Page 77 - 全球气候变化及其影响Global Climate Change and Its Impacts-185×260

Chapter II Evidence for Global Climate Change

Confidence interval is also a crucial means of expressing data uncertainty. Based on
rigorous statistical principles, it provides the potential range of true values at a specific con-
fidence level. In climate change research, 95% or 99% confidence levels are commonly ad-
opted standards that have been widely recognized through long-term practice and extensive
research validation. Taking statistical analysis of long-term precipitation data from meteo-
rological stations as an example, by constructing appropriate statistical models such as time
series analysis models and regression analysis models, while comprehensively considering
interannual variations in precipitation, seasonal fluctuations, and correlations with other me-
teorological factors, the calculated 95% confidence interval for annual precipitation in a re-
gion might be [800 - 1200] mm. This indicates a 95% probability that the true annual precip-
itation for this region falls within this range. Compared to error ranges, confidence intervals
not only account for data dispersion but also incorporate factors such as sample size and sta-
tistical model reliability, enabling more comprehensive reflection of data uncertainty. When
analyzing atmospheric aerosol concentration data retrieved from satellite remote sensing,
using advanced statistical methods such as Bayesian statistics and Monte Carlo simulations
to calculate confidence intervals at 99% confidence levels can provide researchers with more
accurate information about the uncertainty of aerosol concentration estimates. This facilitates
reasonable assessment of how data uncertainty affects research outcomes when analyzing the
relationship between atmospheric radiation balance and climate change.
In terms of model simulation, climate change models also exhibit significant uncertain-
ties in their simulation results due to simplifications and approximations of physical process-
es in the complex climate system, as well as uncertainties in model parameters. The com-
plex climate system encompasses atmosphere, oceans, and landand other subsystems, with
complex nonlinear interactions existing between each subsystem. To enable simulations on
computers, models must simplify these processes. For example, when simulating atmospher-
ic circulation, complex atmospheric motions are simplified into a set of equations, which
inevitably introduces certain errors. The uncertainty in model parameters stems from the lack
of precise measurements or understanding of parameters for many physical processes, such
as radiative properties of cloud layers and ocean mixing parameters. To quantify this uncer-
tainty, ensemble simulation methods can be employed. By running multiple model simula-
tions with different initial conditions or parameter settings, a series of simulation results are
obtained. Statistical analysis is then performed on these simulation results to calculate their
range and probability distribution. For instance, when predicting future global average tem-
perature changes, 100 climate model simulations with different initial conditions produced
a result range of 1.5 - 4.5°C warming. Further calculations can determine the probability of
simulation results within different warming ranges, such as 40% of simulations showing 2.0
- 3.0°C warming. These result ranges and probability distributions obtained through ensem-
ble simulations effectively quantify model simulation uncertainties. When expressing model
simulation uncertainties, besides providing result ranges and probability distributions, the

• 69 •