Search Results
You are looking at 1 - 10 of 17 items for
- Author or Editor: Yongyun Hu x
- Refine by Access: All Content x
Abstract
Regional sea surface temperature (SST) mode variabilities, especially the La Niña–like Pacific Ocean temperature pattern known as the negative phase of the interdecadal Pacific oscillation (IPO) and the associated heat redistribution within the ocean, are the leading mechanisms explaining the recent global warming hiatus. Here version 1 of the Community Earth System Model (CESM) is used to examine how different phases of two leading decadal time scale SST modes, namely the IPO and the Atlantic multidecadal oscillation (AMO), contribute to heat redistribution in the global ocean in the absence of time-evolving external forcings. The results show that both the IPO and AMO contribute a similar magnitude to global mean surface temperature and ocean heat redistribution. Both modes contribute warmer surface temperature and higher upper ocean heat content in their positive phase, and the reverse in their negative phase. Regionally, patterns of ocean heat distribution in the upper few hundred meters of the tropical and subtropical Pacific Ocean depend highly on the IPO phase via the IPO-associated changes in the subtropical cell. In the Atlantic, ocean heat content is primarily associated with the state of the AMO. The interconnections between the IPO, AMO, and global ocean heat distribution are established through the atmospheric bridge and the Atlantic meridional overturning circulation. An in-phase variant of the IPO and AMO can lead to much higher surface temperatures and heat content changes than an out-of-phase variation. This result suggests that changes in the IPO and AMO are potentially capable of modulating externally forced SST and heat content trends.
Abstract
Regional sea surface temperature (SST) mode variabilities, especially the La Niña–like Pacific Ocean temperature pattern known as the negative phase of the interdecadal Pacific oscillation (IPO) and the associated heat redistribution within the ocean, are the leading mechanisms explaining the recent global warming hiatus. Here version 1 of the Community Earth System Model (CESM) is used to examine how different phases of two leading decadal time scale SST modes, namely the IPO and the Atlantic multidecadal oscillation (AMO), contribute to heat redistribution in the global ocean in the absence of time-evolving external forcings. The results show that both the IPO and AMO contribute a similar magnitude to global mean surface temperature and ocean heat redistribution. Both modes contribute warmer surface temperature and higher upper ocean heat content in their positive phase, and the reverse in their negative phase. Regionally, patterns of ocean heat distribution in the upper few hundred meters of the tropical and subtropical Pacific Ocean depend highly on the IPO phase via the IPO-associated changes in the subtropical cell. In the Atlantic, ocean heat content is primarily associated with the state of the AMO. The interconnections between the IPO, AMO, and global ocean heat distribution are established through the atmospheric bridge and the Atlantic meridional overturning circulation. An in-phase variant of the IPO and AMO can lead to much higher surface temperatures and heat content changes than an out-of-phase variation. This result suggests that changes in the IPO and AMO are potentially capable of modulating externally forced SST and heat content trends.
Abstract
A slowdown in the rate of surface warming in the early 2000s led to renewed interest in the redistribution of ocean heat content (OHC) and its relationship with internal climate variability. We use the Community Earth System Model version 1 to study the relationship between OHC and the interdecadal Pacific oscillation (IPO), a major mode of decadal sea surface temperature variability in the Pacific Ocean. By comparing the relative contributions of surface heat flux and ocean dynamics to changes in OHC for different phases of the IPO, we try to identify the underlying physical processes involved. Our results suggest that during IPO phase transitions, changes of 0–300-m OHC across the northern extratropical Pacific are positively contributed by both surface heat flux and oceanic heat transport. By contrast, oceanic heat transport appears to drive the OHC changes in equatorial Pacific whereas surface heat flux acts as a damping term. During a positive IPO phase, weakened wind-driven circulation acts to increase the OHC in the equatorial Pacific while the enhanced evaporation acts to damp OHC anomalies. In the Kuroshio–Oyashio Extension region, a dipole anomaly of zonal heat advection amplifies an OHC dipole anomaly that moves eastward, while strong turbulent heat fluxes act to dampen this OHC anomaly. In the northern subtropical Pacific, both the wind-driven evaporation change and the change of zonal heat advection along Kuroshio Extension contribute to the OHC change during phase transition. For the northern subpolar Pacific, both surface heat flux and enhanced meridional advection contribute to the positive OHC anomalies during the positive IPO phase.
