008. The study area employed 20 gauges across Indonesia to monitor three Indonesian regional rainfall types. The relationship of PR and rain gauge data statistical analysis included the linear correlation coefficient, the mean bias error (MBE), and the root mean square error (RMSE). Data validation was conducted with point-by-point analysis and spatial average analysis. The general results of point-by-point analysis indicated satellite data values of medium correlation, while values of MBE and RMSE tended to indicate underestimations with high square errors. The spatial average analysis indicated the PR data values are lower than gauge values of monsoonal and semi-monsoonal type rainfall, while anti-monsoonal type rainfall was overestimated. The validation analysis showed very good correlation with the gauge data of monsoonal type rainfall, high correlation for anti-monsoonal type rainfall, but medium correlation for semi-monsoonal type rainfall. In general, the statistical error level of monthly seasonal monsoonal type conditions is more stable compared to other rainfall types. Unstable correlations were observed in months of high rainfall for semi-monsoonal and anti-monsoonal type rainfall.     

Regional analysis of annual maximum rainfall using TL-moments method

Information related to distributions of rainfall amounts are of great importance for designs of water-related structures. One of the concerns of hydrologists and engineers is the probability distribution for modeling of regional data. In this study, a novel approach to regional frequency analysis using L-moments is revisited. Subsequently, an alternative regional frequency analysis using the TL-moments method is employed. The results from both methods were then compared. The analysis was based on daily annual maximum rainfall data from 40 stations in Selangor Malaysia. TL-moments for the generalized extreme value (GEV) and generalized logistic (GLO) distributions were derived and used to develop the regional frequency analysis procedure. TL-moment ratio diagram and Z-test were employed in determining the best-fit distribution. Comparison between the two approaches showed that the L-moments and TL-moments produced equivalent results. GLO and GEV distributions were identified as the most suitable distributions for representing the statistical properties of extreme rainfall in Selangor. Monte Carlo simulation was used for performance evaluation, and it showed that the method of TL-moments was more efficient for lower quantile estimation compared with the L-moments.     

Uncovering consistencies in Indian rainfall trends observed over the last half century

Food security in India is tightly linked to rainfall variability. Trends in Indian rainfall records have been extensively studied but the subject remains complicated by the high spatiotemporal variability of rainfall arising from complex atmospheric dynamics. For various reasons past studies have often produced inconsistent results. This paper presents an analysis of recent trends in monthly and seasonal cumulative rainfall depth, number of rainy days and maximum daily rainfall, and in the monsoon occurrence (onset, peak and retreat). A modified version of the Mann-Kendall test, accounting for the scaling effect, was applied to 29 variables derived from square-degree-resolution daily gridded rainfall (1951–2007). The mapping of gridded trend slopes and the regional average Kendall test were used concurrently to assess the field significance of regional trends in areas exhibiting spatial homogeneity in trend directions. The statistics we used account for temporal and spatial correlations, and thus reduce the risk of overestimating the significance of local and regional trends. Our results i/ improve available knowledge (e.g. 5 %-field-significant delay of the monsoon onset in Northern India); ii/ provide a solid statistical basis to previous qualitative observations (e.g. 1 %-field-significant increase/decrease in pre-monsoon rainfall depth in northeast/southwest India); and, iii/ when compared to recent studies, show that the field significance level of regional trends (e.g. in rainfall extremes) is test-dependent. General trend patterns were found to align well with the geography of anthropogenic atmospheric disturbances and their effect on rainfall, confirming the paramount role of global warming in recent rainfall changes.     

Evaluation of the CORDEX regional climate models performance in simulating climate conditions of two catchments in Upper Blue Nile Basin

This study was targeted at evaluating the performance of six Regional Climate Models (RCMs) used in Coordinated Regional Climate Downscaling Experiment (CORDEX). The evaluation is on the bases of how well the RCMs simulate the seasonal mean climatology, interannual variability and annual cycles of rainfall, maximum and minimum temperature over two catchments in western Ethiopia during the period 1990–2008. Observed data obtained from the Ethiopian National Meteorological Agency was used for performance evaluation of the RCMs outputs. All Regional Climate Models (RCMs) have simulated seasonal mean annual cycles of precipitation with a significant bias shown on individual models; however, the ensemble mean exhibited better the magnitude and seasonal rainfall. Despite the highest biases of RCMs in the wet season, the annual cycle showed the prominent features of precipitation in the two catchments. In many aspects, CRCM5 and RACMO22 T simulate rainfall over most stations better than the other models. The highest biases are associated with the highest error in simulating maximum and minimum temperature with the highest biases in high elevation areas. The rainfall interannual variability is less evident in Finchaa with short rainy season experiencing a larger degree of interannual variability. The differences in performance of the Regional Climate Models in the two catchments show that all the available models are not equally good for particular locations and topographies. In this regard, the right regional climate models have to be used for any climate change impact study for local-scale climate projections.     

