Ensemble of Neural Network Emulations for Climate Model Physics: The Impact on Climate Simulations

Author(s):  
M.S. Fox-Rabinovitz ◽  
V. Krasnopolsky ◽  
A. Belochitski
2012 ◽  
Vol 5 (2) ◽  
pp. 313-319 ◽  
Author(s):  
Z. Song ◽  
F. Qiao ◽  
X. Lei ◽  
C. Wang

Abstract. This paper investigates the impact of the parallel computational uncertainty due to the round-off error on climate simulations using the Community Climate System Model Version 3 (CCSM3). A series of sensitivity experiments have been conducted and the analyses are focused on the Global and Nino3.4 average sea surface temperatures (SST). For the monthly time series, it is shown that the amplitude of the deviation induced by the parallel computational uncertainty is the same order as that of the climate system change. However, the ensemble mean method can reduce the influence and the ensemble member number of 15 is enough to ignore the uncertainty. For climatology, the influence can be ignored when the climatological mean is calculated by using more than 30-yr simulations. It is also found that the parallel computational uncertainty has no distinguishable effect on power spectrum analysis of climate variability such as ENSO. Finally, it is suggested that the influence of the parallel computational uncertainty on Coupled General Climate Models (CGCMs) can be a quality standard or a metric for developing CGCMs.


2008 ◽  
Vol 136 (10) ◽  
pp. 3683-3695 ◽  
Author(s):  
Vladimir M. Krasnopolsky ◽  
Michael S. Fox-Rabinovitz ◽  
Alexei A. Belochitski

An approach to calculating model physics using neural network emulations, previously proposed and developed by the authors, has been implemented in this study for both longwave and shortwave radiation parameterizations, or to the full model radiation, the most time-consuming component of model physics. The developed highly accurate neural network emulations of the NCAR Community Atmospheric Model (CAM) longwave and shortwave radiation parameterizations are 150 and 20 times as fast as the original/control longwave and shortwave radiation parameterizations, respectively. The full neural network model radiation was used for a decadal climate model simulation with the NCAR CAM. A detailed comparison of parallel decadal climate simulations performed with the original NCAR model radiation parameterizations and with their neural network emulations is presented. Almost identical results have been obtained for the parallel decadal simulations. This opens the opportunity of using efficient neural network emulations for the full model radiation for decadal and longer climate simulations as well as for weather prediction.


2012 ◽  
Vol 12 (20) ◽  
pp. 9441-9458 ◽  
Author(s):  
A. M. M. Manders ◽  
E. van Meijgaard ◽  
A. C. Mues ◽  
R. Kranenburg ◽  
L. H. van Ulft ◽  
...  

Abstract. Climate change may have an impact on air quality (ozone, particulate matter) due to the strong dependency of air quality on meteorology. The effect is often studied using a global climate model (GCM) to produce meteorological fields that are subsequently used by chemical transport models. However, climate models themselves are subject to large uncertainties and fail to reproduce the present-day climate adequately. The present study illustrates the impact of these uncertainties on air quality. To this end, output from the SRES-A1B constraint transient runs with two GCMs, i.e. ECHAM5 and MIROC-hires, has been dynamically downscaled with the regional climate model RACMO2 and used to force a constant emission run with the chemistry transport model LOTOS-EUROS in a one-way coupled run covering the period 1970–2060. Results from the two climate simulations have been compared with a RACMO2-LOTOS-EUROS (RLE) simulation forced by the ERA-Interim reanalysis for the period 1989–2009. Both RLE_ECHAM and RLE_MIROC showed considerable deviations from RLE_ERA for daily maximum temperature, precipitation and wind speed. Moreover, sign and magnitude of these deviations depended on the region. The differences in average present-day concentrations between the simulations were equal to (RLE_MIROC) or even larger than (RLE_ECHAM) the differences in concentrations between present-day and future climate (2041–2060). The climate simulations agreed on a future increase in average summer ozone daily maximum concentrations of 5–10 μg m−3 in parts of Southern Europe and a smaller increase in Western and Central Europe. Annual average PM10 concentrations increased 0.5–1.0 μg m−3 in North-West Europe and the Po Valley, but these numbers are rather uncertain: overall, changes for PM10 were small, both positive and negative changes were found, and for many locations the two climate runs did not agree on the sign of the change. This illustrates that results from individual climate runs can at best indicate tendencies and should therefore be interpreted with great care.