Abstract
A slowdown in the rate of surface warming in the early 2000s led to renewed interest in the redistribution of ocean heat content (OHC) and its relationship with internal climate variability. We use the Community Earth System Model version 1 to study the relationship between OHC and the interdecadal Pacific oscillation (IPO), a major mode of decadal sea surface temperature variability in the Pacific Ocean. By comparing the relative contributions of surface heat flux and ocean dynamics to changes in OHC for different phases of the IPO, we try to identify the underlying physical processes involved. Our results suggest that during IPO phase transitions, changes of 0–300-m OHC across the northern extratropical Pacific are positively contributed by both surface heat flux and oceanic heat transport. By contrast, oceanic heat transport appears to drive the OHC changes in equatorial Pacific whereas surface heat flux acts as a damping term. During a positive IPO phase, weakened wind-driven circulation acts to increase the OHC in the equatorial Pacific while the enhanced evaporation acts to damp OHC anomalies. In the Kuroshio–Oyashio Extension region, a dipole anomaly of zonal heat advection amplifies an OHC dipole anomaly that moves eastward, while strong turbulent heat fluxes act to dampen this OHC anomaly. In the northern subtropical Pacific, both the wind-driven evaporation change and the change of zonal heat advection along Kuroshio Extension contribute to the OHC change during phase transition. For the northern subpolar Pacific, both surface heat flux and enhanced meridional advection contribute to the positive OHC anomalies during the positive IPO phase.
Abstract
This paper is a continuation of the study of the advection–diffusion problem for stratospheric flow, and deals with the probability distribution function (PDF) of gradients of a freely decaying passive tracer. Theoretical arguments are reviewed and extended showing that mixing of a weakly diffused tracer by random large-scale flows produces a tracer gradient field whose probability distribution function has “stretched exponential” tails P(|∇θ|) ∝ exp(−b|∇θ| γ ) with γ < 1. This contrasts with the lognormal distribution expected for advective mixing in the absence of diffusion. The non-Gaussian distribution of tracer gradients can be derived in terms of the statistics of strain rates of the random driving flow. It is shown that the tails of the gradient PDF provide information about the dissipation scale, the scale selectivity of the dissipation law, and the fluctuations of short-term strain. The gradient PDF is shown to contain information about tracer variability that is not present at all in the power spectrum of the tracer field.
To show that the predictions remain valid for the gradient statistics of passive tracers driven by the well-organized lower-stratospheric flow with mixing barriers, a series of advection–diffusion simulations of a decaying passive tracer are presented. The mixing is driven by ECMWF winds on the 420-K isentropic surface using the high-resolution finite-volume model employed in Part I of this paper. It is found that the probability distribution function of the simulated tracer gradients is indeed stretched exponential, with the stretching parameter γ ≈ 0.55. The largest gradients are not found in the regions of highest Lyapunov exponents, but rather in the surf-zone regions adjacent to the reservoirs of high tracer fluctuation amplitude.
Abstract
This paper is a continuation of the study of the advection–diffusion problem for stratospheric flow, and deals with the probability distribution function (PDF) of gradients of a freely decaying passive tracer. Theoretical arguments are reviewed and extended showing that mixing of a weakly diffused tracer by random large-scale flows produces a tracer gradient field whose probability distribution function has “stretched exponential” tails P(|∇θ|) ∝ exp(−b|∇θ| γ ) with γ < 1. This contrasts with the lognormal distribution expected for advective mixing in the absence of diffusion. The non-Gaussian distribution of tracer gradients can be derived in terms of the statistics of strain rates of the random driving flow. It is shown that the tails of the gradient PDF provide information about the dissipation scale, the scale selectivity of the dissipation law, and the fluctuations of short-term strain. The gradient PDF is shown to contain information about tracer variability that is not present at all in the power spectrum of the tracer field.