Statistical analysis of annual maximum rainfall in North-East India: an application of LH-moments

An attempt has been made to determine the best fitting distribution to describe the annual series of maximum daily rainfall data for the period 1966 to 2007 of nine distantly located stations in North East India. The LH-moments of order zero (L) to order four (L₄) are used to estimate the parameters of three extreme value distributions viz. generalized extreme value distribution (GEV), generalized logistic distribution (GLD), and generalized Pareto distribution (GPD). The performances of the distributions are assessed by evaluating the relative bias (RBIAS) and relative root mean square error (RRMSE) of quantile estimates through Monte Carlo simulations. Then, the boxplot is used to show the location of the median and the associated dispersion of the data. Finally, it can be revealed from the results of boxplots that zero level of LH-moments of the generalized Pareto distribution would be appropriate to the majority of the stations for describing the annual maximum rainfall series in North East India.     

具有对数正态分布几何过程的统计推断

在假定几何过程的初始分布为对数正态分布的前提下 ,研究了几何过程的统计推断问题。对其中的参数分别用极大似然法和改进的矩法进行了估计。然后 ,在理论推导、模拟试验和实际数据分析的基础上提出了一些建议。     

Seasonal variability of rainfall extremes

Monthly rainfall extremes have been analyzed for three stations in Southern Ontario. The double exponential probability distribution was fitted to the extreme values for each month considered, each duration selected, and sets of annual extremes. A station‐year approach yielded monthly and annual extreme value distributions for the lumped region of Southern Ontario. The analysis has revealed a pronounced seasonal pattern in the rainfall extremes – the amount of rain expected with a selected probability of occurrence during the summer being considerably greater than the rainfall that might be expected to be exceeded at the same probability level during the spring or fall. The extent of the seasonal variability was found also to vary with duration. The implications of the variability are seen to be significant for the estimation of the magnitude and frequency of floods.     

A statistical model to downscale local daily temperature extremes from synoptic-scale atmospheric circulation patterns in the Australian region

 The study seeks to describe one method of deriving information about local daily temperature extremes from larger scale atmospheric flow patterns using statistical tools. This is considered to be one step towards downscaling coarsely gridded climate data from global climate models (GCMs) to finer spatial scales. Downscaling is necessary in order to bridge the spatial mismatch between GCMs and climate impact models which need information on spatial scales that the GCMs cannot provide. The method of statistical downscaling is based on physical interaction between atmospheric processes with different spatial scales, in this case between synoptic scale mean sea level pressure (MSLP) fields and local temperature extremes at several stations in southeast Australia. In this study it was found that most of the day-to-day spatial variability of the synoptic circulation over the Australian region can be captured by six principal components. Using the scores of these PCs as multivariate indicators of the circulation a substantial part of the local daily temperature variability could be explained. The inclusion of temperature persistence noticeably improved the performance of the statistical model. The model established and tested with observations is thought to be finally applied to GCM-simulated pressure fields in order to estimate pressure-related changes in local temperature extremes under altered CO₂ conditions. Received: 26 March 1996 / Accepted: 20 September 1996     

Climate projections of future extreme events accounting for modelling uncertainties and historical simulation biases

A methodology is presented for providing projections of absolute future values of extreme weather events that takes into account key uncertainties in predicting future climate. This is achieved by characterising both observed and modelled extremes with a single form of non-stationary extreme value (EV) distribution that depends on global mean temperature and which includes terms that account for model bias. Such a distribution allows the prediction of future “observed” extremes for any period in the twenty-first century. Uncertainty in modelling future climate, arising from a wide range of atmospheric, oceanic, sulphur cycle and carbon cycle processes, is accounted for by using probabilistic distributions of future global temperature and EV parameters. These distributions are generated by Bayesian sampling of emulators with samples weighted by their likelihood with respect to a set of observational constraints. The emulators are trained on a large perturbed parameter ensemble of global simulations of the recent past, and the equilibrium response to doubled CO₂. Emulated global EV parameters are converted to the relevant regional scale through downscaling relationships derived from a smaller perturbed parameter regional climate model ensemble. The simultaneous fitting of the EV model to regional model data and observations allows the characterisation of how observed extremes may change in the future irrespective of biases that may be present in the regional models simulation of the recent past climate. The clearest impact of a parameter perturbation in this ensemble was found to be the depth to which plants can access water. Members with shallow soils tend to be biased hot and dry in summer for the observational period. These biases also appear to have an impact on the potential future response for summer temperatures with some members with shallow soils having increases for extremes that reduce with extreme severity. We apply this methodology for London, using the A1B future emissions scenario to obtain projections of the 50 year return values for the 20 year period centred on 2050. We obtain 10th to 90th percentile ranges of 35.9–42.1 °C for summer daily maximum temperature, 35.5–52.4 mm for summer daily rainfall and 79.2, 97.0 mm for autumn 5 day total rainfall, compared to observed estimates for 1961–1990 of 35.7 °C, 42.1 and 78.4 mm respectively.     