2012 ◽  
Vol 12 (5) ◽  
pp. 12245-12285 ◽  
Author(s):  
A. M. M. Manders ◽  
E. van Meijgaard ◽  
A. C. Mues ◽  
R. Kranenburg ◽  
L. H. van Ulft ◽  
...  

Abstract. Climate change may have an impact on air quality (ozone, particulate matter) due to the strong dependency of air quality on meteorology. The effect is often studied using a global climate model (GCM) to produce meteorological fields that are subsequently used by chemical transport models. However, climate models themselves are subject to large uncertainties and fail to adequately reproduce the present-day climate. The present study illustrates the impact of this uncertainty on air quality. To this end, output from the SRES-A1B constraint transient runs with two GCMs, i.e. ECHAM5 and MIROC-hires, has been dynamically downscaled with the regional climate model RACMO2 and used to force a constant emission run with the chemistry transport model LOTOS-EUROS in a one-way coupled run covering the period 1970–2060. Results from the two climate simulations have been compared with a RACMO2-LOTOS-EUROS (RLE) simulation forced by the ERA-Interim reanalysis for the period 1989–2009. Both RLE_ECHAM and RLE_MIROC showed considerable deviations from RLE_ERA in daily maximum temperature, precipitation and wind speed. Moreover, sign and magnitude of these deviations depended on the region. Differences in average concentrations for the present-day simulations were found of equal to (RLE_MIROC) or even larger than (RLE_ECHAM) the differences in concentration between present-day and future climate (2041–2060). The climate simulations agreed on a future increase in average summer ozone daily maximum concentrations (5–10 μg m−3) in parts of Southern Europe and a smaller increase in Western and Central Europe. Annual average PM10 concentrations increased (0.5–1.0 μg m−3) in North-West Europe and the Po Valley, but these numbers are rather uncertain. Overall, changes for PM10 were small, both positive and negative changes were found, and for many locations the two runs did not agree on the sign of the change. The approach taken here illustrates that results from individual climate runs can at best indicate tendencies and should therefore be interpreted with great care.


2011 ◽  
Vol 4 (4) ◽  
pp. 3295-3312
Author(s):  
Z. Song ◽  
F. Qiao ◽  
X. Lei ◽  
C. Wang

Abstract. This paper investigates the impact of the parallel computational uncertainty on climate simulations using the Community Climate System Model Version 3 (CCSM3). A series of sensitivity experiments have been conducted and the analyses are focused on the Global and Nino3.4 sea surface temperatures. It is shown that the amplitude of the deviation induced by the parallel computational uncertainty is the same order as that of the climate system change. However, the ensemble mean method can reduce the influence and the ensemble member number of 15 is enough to ignore simulated errors. For climatology, the influence can be ignored when the climatological mean is calculated by using more than 30-yr simulations. It is also found that the parallel computational uncertainty has no effect on the simulated periods of climate variability such as ENSO. Finally, it is suggested that the influence of the parallel computational uncertainty on Coupled General Climate Models (CGCMs) can be a quality standard or a metric for developing CGCMs.