To show that the predictions remain valid for the gradient statistics of passive tracers driven by the well-organized lower-stratospheric flow with mixing barriers, a series of advection–diffusion simulations of a decaying passive tracer are presented. The mixing is driven by ECMWF winds on the 420-K isentropic surface using the high-resolution finite-volume model employed in Part I of this paper. It is found that the probability distribution function of the simulated tracer gradients is indeed stretched exponential, with the stretching parameter γ ≈ 0.55. The largest gradients are not found in the regions of highest Lyapunov exponents, but rather in the surf-zone regions adjacent to the reservoirs of high tracer fluctuation amplitude.
Abstract
Using NCEP–NCAR 51-yr reanalysis data, the interannual and decadal variations of planetary wave activity and its relationship to stratospheric cooling, and the Northern Hemisphere Annular mode (NAM), are studied. It is found that winter stratospheric polar temperature is highly correlated on a year-to-year basis with the Eliassen–Palm (E–P) wave flux from the troposphere, implying a dynamical control of the former by the latter, as often suggested. Greater (lower) wave activity from the troposphere implies larger (smaller) poleward heat flux into the polar region, which leads to warmer (colder) polar temperature. A similar highly correlated antiphase relationship holds for E–P flux divergence and the strength of the polar vortex in the stratosphere. It is tempting to extrapolate these relationships found for interannual timescales to explain the recent stratospheric polar cooling trend in the past few decades as caused by decreased wave activity in the polar region. This speculation is not supported by the data. On timescales of decades the cooling trend is not correlated with the trend in planetary wave activity. In fact, it is found that planetary wave amplitude, E–P flux, and E–P flux convergence all show little statistical evidence of decrease in the past 51 yr, while the stratosphere is experiencing a cooling trend and the NAM index has a positive trend during the past 30 yr. This suggests that the trends in the winter polar temperature and the NAM index can reasonably be attributed to the radiative cooling of the stratosphere, due possibly to increasing greenhouse gases and ozone depletion. It is further shown that the positive trend of the NAM index in the past few decades is not through the inhibition of upward planetary wave propagation from the troposphere to the stratosphere, as previously suggested.
Abstract
Using NCEP–NCAR 51-yr reanalysis data, the interannual and decadal variations of planetary wave activity and its relationship to stratospheric cooling, and the Northern Hemisphere Annular mode (NAM), are studied. It is found that winter stratospheric polar temperature is highly correlated on a year-to-year basis with the Eliassen–Palm (E–P) wave flux from the troposphere, implying a dynamical control of the former by the latter, as often suggested. Greater (lower) wave activity from the troposphere implies larger (smaller) poleward heat flux into the polar region, which leads to warmer (colder) polar temperature. A similar highly correlated antiphase relationship holds for E–P flux divergence and the strength of the polar vortex in the stratosphere. It is tempting to extrapolate these relationships found for interannual timescales to explain the recent stratospheric polar cooling trend in the past few decades as caused by decreased wave activity in the polar region. This speculation is not supported by the data. On timescales of decades the cooling trend is not correlated with the trend in planetary wave activity. In fact, it is found that planetary wave amplitude, E–P flux, and E–P flux convergence all show little statistical evidence of decrease in the past 51 yr, while the stratosphere is experiencing a cooling trend and the NAM index has a positive trend during the past 30 yr. This suggests that the trends in the winter polar temperature and the NAM index can reasonably be attributed to the radiative cooling of the stratosphere, due possibly to increasing greenhouse gases and ozone depletion. It is further shown that the positive trend of the NAM index in the past few decades is not through the inhibition of upward planetary wave propagation from the troposphere to the stratosphere, as previously suggested.
