EGU General Assembly 2020
© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Comparing the diurnal cycle of precipitation in models and observations at different spatial scales over China

Mark Muetzelfeldt, Reinhard Schiemann, Andrew Turner, Nicholas Klingaman, and Pier Luigi Vidale
Mark Muetzelfeldt et al.
  • National Centre for Atmospheric Science, University of Reading, Reading, United Kingdom (

Climate models have a long-standing bias in the diurnal cycle of precipitation over land - they produce peak rainfall at local midday, when insolation is at its maximum. As part of the COnvective Scale Modelling In China (COSMIC) project, we investigate this bias over China using high-resolution (13 km) global simulations with the HadGEM3 model. We compare the diurnal cycle of summer precipitation with satellite observations of precipitation from CMORPH. The simulations are run with and without a convection parametrization scheme, as this scheme has been shown to be important for controlling the timing of precipitation. We analyse the amount, frequency and intensity of the precipitation, investigating their diurnal cycle and spatial distribution.

The analysis is performed on a grid-point scale, as well as at larger scales based on the catchment basins across the region. Catchment basins provide a natural way of linking the meteorological precipitation data to the underlying physical geography of the region, in a way which is useful for decision makers and could be used to provide information to hydrological models in the future. We present a simple Python tool for performing the analysis: BAsin-Scale Model Assessment ToolkIt (BASMATI).

In line with previous studies, we find that the simulation performed with parametrized convection produces precipitation over land which peaks too early in the day. The simulation performed with explicit convection generally produces peaks in precipitation which occur later in the day - closer in time to the observed peak. By comparing our results with published work, we find that the presence or absence of a convection parametrization scheme is more important for determining the spatial distribution of the time of peak precipitation than the resolution of the simulations. We present comparisons of precipitation in the simulations and observations performed at grid points and over catchment basins using BASMATI. The catchment basins are chosen based on their size, which allows for the comparison to be done as a function of spatial scale.

How to cite: Muetzelfeldt, M., Schiemann, R., Turner, A., Klingaman, N., and Vidale, P. L.: Comparing the diurnal cycle of precipitation in models and observations at different spatial scales over China, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-7846,, 2020


Display file

Comments on the display

AC: Author Comment | CC: Community Comment | Report abuse

displays version 1 – uploaded on 01 May 2020
  • CC1: sensitivity of skill score to choice of observational product, Margot bador, 06 May 2020

    Hi Mark, thank you for presenting your work here, nice study!

    I was wondering if you had tested the sensitivy of your skill scores to the choice of your observational product (CMORPH)? It might be possible to do that only over your large basins but it would be interesting to check if you still find that higher resolution generally leads to lower skill scores for mean JJA precipitation? I say this because over Asia CMORPH is generally wetter than the regional in situ dataset APHRODITE, which is available at 0.5°x0.5° resolution.

    • AC1: Reply to CC1, Mark Muetzelfeldt, 06 May 2020

      Hi Margot,

      Thanks for your comment. I have done some preliminary analysis with APHRODITE, but only crudely comparing by eye the mean 2009 JJA precipitation over SE China with CMORPH - see slide 17 (in the Bonus Content section). The comparison is between CMORPH (0.25°), APHRODITE (0.25°) and gauge stations take from Li et al. 2018 with a Cressman filter applied ( It certainly bears out what you said - that CMORPH appears to be wetter than APHRODITE over Asia. However, when compared to the gauge data, APHRODITE seems to be drier (perhaps as a consequence of producing a gridded dataset? Perhaps the underlying guage station data is different?).

      We could do some quantative analysis of skill scores using APHRODITE over different basin sizes as well, but it would be restricted to the daily mean. As a part of our study was concerned with the sub-daily diurnal cycle, we didn't progress beyond doing a simple comparison of CMORPH vs APHRODITE mean precipitation mentioned above.

      • CC2: Reply to AC1, Margot bador, 07 May 2020

        Hi Mark,

        Great that you already have your hands on APHRODITE. Even if your analyses with it are limited to daily data it could be good to compare the skill scores w.r.t CMORPH and APHRODITE when you can. It is important to better understand the impact of increasing spatial resolution in climate models and your results tend to show that it will not necessarily lead to a systematic improvement in the simulation of precipitation. This is also our general conclusion from a recent paper (soon to be published in JGRatm) where we evaluate HighResMIP simulations for precipitation extremes at the global scale. However, we also show that observational uncertainties are important, and that they make the evaluation of models difficult, which is why I am suggesting to use a second obs dataset when possible.

        Regarding the differences between APHRODITE and gauge stations, you can indeed expect a dry bias in APHRODITE compared to station data just because of the mismatch in scale, i.e. gridded data representing an average over a grid cell compared to point-based estimates at the station scale.

        Anyway, I'll be interested to have a copy of your paper when it is ready. Thanks!

        • AC2: Reply to CC2, Mark Muetzelfeldt, 07 May 2020

          Hi Margot,

          Thanks for your perspective on this, and for your input. I would say that if APHRODITE had sub-daily data, we probably would have done the analysis already. It would be good to compare with another (non-satellite) observational dataset in a more rigorous manner though - I will have a look at the basin-scale RMSEs of the precipitation in simulations vs APHRODITE (time permitting!).

          Your upcoming paper also sounds very interesting - please let me know when it is available and I will check it out.