19 December 2021 | By: Karisma Yumnam, M.Tech Department of Hydrology, IIT Roorkee

Journey towards my master thesis and publishing it in a reputed journal

When I first come to IIT Roorkee, I was sure that I want to pursue a career in research but I did not know how to do a research. At that time, research sounds very complicated and felt like it requires exceptional knowledge and skills to do one. And I wanted to learn that skill. I wanted to learn “How to do a research?”.

At first, I did not know anything about coding. MATLAB was the first programming I got exposed to during the second semester. My research started with reading journal papers on the “Evaluation of Satellite Precipitation products”, shared by my supervisor Dr. Ankit Agarwal. My first challenge was working with the “netcdf” file format. One day, Prof. Ankit taught us how to extract the “netcdf” file and explained the code line by line. After that, it took me few days to fully understand the file format and how to work with them.

The next step was to merge the Satellite Precipitation Products (SPPs). The SPPs are indirect precipitation data derived from various satellite sensors. Though they are relatively inexpensive, they suffer greatly in their accuracy while representing the actual ground precipitation. Few researchers proposed different merging algorithms. The first one that I got into was the “Dynamic Bayesian Model Averaging”. When I first read the paper, I knew nothing. The method was too complicated for me. Prof. Ankit suggested to work with Ravi Kumar Guntu, PhD student. He took some time to read the paper. Then, he explained me step by step very clearly. I still remember how we video call on Whatsapp for hours and how he wrote the formulas with pen on a paper and tried to show me.

Then, Ravi Kumar shared a R code to implement the method. He taught me how to run the code. I did that and when I saw the result for the first time, I was very happy. Then, the next day, I found out that I was evaluating with the wrong reference data. When I actually evaluate with the right reference data, the result was not what I wanted. Well, that was demotivating. Then, we thought to develop the code in MATLAB only as we are more familiar with it. Then, it took me 2 days to write the code in MATLAB. I was skeptical as it was my first time writing a major algorithm. Then, I discuss every line of the code with Ravi. He helped me correct and refine the algorithm. Well, I ran the algorithm and the results were yet again disappointing. By this point, I completely lost faith in my work and thought that it was unachievable.

When I came back from Germany, Prof. Ankit asked he to prepare a manuscript for the merging work. Then, I prepared the first draft and shared in a group email. Previously, I did modify the method to eliminate the zeros, used more satellite precipitation products and tried evaluating. Then, I started working with Ravi Kumar on the manuscript. First, I explained him how I separate the data based on quantiles and how I develop the Bayesian Model Averaging weights. The idea of quantiles was to separate the wet days and the dry days and also to separate medium rainfall from heavy rainfall events. There was discussion with Ravi Kumar, Prof. Ankit and Prof. Maheswaran, which ended on a positive note. After numerous versions of the manuscript, Prof. Ankit finally submitted the final version of the manuscript to the Journal of Hydrology. It is one of the top reputed journal in my domain. After two revisions, the paper was finally accepted.

I was in euphoria when I saw the acceptance email. I think happiness has different forms and it was a different kind. I felt happy and grateful at the same time: Happy to know that the idea makes sense to the scientific community; Grateful that I found supportive and enthusiastic people, people who believed in me, and that they have helped me reach this point. Having a paper published gave me the confidence in my work and I started believing in myself too.

At the end of my masters, I see research in a different light. I think it is more about communicating, learning and growing together. A person cannot do research alone. We need to learn from different people, share ideas and expertise, support each other and grow together. Research is no longer a mystery to me. Now, I believe that I can do a good research in future. And I am very pleased to have chosen Prof. Ankit Agarwal and his group.

Lastly, thank you COPREPARE team for giving me the opportunity to share my thoughts.

Yumnam, Karisma, Ravi Kumar Guntu, Maheswaran Rathinasamy, and Ankit Agarwal. "Quantile-based Bayesian Model Averaging approach towards merging of precipitation products." Journal of Hydrology (2021): 127206.