The progress and advances of the detection and attribution of changes in the hydrological cycle in the IPCC Assessment Reports of WGI and WGII from 1990 to 2007 are reviewed. Accomplishment and endorsed by the joint Expert Meeting on Detection and Attribution in 2009, the Good Practice Guidance Paper (GPGP) for IPCC Lead Authors with its main content and characteristics are briefly introduced. Based on the review and the purpose of the GPGP, some characteristics on the detection and attribution of global warming and of changes in the hydrological cycle are presented.
IPCC Assessment Report ; climate change ; hydrological cycle ; detection and attribution
Have the climatic variables and relevant natural and human systems been affected by external forcing? Have their effects already exceeded the influence of internal natural variability? These questions are asked by climate scientists and policy makers. In climate science, the comparison of simulated results of climate models with observed climate variables and the evaluation of consistency of observed changes with the expected response by external forcing will enhance the understanding of the interaction and mechanisms between natural climate variability and external forcing. It also helps to recognize the causes of climate change and to improve climate models and their capacity of climate prediction. In the research fields of hydrology and water resources, the use of physical and statistical models to study the changes of the hydrological cycles affected by anthropogenic forcing and multi-drivers, and their variations and trends in river basin can be revealed. Further, the prediction availability of flood, drought and water resource will be provided. Policy makers have to understand the current climate situation to design adequate climate policies for a successful adaptation and risk management. Especially they need to know what climate change really means, and when, where and by which mechanisms the hydrologic conditions and water resources are affected by climate change.
Therefore, the detection and attribution (D&A) of climate change are critical components of the IPCC Assessment Reports. Many advances and new findings with respect to D&A have been achieved since the IPCC First Assessment Report (FAR). Due to the complexity of temporal-spatial variations in atmospheric system, involving internal and external multidrivers and nonlinear interaction between natural and anthropogenic forcing, there are still great gaps in D&A studies. Especially the different temporal and spatial variations in climate and relevant systems, their causes and mechanisms, and their levels of confidence are of main interest. In order to promote and accelerate further studies, the IPCC Working Group I (WGI) and Working Group II (WGII) held a joint Expert Meeting on Detection and Attribution Related to Anthropogenic Climate Change in September 2009. The definitions and terminology, methods, data, and other requirements were discussed, based on which a stand-alone Good Practice Guidance Paper (GPGP) for IPCC Lead Authors was produced. The GPGP covered the full set of fundamental D&A issues. This will benefit the understanding of the scientific basis of climate change, the effects on physical and human systems and in enabling decision makers to manage climate-related risks.
This paper reviewed the processes and advances of D&A in four IPCC Assessment Reports from 1990 to 2007. It introduces the GPGP briefly. Based on which, views on D&A of observed changes in the hydrological cycles affected by climate change have been summarized. We hope that this paper can support further studies in this field.
The IPCC Assessment Reports present the scientific basis and predictions of climate change. They also assess the impacts of climate change on the environment and socio-economy systems. Since the IPCC FAR published in 1990, more knowledge and findings on D&A of anthropogenic climate change and its impacts have been achieved. This is due to the accumulation of past and recent observation data as well as the improvement of climate (and impact) models based on the progress in simulation and analysis techniques. Updated every five or six years, the progress in the D&A of climate change is presented in the reports on the basis of climate change (by WGI) and its impacts (by WGII). In order to outline this progress within every IPCC Assessment Report, the findings on global warming by WGI and on impacts on the hydrological cycle by WGII are listed in Table 1 ; Table 2 , respectively.