Abstract
Using NCEP–NCAR reanalysis data, decadal trends in planetary wave activity in Northern Hemisphere high latitudes (50°–90°N) in late winter and early spring (January–February–March) were studied. Results show that wave activity in both the stratosphere and the troposphere has been largely reduced and exhibits statistically significant downward trends since the 1980s. In the stratosphere, the wave activity is decreased by about 30%, which is mainly due to less Eliassen–Palm (E–P) flux from the troposphere into the stratosphere. In the troposphere, the vertical E–P flux is reduced by about 30%, while equatorward horizontal E–P flux is increased by 130%. This suggests a significant refraction of planetary waves away from the high latitudes. The significant negative trends in wave activity in late winter are in contrast to the authors' previous finding of no significant changes in planetary wave activity in early winter.
The timing of the significant decline in wave activity, which starts from the early 1980s and exists only in late winter and springtime, suggests that such a decrease of wave activity is possibly a result of stratospheric ozone depletion in the Arctic. Therefore, a mechanism is proposed whereby Arctic ozone depletion leads to an enhanced meridional temperature gradient near the subpolar stratosphere, strengthening westerly winds. The strengthened winds refract planetary waves toward low latitudes and cause the reduction in wave activity in high latitudes.
Decreasing vertical E–P fluxes are found to extend to near the surface. At 850 mb, vertical E–P fluxes have been reduced by about 10% since 1979. Such a reduction in wave activity might be responsible for the observed late-winter and springtime warming over Northern Hemisphere high-latitude continents during the last two decades.
Abstract
Using NCEP–NCAR reanalysis data, decadal trends in planetary wave activity in Northern Hemisphere high latitudes (50°–90°N) in late winter and early spring (January–February–March) were studied. Results show that wave activity in both the stratosphere and the troposphere has been largely reduced and exhibits statistically significant downward trends since the 1980s. In the stratosphere, the wave activity is decreased by about 30%, which is mainly due to less Eliassen–Palm (E–P) flux from the troposphere into the stratosphere. In the troposphere, the vertical E–P flux is reduced by about 30%, while equatorward horizontal E–P flux is increased by 130%. This suggests a significant refraction of planetary waves away from the high latitudes. The significant negative trends in wave activity in late winter are in contrast to the authors' previous finding of no significant changes in planetary wave activity in early winter.
The timing of the significant decline in wave activity, which starts from the early 1980s and exists only in late winter and springtime, suggests that such a decrease of wave activity is possibly a result of stratospheric ozone depletion in the Arctic. Therefore, a mechanism is proposed whereby Arctic ozone depletion leads to an enhanced meridional temperature gradient near the subpolar stratosphere, strengthening westerly winds. The strengthened winds refract planetary waves toward low latitudes and cause the reduction in wave activity in high latitudes.
Decreasing vertical E–P fluxes are found to extend to near the surface. At 850 mb, vertical E–P fluxes have been reduced by about 10% since 1979. Such a reduction in wave activity might be responsible for the observed late-winter and springtime warming over Northern Hemisphere high-latitude continents during the last two decades.
Abstract
Current climate models project that Antarctic sea ice will decrease by the end of the twenty-first century. Previous studies have suggested that Antarctic sea ice changes have impacts on atmospheric circulation and the mean state of the Southern Hemisphere. However, little is known about whether Antarctic sea ice loss may have a tangible impact on climate extremes over the southern continents and whether ocean–atmosphere coupling plays an important role in changes of climate extremes over the southern continents. In this study, we conduct a set of fully coupled and atmosphere-only model experiments forced by present and future Antarctic sea ice cover. It is found that the projected Antarctic sea ice loss by the end of the twenty-first century leads to an increase in the frequency and duration of warm extremes (especially warm nights) over the southern continents and a decrease in cold extremes over most regions. The frequency and duration of wet extremes are projected to increase over South America and Antarctica, whereas changes in dry days and the longest dry spell vary with regions. Further Antarctic sea ice loss under a quadrupling of CO2 leads to similar but larger changes. Comparison between the coupled and atmosphere-only model experiments suggests that ocean dynamics and their interactions with the atmosphere induced by Antarctic sea ice loss play a key role in driving the identified changes in temperature and precipitation extremes over southern continents. By comparing with global warming experiments, we find that Antarctic sea ice loss may affect temperature and precipitation extremes for some regions under greenhouse warming, especially Antarctica.
