Dr. Chingka Kalai

 

Brief Details:

I am Chingka Kalai, Ph.D. in Water Resources Engineering from Indian Institute of Technology Bombay. I have pursued my Bachelor of Technology in Civil Engineering from NIT Agartala, and my Master of Technology (M.Tech.) in Water Resources Engineering and Management from NIT Karnataka (Surathkal). During the summer break of my Master's, I participated in an internship at NIH Belgam. Prior to joining as a Ph.D. student, I worked as an adhoc Assistant Professor in the Department of Civil Engineering at NIT Warangal after receiving M.Tech. degree. During my Ph.D., I have applied statistical and machine learning approaches to estimated floods at ungauged locations with a focus on seasonality and nonstationarity. Alongside the Centre for Ecology and Hydrology UK, we have worked on a project “Flood Estimation in Maharashtra” and applied some of the methods to real-world data in Maharashtra, India.

After my Ph.D., I have worked as a post-doctoral researcher at North Carolina State University, Raleigh, USA. My work was focused on the downscaling of the CMIP6 GCMs. During my stay at NC State University, I have dealt with large climatic GCM datasets and used Climate Data Operator (CDO) along with High-Performance Computing and parallel coding for computations. I have handled large datasets of NetCDF files and other formats. Later, I had to return to India due to some personal issues; however, I got another opportunity from UMass Amherst, USA, but could not join due to visa issues. Currently, I am working as a Postdoctoral Fellow at the International Centre of Excellence for Dams, IIT Roorkee, India, and my current focus is to forecast future water levels. Additionally, I have also performed analyses related to record-breaking rainfall and changes in the seasonality of rainfall in India.

Research: Approaches for improved regional flood frequency analysis within nonstationarity and seasonality frameworks

Regional Frequency Analysis (RFA) is an important tool for estimating flood magnitude at critical locations of interest. It involves (i) identification of sites with similar flood characteristics, called homogeneous regions, and (ii) pooling records from homogeneous sites to estimate the desired variable at the site of interest. Recent climate change and land-use modifications have drastically altered flood characteristics by inducing nonstationarity in flood magnitudes, leading to more uncertainty, thus signifying the necessity for a pooled RFA approach to obtain reliable estimates. However, to incorporate nonstationary flooding characteristics, the RFA approach needs to undergo certain modifications in its approaches. Also, the importance of floods seasonal due to their utility in several hydrological applications such as cleaning of the water conveyance system before flooding, yearly regulation of water control structures, and scheduling of crop rotations, calls for the implementation of RFA approach that could favour estimating flood seasonality for the ungauged site.

In the first part, a new test is proposed to validate homogeneous regions in the presence of nonstationarity known as the Proposed Homogeneity Test (PHT). Additionally, to incorporate the effect of nonstationarity in flood estimates, this work also proposes a modified Basu and Srinivas method (BSM*). The PHT shows better performance with H > 2 for heterogeneous regions while the flood estimates using BSM* produce lesser bias in the presence of nonstationarity. In the latter part, RFA framework is proposed for the estimation of flood seasonality at ungauged basins that consist of (i) selection of drivers or attributes that influence flood seasonality using correlation of similarity, circular Generalized Linear Models (GLM), and circular Classification and Regression Trees (CART). Then, (ii) Region of Influence (ROI) is employed for region identification, followed by validation of homogeneity using a proposed homogeneity test. (iii) Finally, the von Mises kernel is used for capturing the multimodality of flood seasonality at ungauged sites of interest. The approach estimates seasonality with absolute bias < 16 days for 75% of the sites.

Teaching: Frequency analysis and regional-frequency analysis of floods

Estimation of flood characteristics is beneficial for water resources utilities such as drinking water schemes, water control designs, culverts, water allocation strategies, and even identifying sites suitable for settlement. Reliable estimates of flood characteristics like its magnitudes can be obtained from observed data with long years of record. A frequency analysis of floods can provide reliable estimates of high flows that can have utility in several applications. However, the absence (or presence of a small amount) of data at several sites necessitates the requirement of a robust approach for the estimation of flood magnitude at these locations. Also, for example, the flood estimates with a recurrence interval of 100 years require a record length longer than or equal to 100 years. However, sites rarely have records of sufficient length, with most gauges being installed recently. Thus, the presentation will provide an overview of flood frequency at the site interest and deal with the situation when there is less or no data available for the site.

Google Scholar Link:

https://scholar.google.co.in/citations?user=qIcyjvYAAAAJ&hl=en