Climate ◽  
2020 ◽  
Vol 8 (3) ◽  
pp. 44
Author(s):  
Nadia Politi ◽  
Athanasios Sfetsos ◽  
Diamando Vlachogiannis ◽  
Panagiotis T. Nastos ◽  
Stylianos Karozis

In the present study, the ability of the Advanced Weather Research and Forecasting numerical model (WRF-ARW) to perform climate regionalization studies in the topographically complex region of Greece, was examined in order to explore the possibility of a more reliable selection of physical schemes for the simulation of historical and future high resolution (5 km) climate model experiments to investigate the impact of climate change. This work is directly linked to a previous study investigating the performance of seven different model setups for one year, from which the need was derived for further examination of four different simulations to investigate the model sensitivity on the representation of surface variables statistics during a 5-year period. The results have been compared with observational data for maximum and minimum air temperature and daily precipitation through statistical analysis. Clear similarities were found in precipitation patterns among simulations and observations, yielding smoothly its inter-annual variability, especially during the wettest months and summer periods, with the lowest positive percentage BIAS calculated at about 19% for the selected combination of physics parameterizations (PP3). Regarding the maximum and minimum temperature, statistical analysis showed a high correlation above 0.9, and negative bias around 1−1.5 °C, and positive bias near 2 °C, respectively.


2020 ◽  
Vol 11 (1) ◽  
Author(s):  
Yunqian Zhu ◽  
Owen B. Toon ◽  
Eric J. Jensen ◽  
Charles G. Bardeen ◽  
Michael J. Mills ◽  
...  

Abstract Volcanic ash is often neglected in climate simulations because ash particles are assumed to have a short atmospheric lifetime, and to not participate in sulfur chemistry. After the Mt. Kelut eruption in 2014, stratospheric ash-rich aerosols were observed for months. Here we show that the persistence of super-micron ash is consistent with a density near 0.5 g cm−3, close to pumice. Ash-rich particles dominate the volcanic cloud optical properties for the first 60 days. We also find that the initial SO2 lifetime is determined by SO2 uptake on ash, rather than by reaction with OH as commonly assumed. About 43% more volcanic sulfur is removed from the stratosphere in 2 months with the SO2 heterogeneous chemistry on ash particles than without. This research suggests the need for re-evaluation of factors controlling SO2 lifetime in climate model simulations, and of the impact of volcanic ash on stratospheric chemistry and radiation.


2005 ◽  
Vol 133 (5) ◽  
pp. 1370-1383 ◽  
Author(s):  
Vladimir M. Krasnopolsky ◽  
Michael S. Fox-Rabinovitz ◽  
Dmitry V. Chalikov

Abstract A new approach based on a synergetic combination of statistical/machine learning and deterministic modeling within atmospheric models is presented. The approach uses neural networks as a statistical or machine learning technique for an accurate and fast emulation or statistical approximation of model physics parameterizations. It is applied to development of an accurate and fast approximation of an atmospheric longwave radiation parameterization for the NCAR Community Atmospheric Model, which is the most time consuming component of model physics. The developed neural network emulation is two orders of magnitude, 50–80 times, faster than the original parameterization. A comparison of the parallel 10-yr climate simulations performed with the original parameterization and its neural network emulations confirmed that these simulations produce almost identical results. The obtained results show the conceptual and practical possibility of an efficient synergetic combination of deterministic and statistical learning components within an atmospheric climate or forecast model. A developmental framework and practical validation criteria for neural network emulations of model physics components are outlined.


2020 ◽  
Vol 39 (6) ◽  
pp. 8927-8935
Author(s):  
Bing Zheng ◽  
Dawei Yun ◽  
Yan Liang

Under the impact of COVID-19, research on behavior recognition are highly needed. In this paper, we combine the algorithm of self-adaptive coder and recurrent neural network to realize the research of behavior pattern recognition. At present, most of the research of human behavior recognition is focused on the video data, which is based on the video number. At the same time, due to the complexity of video image data, it is easy to violate personal privacy. With the rapid development of Internet of things technology, it has attracted the attention of a large number of experts and scholars. Researchers have tried to use many machine learning methods, such as random forest, support vector machine and other shallow learning methods, which perform well in the laboratory environment, but there is still a long way to go from practical application. In this paper, a recursive neural network algorithm based on long and short term memory (LSTM) is proposed to realize the recognition of behavior patterns, so as to improve the accuracy of human activity behavior recognition.


Sign in / Sign up

Export Citation Format

Share Document