Abstract
Current climate models project that Antarctic sea ice will decrease by the end of the twenty-first century. Previous studies have suggested that Antarctic sea ice changes have impacts on atmospheric circulation and the mean state of the Southern Hemisphere. However, little is known about whether Antarctic sea ice loss may have a tangible impact on climate extremes over the southern continents and whether ocean–atmosphere coupling plays an important role in changes of climate extremes over the southern continents. In this study, we conduct a set of fully coupled and atmosphere-only model experiments forced by present and future Antarctic sea ice cover. It is found that the projected Antarctic sea ice loss by the end of the twenty-first century leads to an increase in the frequency and duration of warm extremes (especially warm nights) over the southern continents and a decrease in cold extremes over most regions. The frequency and duration of wet extremes are projected to increase over South America and Antarctica, whereas changes in dry days and the longest dry spell vary with regions. Further Antarctic sea ice loss under a quadrupling of CO2 leads to similar but larger changes. Comparison between the coupled and atmosphere-only model experiments suggests that ocean dynamics and their interactions with the atmosphere induced by Antarctic sea ice loss play a key role in driving the identified changes in temperature and precipitation extremes over southern continents. By comparing with global warming experiments, we find that Antarctic sea ice loss may affect temperature and precipitation extremes for some regions under greenhouse warming, especially Antarctica.
Abstract
The “Snowball Earth” hypothesis, proposed to explain the Neoproterozoic glacial episodes in the period 750–580 million years ago, suggested that the earth was globally covered by ice/snow during these events. This study addresses the problem of the forcings required for the earth to enter such a state of complete glaciation using the Community Climate System Model, version 3 (CCSM3). All of the simulations performed to address this issue employ the geography and topography of the present-day earth and are employed to explore the combination of factors consisting of total solar luminosity, CO2 concentration, and sea ice/snow albedo parameterization that would be required for such an event to occur. The analyses demonstrate that the critical conditions beyond which runaway ice–albedo feedback will lead to global freezing include 1) a 10%–10.5% reduction in solar radiation with preindustrial greenhouse gas concentrations; 2) a 6% reduction in solar radiation with 17.5 ppmv CO2; or 3) 6% less solar radiation and 286 ppmv CO2 if sea ice albedo is equal to or greater than 0.60 with a snow albedo of 0.78, or if sea ice albedo is 0.58 with a snow albedo equal to or greater than 0.80. These bifurcation points are very sensitive to the sea ice and snow albedo parameterizations. Moreover, “soft Snowball” solutions are found in which tropical open water oceans stably coexist with year-round snow-covered low-latitude continents, implying that tropical continental ice sheets would actually be present. The authors conclude that a “soft Snowball” is entirely plausible, in which the global sea ice fraction may reach as high as 76% and sea ice margins may extend to 10°S(N) latitudes.
Abstract
The “Snowball Earth” hypothesis, proposed to explain the Neoproterozoic glacial episodes in the period 750–580 million years ago, suggested that the earth was globally covered by ice/snow during these events. This study addresses the problem of the forcings required for the earth to enter such a state of complete glaciation using the Community Climate System Model, version 3 (CCSM3). All of the simulations performed to address this issue employ the geography and topography of the present-day earth and are employed to explore the combination of factors consisting of total solar luminosity, CO2 concentration, and sea ice/snow albedo parameterization that would be required for such an event to occur. The analyses demonstrate that the critical conditions beyond which runaway ice–albedo feedback will lead to global freezing include 1) a 10%–10.5% reduction in solar radiation with preindustrial greenhouse gas concentrations; 2) a 6% reduction in solar radiation with 17.5 ppmv CO2; or 3) 6% less solar radiation and 286 ppmv CO2 if sea ice albedo is equal to or greater than 0.60 with a snow albedo of 0.78, or if sea ice albedo is 0.58 with a snow albedo equal to or greater than 0.80. These bifurcation points are very sensitive to the sea ice and snow albedo parameterizations. Moreover, “soft Snowball” solutions are found in which tropical open water oceans stably coexist with year-round snow-covered low-latitude continents, implying that tropical continental ice sheets would actually be present. The authors conclude that a “soft Snowball” is entirely plausible, in which the global sea ice fraction may reach as high as 76% and sea ice margins may extend to 10°S(N) latitudes.