Storylines: an alternative approach to representing uncertainty in physical aspects of climate change

Investigating the relationships between climate extremes and crop yield can help us understand how unfavourable climatic conditions affect crop production. In this study, two statistical models, multiple linear regression and random forest, were used to identify rainfall extremes indices affecting wheat yield in three different regions of the New South Wales wheat belt. The results show that the random forest model explained 41–67% of the year-to-year yield variation, whereas the multiple linear regression model explained 34–58%. In the two models, 3-month timescale standardized precipitation index of Jun.–Aug. (SPI_JJA), Sep.–Nov. (SPI_SON), and consecutive dry days (CDDs) were identified as the three most important indices which can explain yield variability for most of the wheat belt. Our results indicated that the inter-annual variability of rainfall in winter and spring was largely responsible for wheat yield variation, and pre-growing season rainfall played a secondary role. Frequent shortages of rainfall posed a greater threat to crop growth than excessive rainfall in eastern Australia. We concluded that the comparison between multiple linear regression and machine learning algorithm proposed in the present study would be useful to provide robust prediction of yields and new insights of the effects of various rainfall extremes, when suitable climate and yield datasets are available.     

Downscaling daily precipitation over the Yellow River source region in China: a comparison of three statistical downscaling methods

Three statistical downscaling methods are compared with regard to their ability to downscale summer (June–September) daily precipitation at a network of 14 stations over the Yellow River source region from the NCEP/NCAR reanalysis data with the aim of constructing high-resolution regional precipitation scenarios for impact studies. The methods used are the Statistical Downscaling Model (SDSM), the Generalized LInear Model for daily CLIMate (GLIMCLIM), and the non-homogeneous Hidden Markov Model (NHMM). The methods are compared in terms of several statistics including spatial dependence, wet- and dry spell length distributions and inter-annual variability. In comparison with other two models, NHMM shows better performance in reproducing the spatial correlation structure, inter-annual variability and magnitude of the observed precipitation. However, it shows difficulty in reproducing observed wet- and dry spell length distributions at some stations. SDSM and GLIMCLIM showed better performance in reproducing the temporal dependence than NHMM. These models are also applied to derive future scenarios for six precipitation indices for the period 2046–2065 using the predictors from two global climate models (GCMs; CGCM3 and ECHAM5) under the IPCC SRES A2, A1B and B1scenarios. There is a strong consensus among two GCMs, three downscaling methods and three emission scenarios in the precipitation change signal. Under the future climate scenarios considered, all parts of the study region would experience increases in rainfall totals and extremes that are statistically significant at most stations. The magnitude of the projected changes is more intense for the SDSM than for other two models, which indicates that climate projection based on results from only one downscaling method should be interpreted with caution. The increase in the magnitude of rainfall totals and extremes is also accompanied by an increase in their inter-annual variability.     

Precipitation and temperature space–time variability and extremes in the Mediterranean region: evaluation of dynamical and statistical downscaling methods

