I currently have monthly precipitation and potential evapotranspiration for each month in my 17-year study. I calculated the water balance for each month by subtracting PET from precipitation. I then summed the water balance in three month chunks since I want the 3-month SPEI.
From here, I'm confused about calculating the probability-weighted moments (PWM) of the water balance series, which is necessary to determine SPEI.
My question is: How do I calculate the 3-month SPEI across my entire 17-year time scale? I understand how to calculate the alpha, beta, and gamma coefficients, etc. I attached a picture of the PWM formula (from Wagan 2015 "Drought Trends and Temperature Influence in Zhanghe River Basin, China".) From what I understand, the Ws is computed at different orders of s, and these results are inputs to the formulas for alpha, beta, gamma coefficients, which are themselves inputs into the SPEI formula.
In terms of Di, how do I combine my time series into this formula? For instance, I have 17 files for the 3-month June water balance, which is the sum of the April, May and June water balance in a given year. In this PWM equation, I imagine I would include all 17 years for June (and then would run the equation again for all 17 data points for July, etc.) If I did that though, then how would I determine SPEI for a particular year? Do I use this equation twice, once for the whole time series, and then again for the particular 3 month span I'm interested in?
I know there is SPEI R code, but this code cannot handle NaNs and I have a lot of NaNs. These cannot be removed since the location of the data values is very important. Thanks for any feedback.
http://spei.csic.es/home.html
https://www.researchgate.net/publication/238101515_CLIMATE_CHANGE_IMPACT_ON_ARMENIAN_WATER_RESOURCES?ev=prf_high
Have a look at this article of mine
Article CLIMATE CHANGE IMPACT ON ARMENIAN WATER RESOURCES
Dear Dion: From your message I think that your calculations are ok. Answering to your question "how do I combine my time series into this formula?", the procedure is as follows:
1) You have time series of D, isn´t it?
2) You should calculate F(x) being the xi = Di, that is, apply the log-logistic distribution with your time series.
3) p (i) = 1 - F(x), then you will have the probability of exceedance for your D time series and following the log-logistic distribution.
4) w(i)=sqrt(-2*log(p(i)) (valid for p(i) lower than 0.5)
5) By applying the formula suggested by Vicente-Serrano (SPEI web page) you will have a good estimation of the SPEI.
I hope these suggestions are useful for your study. If you have any question, please do not hesitate to contact me.
Also, I suggest you to read the attached work for the calculation of the 3-month SPEI for large regions. Some outputs could also be useful for you. Finally, in two weeks aprox. we will have an accepted paper related to this topic, and we will be happy to share it.
Regards, Alvaro Sordo-Ward
Conference Paper MACRO SCALE ANALYSIS OF DROUGHTS IN A CLIMATE CHANGE CONTEXT
Book BASE - ADAPTATION INSPIRATION BOOK 23 European cases of clim...
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts