Authors: Abraham Elmir, Dr. Anjuli Jain Figueroa, Dr. Steven Gorelick
Droughts have been recognized by many governments and scientific researchers as one of the most impactful disasters affecting vulnerable communities, particularly affecting food, water and energy resources. Consequently, there is great interest in monitoring, predicting and preparing for droughts and their impacts. Droughts can be categorized as meteorological, hydrological, agricultural, and socioeconomic and their severity can be measured through indices. In this study, we compare and evaluate meteorological drought indices including Standardized Precipitation Index (SPI) and agricultural drought indices such as Normalized Difference Vegetation Index (NDVI) to measure drought-related stress to vegetation within the semi-arid, agricultural Bhīma Basin in India. Historical values of these indices are compared with future values based on modeled climate projections till 2100 to identify possible shifts and changes in drought severity caused by climate change.
Click HERE to view my slides.
Click HERE to view my video presentation.
●Anjuli Jain Figueroa
●Steven M Gorelick
●Abigail Nora Birnbaum
Great to work with you this summer, Abe! For the SPI calculations in the projected years, it seems like there is also a considerable amount of floods which will occur, as well. Could this cancel out the negative precipitation trend you are describing? Or would you say the periods of drought are enough to overtake the periods of flooding?
Abe – Nice job. I appreciate that you described what each acronym and other terminology meant. Very helpful. What do you think you would need to do to look at the seasonal signal? I don’t remember if you computed the monthly numbers – i”m wondering how the droughts might look at a 10-day scale, rather than the monthly (about 30-day) scale. – Jenny
To look at the seasonal scale through the data I would have to take into account when the monsoon season period is during the year towards the Bhima river basin. For the basin it occurs from July to September. The precipitation data for the historical and projected data time frames is based on a monthly analysis so it would be ideal to focus on looking at one specific month, such as September, since this is the month where the monsoon season ends and would be a good target month to notice trends in the severity of drought. If the the 30 day scale was reduced to a 10 day scale it would be more effective in measuring at agricultural drought than it would meteorological drought. I have read papers that SPI could be used to measure agricultural drought but it normally used to measure meterological drought and because of this a 30 day scale is more ideal.
I’m sad to see the regression of vegetation. But this is important work you’re doing. I’m curious more about the methods used in the projection of precipitation until 2035.
In terms of the methods for presenting the projection data, I received RCP rain data from a member of my Stanford lab, Peter Burik, which essentially is data that is meant to model precipitation trends towards the future. I uploaded that data via jupyter notebook and used python coding to present the visualization of the data as well as used a climate indices package to perform a SPI analysis with the variable needed which is precipitation in this case.
Great talk, and great work Abraham. Looks like you were able to come up with some interesting findings about a very important problem. I have a related question to the commenter above. Given how the future precipitation projections are made, should I interpret the prediction of droughts in the 2030s as a forecast for those specific years, or more as indicative of likely precipitation events in the future period? This seems to touch upon the difference between weather events and climate, which can be an important distinction to make.
So the forecast of the projected precipitation data via SPI analysis suggest more extremity in drought events to occur in the basin so the SPI indices is more of a measure of climate than it is weather pattern. with lower annual average rainfall, there can be expected severity of drought events in the future. Thank you for the great question!
Nice job on the presentation! I particularly like the graphics you provide like the NDVI anomaly map.
Here are some questions that came to mind.
From my immediate reaction, I feel like a ten year time period you chose is too short to conclude your results. The decrease of precipitation in the three year time period (2032 – 2034) you chose in the projection might just be due to the noise caused by the climate models. The signal to noise ratio for the time period seems a little too small. Why not pick 2090-2100 so any changes in precipitation would be much more pronounced? Are you sure about concluding that there is a drought from two ten-year-time series? Or do you think that the decrease in SPI between the historical time series and the projections a good enough evidence?
Otherwise, your NDVI anomaly map is really well done since it clearly shows the effect of the October 2018 drought in Maharashtra very well. I wonder how you obtain such high resolution data for the plot.
Some suggestions from me for future presentation is that you try to link more between the NDVI and SPI, and make figures or map that link the two. Maybe you pick 2018 as the example and you show both the two measures and maps for that year, so the link between meteorological and agricultural drought is more pronounced.
Thank you for your great questions. So yes I agree with you that a ten year time frame isn’t a ideal period of time to use when performing SPI analysis. I recently came upon a paper reviewing SPI analysis that suggested that a 30 year time frame would be optimal in measuring the extremity of drought. when I further work on this project I plan to a longer time frame. With your question about concluding that there is a drought from two ten-year-time series well there is actually consecutive years in both types of data where drought periods have occurred. The focus isn’t whether or not drought occur in the basin (which they do) it’s more trying to measure their severity and determine if they are increasing in extremity throughout the time frames. A decrease of average SPI when comparing historical and projected data is a good indicator but I agree with you that I should have used a longer time frame (30 years) for both historical and projected. With the NDVI anomaly map I was able to get high resolution data thanks to a NASA Modis image website. Also, Thank you for the suggestions I will take them in account when working to improve my project.
Congrats on graduating! I enjoyed watching your presentation! I was wondering on your graph of the projected precipitation, do you know what might account for the big spike in 2027 and 2031? Your final visual of the decrease in vegetation quite shocking. It definitely drove home the importance of the research you’re doing.
It was also a great experience working with you as well this summer!
To determine whether the periods of flooding can actually negate the effects of the drought seasons in both the historical and projected data would require to look at another type of climate indices that has been recommended to analyze flooding events. SPI is recommended as being a simple yet effective drought indices but in terms of measuring floods, it would be better to using a recommended flood index such as stream flow index that takes into account more variables than precipitation with SPI.