This chapter assesses the scientific literature on atmospheric and surface observations since AR4 (IPCC, 2007). The most likely changes in physical climate variables or climate forcing agents are identified based on current knowledge, following the IPCC AR5 uncertainty guidance (Mastrandrea et al., 2011).

As described in AR4 (Trenberth et al., 2007), the climate comprises a variety of space- and timescales: from the diurnal cycle, to interannual variability such as the El Niño-Southern Oscillation (ENSO), to multi- decadal variations. ‘Climate change’ refers to a change in the state of the climate that can be identified by changes in the mean and/or the variability of its properties and that persists for an extended period of time (Annex III: Glossary). In this chapter, climate change is examined for the period with instrumental observations, since about 1850. Change prior to this date is assessed in Chapter 5. The word ‘trend’ is used to designate a long-term movement in a time series that may be regarded, together with the oscillation and random component, as composing the observed values (Annex III: Glossary). Where numerical values are given, they are equivalent linear changes (Box 2.2), though more complex nonlinear changes in the variable will often be clear from the description and plots of the time series.

In recent decades, advances in the global climate observing system have contributed to improved monitoring capabilities. In particular, satellites provide additional observations of climate change, which have been assessed in this and subsequent chapters together with more traditional ground-based and radiosonde observations. Since AR4, substantial developments have occurred including the production of revised data sets, more digital data records, and new data set efforts. New dynamical reanalysis data sets of the global atmosphere have been published (Box 2.3). These various innovations have improved understanding of data issues and uncertainties (Box 2.1).

Developing homogeneous long-term records from these different sources remains a challenge. The longest observational series are land surface air temperatures (LSATs) and sea surface temperatures (SSTs). Like all physical climate system measurements, they suffer from non- climatic artefacts that must be taken into account (Box 2.1). The global combined LSAT and SST remains an important climate change measure for several reasons. Climate sensitivity is typically assessed in the context of global mean surface temperature (GMST) responses to a doubling of CO2 (Chapter 8) and GMST is thus a key metric in the climate change policy framework. Also, because it extends back in time farther than any other global instrumental series, GMST is key to understanding both the causes of change and the patterns, role and magnitude of natural variability (Chapter 10). Starting at various points in the 20th century, additional observations, including balloon-borne measurements and satellite measurements, and reanalysis products allow analyses of indicators such as atmospheric composition, radiation budgets, hydrological cycle changes, extreme event characterizations and circulation indices. A full understanding of the climate system characteristics and changes requires analyses of all such variables as well as ocean (Chapter 3) and cryosphere (Chapter 4) indicators. Through such a holistic analysis, a clearer and more robust assessment of the changing climate system emerges (FAQ 2.1).

This chapter starts with an assessment of the observations of the abundances of greenhouse gases (GHGs) and of aerosols, the main drivers of climate change (Section 2.2). Global trends in GHGs are indicative of the imbalance between sources and sinks in GHG budgets, and play an important role in emissions verification on a global scale. The radiative forcing (RF) effects of GHGs and aerosols are assessed in Chapter 8. The observed changes in radiation budgets are discussed in Section 2.3. Aerosol–cloud interactions are assessed in Chapter 7. Section 2.4 provides an assessment of observed changes in surface and atmospheric temperature. Observed change in the hydrological cycle, including precipitation and clouds, is assessed in Section 2.5. Changes in variability and extremes (such as cold spells, heat waves, droughts and tropical cyclones) are assessed in Section 2.6. Section 2.7 assesses observed changes in the circulation of the atmosphere and its modes of variability, which help determine seasonal and longer-term anomalies at regional scales (Chapter 14).

Trends have been assessed where possible for multi-decadal periods starting in 1880, 1901 (referred to as long-term trends) and in 1951, 1979 (referred to as short-term trends). The time elapsed since AR4 extends the period for trend calculation from 2005 to 2012 for many variables. The GMST trend since 1998 has also been considered (see also Box 9.2) as well as the trends for sequential 30-year segments of the time series. For many variables derived from satellite data, information is available for 1979–2012 only. In general, trend estimates are more reliable for longer time intervals, and trends computed on short intervals have a large uncertainty. Trends for short intervals are very sensitive to the start and end years. An exception to this is trends in GHGs, whose accurate measurement and long lifetimes make them well-mixed and less susceptible to year-to-year variability, so that trends computed on relatively short intervals are very meaningful for these variables. Where possible, the time interval 1961–1990 has been chosen as the climatological reference period (or normal period) for averaging. This choice enables direct comparisons with AR4, but is different from the present-day climate period (1986–2005) used as a reference in the modelling chapters of AR5 and Annex I: Atlas of Global and Regional Climate Projections.

It is important to note that the question of whether the observed changes are outside the possible range of natural internal climate variability and consistent with the climate effects from changes in atmospheric composition is not addressed in this chapter, but rather in Chapter 10. No attempt has been undertaken to further describe and interpret the observed changes in terms of multi-decadal oscillatory (or low-frequency) variations, (long-term) persistence and/or secular trends (e.g., as in Cohn and Lins, 2005; Koutsoyiannis and Montanari, 2007; Zorita et al., 2008; Lennartz and Bunde, 2009; Mills, 2010; Mann, 2011; Wu et al., 2011; Zhou and Tung, 2012; Tung and Zhou, 2013). In this chapter, the robustness of the observed changes is assessed in relation to various sources of observational uncertainty (Box 2.1). In addition, the reported trend significance and statistical confidence intervals provide an indication of how large the observed trend is compared to the range of observed variability in a given aspect of the climate system (see Box 2.2 for a description of the statistical trend model applied). Unless otherwise stated, 90% confidence intervals are given. The chapter also examines the physical consistency across different observations, which helps to provide additional confidence in the reported changes. Additional information about data sources and methods is described in the Supplementary Material to Chapter 2.