This study evaluates how statistical and dynamical downscaling models as well as combined approach perform in retrieving the space–time variability of near-surface temperature and rainfall, as well as their extremes, over the whole Mediterranean region. The dynamical downscaling model used in this study is the Weather Research and Forecasting (WRF) model with varying land-surface models and resolutions (20 and 50 km) and the statistical tool is the Cumulative Distribution Function-transform (CDF-t). To achieve a spatially resolved downscaling over the Mediterranean basin, the European Climate Assessment and Dataset (ECA&D) gridded dataset is used for calibration and evaluation of the downscaling models. In the frame of HyMeX and MED-CORDEX international programs, the downscaling is performed on ERA-I reanalysis over the 1989–2008 period. The results show that despite local calibration, CDF-t produces more accurate spatial variability of near-surface temperature and rainfall with respect to ECA&D than WRF which solves the three-dimensional equation of conservation. This first suggests that at 20–50 km resolutions, these three-dimensional processes only weakly contribute to the local value of temperature and precipitation with respect to local one-dimensional processes. Calibration of CDF-t at each individual grid point is thus sufficient to reproduce accurately the spatial pattern. A second explanation is the use of gridded data such as ECA&D which smoothes in part the horizontal variability after data interpolation and damps the added value of dynamical downscaling. This explains partly the absence of added-value of the 2-stage downscaling approach which combines statistical and dynamical downscaling models. The temporal variability of statistically downscaled temperature and rainfall is finally strongly driven by the temporal variability of its forcing (here ERA-Interim or WRF simulations). CDF-t is thus efficient as a bias correction tool but does not show any added-value regarding the time variability of the downscaled field. Finally, the quality of the reference observation dataset is a key issue. Comparison of CDF-t calibrated with ECA&D dataset and WRF simulations to local measurements from weather stations not assimilated in ECA&D, shows that the temporal variability of the downscaled data with respect to the local observations is closer to the local measurements than to ECA&D data. This highlights the strong added-value of dynamical downscaling which improves the temporal variability of the atmospheric dynamics with regard to the driving model. This article highlights the benefits and inconveniences emerging from the use of both downscaling techniques for climate research. Our goal is to contribute to the discussion on the use of downscaling tools to assess the impact of climate change on regional scales.     

Frequency analysis of short‐duration rainfalls

Abstract

An investigation was carried out to determine the appropriateness of a probability distribution and fitting technique commonly used in Canada for rainfall frequency analysis.

The extreme value type 1 (EV1) distribution was assessed for three long‐term Canadian stations: Victoria, St Thomas and Québec. The EV1 distribution appears to provide a reasonable fit for durations varying from 5 min to 6 h, but is not clearly superior to another two‐parameter distribution, the lognormal.

The fitting technique, known as modified moments or regression, was assessed by comparing it with three other fitting techniques: moments, maximum likelihood and adjusted maximum likelihood. This comparison was carried out using Monte Carlo simulation techniques over the parameter space deemed to be representative of short duration rainfall data for Canada. In terms of estimating rainfall amounts of specified return period, the modified moments technique was the poorest with regard to both bias and efficiency. In general, the maximum likelihood estimates were the most efficient and were relatively unbiased.     

Downscaling over Vietnam using the stretched-grid CCAM: verification of the mean and interannual variability of rainfall

Rainfall over Vietnam is highly variable from north to south, due to the interaction of the monsoonal winds with the terrain. There is high rainfall from April to September, and little rainfall from October to March (except along the central Vietnam coast). In order to study the ability of the Commonwealth Scientific and Industrial Research Organisation stretched-grid Conformal Cubic Atmospheric Model (CCAM) to capture the climatic and interannual variability of rainfall, downscaled simulations at approximately 20 km horizontal resolution over the region were produced for the period 1979–2001. A scale-selective digital filter was used to force the winds, temperature and sea-level pressure from the ERA-Interim reanalysis for length scales greater than about 700 km. For wind and temperature, the forcing is applied for pressure-sigma levels above about 0.9. ERA-Interim sea surface temperatures were used over the oceans. The simulations were primarily validated against the gridded Asian Precipitation Highly Resolved Observational Data Integration Toward Evaluation of the Water Resources rainfall dataset and station observations using standard statistical methods. It was found that CCAM reproduces well the amount and spatial variability of rainfall, with an area-averaged bias for the entire study domain of less than 1 mm day^?1; CCAM is also able to capture the rainfall pattern under different El Niño Southern Oscillation phases reasonably well for the dry season. For interannual variability, the simulation generally performed better for North and Central Vietnam than for South Vietnam, where rainfall variability was overestimated.     

Verification of a scattering-based algorithm for estimating rainfall over the open ocean

Summary An algorithm for estimating rainfall (based upon scattering of upwelling radiation at 85 GHz) using the Special Sensor Microwave Imager (SSM/I) 85 GHz sensor data has recently been developed at the National Oceanographic and Atmospheric Administration's Office of Research and Applications (NOAA ORA). Spatial and temporal variability in the differences between simple areal-averaged rainfall estimates from a raingage and satellite-derived estimates are examined. These initial comparisons provide insight into the nature of the differences; however, due to substantial amounts of spatial sampling error in the raingage spatial averages, this approach is ineffective at identifying statistically significant differences.Thus, rainfall estimates for this algorithm are examined over the tropical Pacific using a statistical technique, called the noncontiguous raingage method (NCR), which minimizes spatial sampling error so that the statistical significance of relationships may be determined. The results indicate that, adjusted with calibration coefficients, the ORA algorithm could produce unbiased estimates of areal rainfall for the tropical Pacific region examined in this study.With 8 Figures