|IPCC Assessment Report||External forcing||Methodology||Observed global warming (over land) and its causes||Major progress and achievements|
|First Assessment Report (FAR)[ IPCC, 1990a ]||CO2 and CH4 increase||Atmosphere energy balance approach and GCMs for responses of CO2 forcing||Over a century, the average global land temperature rose by 0.3–0.6 °C,which is close to the predictions of climate models and corresponds to the natural variability||The temperature rise is mainly caused by natural variations, or such natural variations and other human factors have set off anthropogenic induced temperature rise|
|Second Assessment Report (SAR) [ IPCC, 1996a ]||Greenhouse gases (GHGs), sulphur, aerosols, and stratospheric ozone||Comparison of coupled atmosphere-ocean model simulations of temporal-spatial responses to anthropogenic forcing with the temporal-spatial distribution of observations||Global warming is unequal distributed, warmest at 40°–70°N, with a few areas of cooling in the North Atlantic Ocean and adjacent land during the last 100 years. The increase of observed global average temperature is unlikely to be caused by natural variability||Impacts due to anthropogenic forcing are more intense than natural variability. All the results point to human activities being the cause of global warming. It is very difficult to quantitatively estimate these causes and impacts|
|Third 9 Assessment Report (TAR) [ IPCC, 2001a ]||Multiple anthropogenic and natural forcing, e.g., halocarbons, CO2 , CH4 , N2 O, CFCs, O3 , aerosols||Multiple coupled atmosphere-ocean models are used to simulate spatial-temporal changes from different forcing. Internal natural climate variability is estimated from long controlled simulations for centuries. The optimal fingerprinting method is used to enhance the ratio of anthropogenic climate change signals to natural variation noises||The global warming by (0.6±0.2) °C in the recent 100 years is very unlikely to be caused by natural variability. The response to solar radiation and volcanic eruptions can not reflect the warming in the past 50 years; the global warming over the past 50 years is likely caused by an increasing concentration of GHGs||Definitions of D&A of climate change are presented for the first time. D&A studies on climate change due to anthropogenic forcing include statistical analysis and multiple verifications. The observed changes are unlikely caused by natural variability. Results are consistent with expected joint responses of human and natural forcing. It is not consistent with other interpretations in physics|
|Fourth Assessment Report (AR4) [ IPCC, 2007a ]||Multiple anthropogenic and natural forcing, e.g., GHGs (CO2 , CH4 , N2 O), O3 , halocarbons, volcanic aerosols and solar irradiance||Multiple coupled atmosphere-ocean models are used to simulate multiple forcing. Bayesian law and filter technology are introduced into the studies on optimal fingerprinting||The rise in average temperature over the six continents, except the Arctic, is very likely caused by an increase of observed anthropogenic GHG emissions. It is difficult to detect and attribute external forcing impacts on a spatial scale of higher resolution, and on a temporal scale of less than 50 years, due to various natural variables||Correct D&A is given by the use of objective statistical detection methods to estimate whether the natural variability is different within the climate system, and whether evidences of responses are found for external forcing. This detection method uses spatial and temporal distributions to identify responses to one or multiple external forcing|
|IPCC Assessment Report||External forcing, external driver||Methodology||Observed changes in the hydrological cycle and water resources||Major progress and achievements|
|FAR [ IPCC, 1990b ]||Rise in CO2 concentration||Changes in temperature and precipitation, statistic analysis of observed longterm streamflow||N/A||N/A|
|SAR [ IPCC , 1996b ]||GHGs emission||Comparison of observed longterm streamflow with temperature and precipitation time-series||Increasing trend in precipitation and streamflow in the Northern Hemisphere and high altitudes, attribution of anthropogenic forcing has not been carried out||Analysis of paleo-climate data, proxy data and instrumental observations on long-time hydrological series, and development of large-scale hydrological models|
|TAR [ IPCC , 2001b ]||GHGs, sulphur, aerosols, and natural forcing||Statistical detection methods are used to analyze the observed long-term series of streamflow, temperature, and precipitation||Over the last 20 years, increases in regional temperature had discernible impact on the retreat of glaciers, thawing of permafrost, late frost, and earlier break-up of ice on rivers and lakes (high confidence level of around 80%)||Studies on the impacts of natural climatic variability on streamflow are implemented; patterns of large-scale variables such as ENSO, Atlantic Oscillations, decadal oscillations of the Pacific, and the monsoon are analysed with focus on the hydrological variation to reveal their relation in inter-annual and inter-decadal variation|
