Supported by a research grant from NOAA, this project has introduced an aggregated extreme sea level indicator for the contiguous U.S. coast comprised of separate indicators for mean sea level (MSL) and storm surge climatology (SSC). Non-Stationary extreme value analysis is performed to derive multi-decadal variability of SSC and employed K-means clustering to identify regions of coherent variability. This research highlights the important role of interannual to multi-decadal variability in different sea level components in exacerbating, or reducing, the impacts of long-term MSL rise over time scales relevant for planning and management of coastal cities. Different models are developed to predict the SSC indicator using climate indices, sea level pressure, and sea surface temperature as predictor variables. Predictability of the indicator using initialized decadal climate model experiments are also been assessed. Additionally, I have also been working to assess coastal flood risk from concurrent or successive events of multiple drivers such as storm surge, high tide, extreme rainfall, and river discharge using copulas. I look forward to understad how these flood events are modulated by the large scale climate variability using weather typing.
This research investigates the relationships of drought and reanalysis climate variables filtered to specific frequency bands using wavelet transforms. The key innovation of this study is to filter the predictor variables using wavelets to enhance the information content at the relevant frequencies for droughts that has potential application in monitoring and projection of future droughts from GCMs datasets. Additionally, the merit of distorting the time aggregation of large scale climate indices (Nino3.4 and DMI) have investigated before casting them in a drought predictive model. Aggregated climate indices are used to predict sustained drought and wet anomalies characterized here using a drought index (i.e. Standardize Precipitation Index, SPI) as a response and Australia as the study region of interest. The aim is to enhance the strength of relationships of drought index and climate indices (predictors) by tuning the frequency of climate predictors using an aggregation technique. The result shows that aggregated climate indices provide significant improvement in the prediction of SPI over raw climate indices across Australia.
Droughts are often monitored based on the changes of aggregated hydro-climatic variables e.g. rainfall, soil moisture, and streamflow. Although hydro-climatic time series are often nonstationary, traditional standardized drought monitoring indices, for example, Standardized Precipitation Index (SPI) do not take nonstationary into account make these indices unreliable for monitoring droughts in changing climates. Therefore, I developed a nonstationary SPI (NSPI) within the Generalized Additive Model in Location, Scale and Shape (GAMLSS) modeling framework and demonstrated that the NSPI outperform to identify droughts in a changing climate.
Hydroclimatic variable such as rainfall is non-stationary and low frequency variability (i.e. persistence) is vital to reproduce hydrological extreme events such as flood and drought. This suggests that a non-stationary model capable to reproduce low frequency variability of rainfall is essential. Therefore, I developed a noble downscaling approach where predictand (i.e. rainfall) and predictors (i.e. climate variables from reanalysis and /or GCM outputs) are first decomposed by wavelet transforms (WT) and rainfall is modeled using the semi-parametric additive model formulation of GAMLSS. When compared with the traditional downscaling approach (i.e. application of GAMLSS to the original rainfall series without wavelet decomposition), the proposed model outperforms at reproducing the time-frequency properties of the observed rainfall and still preserve the statistics produced by traditional modelling approach.
This project investigates the future changes in hydrologically relevant rainfall metrics compare to the historical period. I developed a multi-site statistical downscaling model using Generalized Linear Model (GLM). Logistic regression was used to identify the wet days and an amount model was fitted using a gamma distribution to estimate the amount of rain on wet days. I downscaled rainfall for future periods from CMIP5 General Circulation Models (GCMs). I developed an innovative Frequency Adapted Quantile Mapping (FAQM) technique for bias correction of the downscaled rainfall. Additionally, I developed a nonstationary model to estimate the return periods of annual maximum daily rainfall (AMDR). I conducted this project as a part of the largest ever climate change research project in South Australia and was funded by Goyder Institute for Water Research.
Monitoring droughts with non-stationary drought index
Modeling of sustained drought and wet anomalies
Statistical downscaling: non-stationary and multi-resolution approach
Statistical downscaling of GCMs output to rainfall
Rainfall-streamflow model integrating wavelet and GAMLSS models
A rainfall–streamflow model is proposed, in which a downscaled rainfall series and it's wavelet-based decomposed sub-series at optimum lags were used as covariates in GAMLSS (Generalized Additive Model in Location, Scale and Shape). GAMLSS is applied in climate change impact assessment using CMIP5 general climate model to simulate daily streamflow in three sub-catchments of the Onkaparinga catchment, South Australia. Study shows that the GAMLSS has the capability to capture non-stationarity in the rainfall–streamflow process. It was also observed that the use of wavelet-based decomposed rainfall sub-series with optimum lags as covariates in the GAMLSS model captures the underlying physics of the rainfall–streamflow process. The development and application of an empirical rainfall–streamflow model that can be used to assess the impact of catchment-scale climate change on streamflow is demonstrated.
Risk of extreme sea level and compound flood
Characterization of droughts in a multivariate perspective
This study considers the Standardized Precipitation Index (SPI) at time scales of 3, 6, 12 and 24 months to assess both short‐term and long‐term droughts in South Australia (SA). Trends in the SPI at 46 selected rainfall stations across South Australia (SA) are investigated. Furthermore, various drought parameters such as frequency, severity, and duration are estimated, and their spatial distributions are presented. Drought risks are estimated in terms of copula based bivariate joint return periods and compared with traditional univariate return periods. Finally, the dominant climatic modes of sustained drought and wet anomalies are identified using correlation analysis. The results highlight the suitability of a bivariate joint return period over a univariate return period for drought risk assessment because the drought is a multi‐faceted phenomenon. The dominant climatic modes identified in this study are Niño3.4 and the dipole mode index (DMI), which are able to represent the SPI across SA at the selected time scales.
Frequency level trend detection technique using wavelet decomposition
This study focuses on identifying existing trends in annual, seasonal and monthly rainfall at thirteen stations in the Onkaparinga catchment in South Australia during the period 1960-2010. A new trend detection approach was developed, which combines a Continuous Wavelet Transform (CWT) with the Mann Kendall (MK) test. The original rainfall time series was decomposed to different periodic components using a CWT and then the MK test was applied to detect the trends. The study showed that the periodic components (generally low-frequency signals) might have significant trends even when there are no significant trends in the original data. Thus signifies the potential of this trend detection approach to detect climate change compare to the conventional approach of considering the original series of hydroclimatic variables. A sequential Mann-Kendall analysis was found useful for identifying the trend turning points. Most of the trends, whether positive or negative, started during the mid-1970s to mid-1980s. The technique developed in this study may also be applied for trend detection of other hydro-climatic variables in other catchments, particularly where temporal and spatial variabilities are high.