Abstract
This study investigates the climate dynamic feedbacks during a transition from the present climate to the extremely cold climate of a “Snowball Earth” using the Community Climate System Model, version 3 (CCSM3). With the land–sea distribution fixed to modern, it is found that by reducing solar luminosity and/or carbon dioxide concentration: 1) the amount of atmospheric water vapor and its attendant greenhouse effect decrease with the logarithm of sea ice cover, thereby promoting the expansion of sea ice; 2) over the sea ice, the cloud radiative feedback is positive, thus enhancing sea ice advance; over the ocean, the cloud radiative feedback is first negative and then becomes positive as sea ice enters the tropics; and 3) the strength of the atmospheric Hadley cell and the wind-driven ocean circulation increases significantly in the Southern Hemisphere, inhibiting the expansion of sea ice into the tropics. Meanwhile, the North Atlantic Deep Water cell disappears and the Antarctic Bottom Water cell strengthens and expands to occupy almost the entire Atlantic basin. In the experiment with 6% less solar radiation and 70 ppmv CO2 compared to the control experiment with 100% solar radiation and 355 ppmv CO2 near the ice edge (28°S latitude), the changes of solar radiation, CO2 forcing, water vapor greenhouse effect, longwave cloud forcing at the top of the model, and atmospheric and oceanic energy transport are −22.4, −6.2, −54.4, +6.2, and +16.3 W m−2, respectively. Therefore, the major controlling factors in producing global ice cover are ice albedo feedback (Yang et al., Part I) and water vapor feedback.
Abstract
This study investigates the climate dynamic feedbacks during a transition from the present climate to the extremely cold climate of a “Snowball Earth” using the Community Climate System Model, version 3 (CCSM3). With the land–sea distribution fixed to modern, it is found that by reducing solar luminosity and/or carbon dioxide concentration: 1) the amount of atmospheric water vapor and its attendant greenhouse effect decrease with the logarithm of sea ice cover, thereby promoting the expansion of sea ice; 2) over the sea ice, the cloud radiative feedback is positive, thus enhancing sea ice advance; over the ocean, the cloud radiative feedback is first negative and then becomes positive as sea ice enters the tropics; and 3) the strength of the atmospheric Hadley cell and the wind-driven ocean circulation increases significantly in the Southern Hemisphere, inhibiting the expansion of sea ice into the tropics. Meanwhile, the North Atlantic Deep Water cell disappears and the Antarctic Bottom Water cell strengthens and expands to occupy almost the entire Atlantic basin. In the experiment with 6% less solar radiation and 70 ppmv CO2 compared to the control experiment with 100% solar radiation and 355 ppmv CO2 near the ice edge (28°S latitude), the changes of solar radiation, CO2 forcing, water vapor greenhouse effect, longwave cloud forcing at the top of the model, and atmospheric and oceanic energy transport are −22.4, −6.2, −54.4, +6.2, and +16.3 W m−2, respectively. Therefore, the major controlling factors in producing global ice cover are ice albedo feedback (Yang et al., Part I) and water vapor feedback.
Abstract
Decadal trends are compared in various fields between Northern Hemisphere early winter, November–December (ND), and late-winter, February–March (FM), months using reanalysis data. It is found that in the extratropics and polar region the decadal trends display nearly opposite tendencies between ND and FM during the period from 1979 to 2003. Dynamical trends in late winter (FM) reveal that the polar vortex has become stronger and much colder and wave fluxes from the troposphere to the stratosphere are weaker, consistent with the positive trend of the Arctic Oscillation (AO) as found in earlier studies, while trends in ND appear to resemble a trend toward the low-index polarity of the AO. In the Tropics, the Hadley circulation shows significant intensification in both ND and FM, with stronger intensification in FM. Unlike the Hadley cell, the Ferrel cell shows opposite trends between ND and FM, with weakening in ND and strengthening in FM. Comparison of the observational results with general circulation model simulations is also discussed.
