More research is needed for assessing the performance of precipitation datasets for hydrological applications.
See our short comment just posted on HESS journal and related to Beck et al. paper (https://www.hydrol-earth-syst-sci-discuss.net/hess-2018-481/):
https://www.hydrol-earth-syst-sci-discuss.net/hess-2018-481/hess-2018-481-SC3-print.pdf
More info here: https://twitter.com/Hydrology_IRPI/status/1057562077399826432
Please, give your comments and feedbacks, and participate to this 1-minute survey! https://twitter.com/Hydrology_IRPI/status/1057637422064783360
Hi Dear Brocca
I think it depends on the objective of a study or research, hydrological model type, the density of ground gauges, interpolation method to create gridded dataset from observed rainfalls, spatial and temporal resolution of RS based datasets and . . . . In one study we found that Era-Interim performs very well in estimating depth and spatial pattern of rainfall, when compared to PERSIAN-CDR and TRMM datasets. But in hydrological modeling by VIC-3L we found that PERSIAN-CDR led to better results in simulating streamflow hydrograph. Hydrological analyses by VIC-3L model showed that despite having low efficiency in estimating rainfall, the PERSIAN dataset led to better simulation of runoff when it compared to ECMWF. For example, the Nash-Sutcliffe (NS) coefficients between daily and monthly simulated runoff using PERSIAN and observed runoff at the outlet of SefidRood catchment are about 0.80 and 0.88, respectively, while in the case of Era-Interim these coefficients are about 0.67 and 0.72. In other words, in calibration process, the combination of input parameters compensate the low performance of rainfall datasets..
I added a pic that shows the performance of two rainfall datasets in simulation of stream flow...
the best
Azizian
Asghar Azizian That's exactly the point! And the results you have shown are very interesting.
We perform a lot of studies for satellite precipitation products validation through the comparison with "ground reference data" and some "statistical scores".
Then, we classify the products and we say which one is better.
However, this assessment is often not in agreement with the results we obtain by using satellite precipitation in "impact studies" (floods, landslides, water management, agriculture, ...)...exactly as you have shown.
Food for thought!!!
Hi Dear Brocca
I searched and found that Qi et al. 2016 state the same results that I found...
they concluded:
Although large uncertainty exists in heavy precipitation, hydrological models contribute most of the uncertainty in extreme discharges. Interactions between precipitation products and hydrological models can have the similar magnitude of contribution to discharge uncertainty as the hydrological models. A better precipitation product does not guarantee a better discharge simulation because of interactions. It is also found that a good discharge simulation depends
on a good coalition of a hydrological model and a precipitation product, suggesting that, although the satellite-based precipitation products are not as accurate as the gauge-based products, they could have better performance in discharge simulations when appropriately combined with hydrological models.
at the following link you can find their paper..
doi:10.5194/hess-20-903-2016
the best
Azizian
I believe that the structure of the error of our precipitation products is the reason for their performances in terms of flood reproduction. Statistical scores are likely not appropriate...we should investigate something new for this assessment...
We have developed a new rainfall estimation technique using Meteosat and GPM data. The algorithm is quite good for heavy rainfall events but the light rainfall event estimation has large errors. We have also reviewed all the available products and we came into conclusion that more research is need to estimate this highly variable phenomenon precisArticle Rainfall estimation techniques over India and adjoining ocea...
ely.Article Towards combining GPM and MFG observations to monitor near r...
In our recent paper, we elaborately assessed Machine learning based ensemble precipitation, satellite precipitation, and reanalysis precipitation with hydrological evaluation using multiple land surface and global hydrologic models. In that study we showed not only the score(skill score) but also many other factors, such co efficient of variation ratio, ensemble spreads etc are required to quantify precipitation uncertainty.
Article Assessment of precipitation error propagation in multi-model...
Sir, How we can do machine learning based ensemple precipiation analysis. Please kindly provide refeence papers.
Thanks and regards
You can go through our recent published paper
Article A nonparametric statistical technique for combining global p...
This paper is very interesting, particularly for me that you have shown the relevance of satellite 🛰 soil moisture for improving precipitation ☔
- Badges
- Science topic