|AR4 [ IPCC , 2007b ]||GHGs, sulphur, aerosols, and natural forcing||Multiple coupled atmosphere-ocean models are used separately to simulate temperature changes caused by natural and anthropogenic forcing. Using statistical pattern comparison methods to study the consistency of spatial pattern of changes in temperature to spatial changes in the components of the hydrological cycle and hydrological systems||Very high confidence levels (90%) based on 75 studies show that many physical and biological systems have been impacted by anthropogenic climate change, changes in the hydrological cycle, such as runoff increase and earlier spring peak discharge of many glacier and snow-fed rivers are already affected by recent climate change (particularly due to the regional temperature increase). Anthropogenic climate forcing has influenced the global runoff pattern with significant increase in the northern part of Eurasia and in the high latitudes and decrease in low latitude of the northwest part of North America [ Milly et al., 2005 ]||Definitions of detection and joint attribution in impact studies are presented for the first time. The joint attribution involves both attribution of observed changes to regional climate change and attribution of a measurable proportion of either regional climate change or the associated observed changes in the system to anthropogenic causes, beyond natural variability. The two steps considered together necessarily involve a greater amount of uncertainty than that associated with each step when considered in isolation. Confidence in joint attribution will generally be similar to or weaker than the weakest step [ Rosenzweig et al ., 2008 ; De’ath et al ., 2009 ]|
The progress in AR4 has been achieved by following studies on D&A of observed changes in the hydrological cycle at global and river basin scale.
(1) Santer et al.  used formal methods of D&A to simulate fingerprint patterns of changes in tropospheric moisture content caused by anthropogenic forcing in the 20th century by means of 22 climate models. With high confidence, they prove that the increase in moisture content is mainly caused by anthropogenic GHG emissions.
(2) Willett et al.  used coupled climate models and global land surface gridded observational data sets of specific humidity to reveal that a significant increase of land surface specific humidity in last century is mainly caused by anthropogenic climate forcing.
(3) Zhang et al.  found detectable impacts by anthropogenic forcing using optimal fingerprinting method on global latitudinal band average precipitation in the 20th century.
(4) Barnett et al.  studied hydrological changes of three river basins in the western United States by means of two climate models under multiple anthropogenic drivers, high-resolution hydrologic model, and a single-variable method of D&A. As a result, he found that over 60% of the changes in stream-flow, winter temperature, and snow pack in 1950–1999 can be explained by anthropogenic forcing. They provided a method for the D&A of multivariable anthropogenic climate changes at a regional scale.
The GPGP [ IPCC, 2010 ] covers the definition of D&A of climate change, methods, and information requirements, etc.. The consensus is obtained through communication and discussions of experts from the IPCC WGI and WGII to standardize their studies on D&A and provide references for the IPCC Fifth Assessment Report (AR5). The GPGP features the following:
(1) A unified definition of D&A given by WGI and WGII: “Detection of change is defined as the process of demonstrating that climate or a system affected by climate has changed in some defined statistical sense, without providing a reason for that change. Attribution is defined as the process of evaluating the relative contribution of multiple causal factors to a change or event with an assignment of statistical confidence. The observed changes must be able to be detected”. Compared with previous definitions, this definition highlights multiple causal factors which are not only contributed to anthropogenic climate change.
(2) Multiple factors that cause changes in the climate system and related systems are clarified and defined. 1) External forcing refers to the forcing causing climate change outside the climate system, which may include solar variation, volcanic eruption, and anthropogenic changes in atmospheric composition and land use. Such forcing may impact upon the climate system and non-climate system directly by no means of climate adjustment. 2) External driver means any external forcing factor outside the system of interest that causes a change in the system. As for WGII, external driver is an extensive group of external impacts, which may include or exclude the climate. Nonclimate drivers may exert an impact upon the natural and human systems. If the response to GHG forcing can be segregated from other external forcing and driver, the contribution of greenhouse forcing to system changes can be estimated. 3) Confounding factor is a forcing that affects the variable or system of interest but is not explicitly accounted for in the design of study. A confounding factor could lead to erroneous conclusions within attribution studies if not properly considered or controlled for. Examples of possible confounding factors for attribution studies include pervasive biases and errors in instrumental records, model errors and uncertainties, improper or missing representation of factors in climate and impact models, uncertain or unaccounted internal variability, and nonlinear interactions between forcing and responses [ IPCC, 2010 ].