Abstract
Decadal trends are compared in various fields between Northern Hemisphere early winter, November–December (ND), and late-winter, February–March (FM), months using reanalysis data. It is found that in the extratropics and polar region the decadal trends display nearly opposite tendencies between ND and FM during the period from 1979 to 2003. Dynamical trends in late winter (FM) reveal that the polar vortex has become stronger and much colder and wave fluxes from the troposphere to the stratosphere are weaker, consistent with the positive trend of the Arctic Oscillation (AO) as found in earlier studies, while trends in ND appear to resemble a trend toward the low-index polarity of the AO. In the Tropics, the Hadley circulation shows significant intensification in both ND and FM, with stronger intensification in FM. Unlike the Hadley cell, the Ferrel cell shows opposite trends between ND and FM, with weakening in ND and strengthening in FM. Comparison of the observational results with general circulation model simulations is also discussed.
Abstract
Sea surface temperature (SST) forecast products from the NCEP Climate Forecast System (CFSv2) that are widely used in climate research and prediction have nonstationary bias. In this study, we develop single- (ANN1) and three-hidden-layer (ANN3) neural networks and examine their ability to correct the SST bias in the NCEP CFSv2 extended seasonal forecast starting from July in the extratropical Northern Hemisphere. Our results show that the ensemble-based ANN1 and ANN3 can reduce the uncertainty associated with parameters assigned initially and dependence on random sampling. Overall, ANN1 reduces RMSE of the CFSv2 forecast SST substantially by 0.35°C (0.34°C) for the testing (training) data and ANN3 further reduces RMSE relatively by 0.49°C (0.47°C). Both the ensemble-based ANN1 and ANN3 can significantly reduce the spatially and temporally varying bias of the CFSv2 forecast SST in the Pacific and Atlantic Oceans, and ANN3 shows better agreement with the observation than that of ANN1 in some subregions.
Significance Statement
Global coupled climate models are the primary tool for climate simulation and prediction and provide initial and boundary conditions to drive regional climate models. SST is an essential climate variable simulated and forecast by global climate models, which suffers substantial biases both spatially and temporally. We apply the ensemble averaging of both single- and three-hidden-layer neural networks on the NCEP CFSv2 SST forecast. They can correct the identified SST error, though ANN3 performs relatively better than that of ANN1. Thus, ensemble-based ANN3 is a useful SST bias correction approach.
Abstract
Sea surface temperature (SST) forecast products from the NCEP Climate Forecast System (CFSv2) that are widely used in climate research and prediction have nonstationary bias. In this study, we develop single- (ANN1) and three-hidden-layer (ANN3) neural networks and examine their ability to correct the SST bias in the NCEP CFSv2 extended seasonal forecast starting from July in the extratropical Northern Hemisphere. Our results show that the ensemble-based ANN1 and ANN3 can reduce the uncertainty associated with parameters assigned initially and dependence on random sampling. Overall, ANN1 reduces RMSE of the CFSv2 forecast SST substantially by 0.35°C (0.34°C) for the testing (training) data and ANN3 further reduces RMSE relatively by 0.49°C (0.47°C). Both the ensemble-based ANN1 and ANN3 can significantly reduce the spatially and temporally varying bias of the CFSv2 forecast SST in the Pacific and Atlantic Oceans, and ANN3 shows better agreement with the observation than that of ANN1 in some subregions.
Significance Statement
Global coupled climate models are the primary tool for climate simulation and prediction and provide initial and boundary conditions to drive regional climate models. SST is an essential climate variable simulated and forecast by global climate models, which suffers substantial biases both spatially and temporally. We apply the ensemble averaging of both single- and three-hidden-layer neural networks on the NCEP CFSv2 SST forecast. They can correct the identified SST error, though ANN3 performs relatively better than that of ANN1. Thus, ensemble-based ANN3 is a useful SST bias correction approach.