(3) Four methods of D&A of climate change are summarized. 1) Single-step attribution of external forcing is based on explicit modeling of the response of the variable to external forcing and drivers. Modeling can involve a single comprehensive model or a sequence of models. The attribution step involves detection of a significant change in the variable of interest and comparison of observed changes in the variable of interest with expected changes due to external forcing and drivers [ Gillet et al., 2004 ]. 2) Multi-step attribution to external forcing: The changes of observed variables are attributed to climatic or environmental changes, which are attributed to an estimation of external forcing and external drivers respectively. This method may include a series of comprehensive applications of observed data and models. The last step may use process simulation or related statistics models, each step can demonstrate its own level of confidence. However the reliability of the total estimation will be similar to or less than that of the weakest step [Ee’ath et al., 2009]. 3) Associative pattern attribution to external forcing comprises a synthesis of large numbers of results (possibly across multiple systems) demonstrating the sensitivity of impacts to a change in climate conditions and other external drivers. The link between externally forced climate change and its ensemble of results is made using spatial and temporal measures of association [ Rosenzweig et al., 2008 ]. 4) Attribution to a change in climatic conditions (climate change) comprises assessments that attribute an observed change in a variable of interest to an observed change in climate conditions based on process knowledge and relative importance of a change in climate condition in determining the observed effects [ Hao et al., 2008 ]. The first three methods attribute observed impacts or climate change to external forcing, such as the increase of GHG concentration, while the fourth method takes climate as the driver to study the link between impacts and climate change. No matter which method is adopted, the authors should specifically state the causal factor to which a particular change is being attributed and should identify whether the studied issues are concerned with a response to a change in climate and/or environmental conditions and/or other external drivers and forcing. The confidence levels should be evaluated for the data, model, methods, and confounding factors used in the study [ IPCC , 2010 ].
(1) The D&A is to ascertain whether the observed changes in the climate system are significant, and whether these changes are caused by external forcing. This is done by an interactive cause-effect chain from external forcing to the climate system, physical and biological system, and socio-economic systems, and their interaction. The GPGP synthesized four different methods of D&A. The first three methods are applicable to the attribution of impacts of anthropogenic forcing upon climate change which can be separated from other external forcing or drivers. The fourth method is applicable to attribute the observed changes in the hydrological cycle to climatic changes but not to identify whether such changes are caused by natural or anthropogenic forcing.
(2) With the accumulation of observed data, and improvements in climate models and cognition, the resolution of temporal and spatial scales used for D&A is being refined. This is reflected by the shift of predictions of climate warming at global scale in 1990, hemispheric, marine and terrestrial scales from 1995 to 2001, and in six continental scales in 2007 (currently expanded to the Antarctic) [ Gillett et al., 2008 ]. The optimal fingerprint and formal D&A methods of climate change have been widely applied in the studies on global hydrological elements, such as specific humidity, moisture content, precipitation, and streamflow [ Stott et al., 2010 ].
(3) Since 2007, many studies on D&A focus on the regional scale. This is reflected by the improvement in simulation of natural variability in climate models, the description of different external forcing, as well as the enhancement in modelled separation of impacts caused by different external forcing or drivers. For instance, the study on D&A of anthropogenic influences on the hydrological cycle in the western United States used the fingerprinting method. However, it is still rather difficult to carry out D&A of climatic impacts upon the hydrological cycle with a spatial scale lower than continental scale and a time scale of less than 50 years. This is mainly because the signal of anthropogenic forcing at a small scale is generally covered by confounding factors and large natural variables. Hence, the impacts from external forcing and drivers are difficult for climate models to separate.
(4) At present, we are still unable to detect anthropogenic influences on observed changes in all climatic variables. This is due to some of the variables which are not sensitive to the increase in GHG concentration and the reliability of climate models. For example, our ability to detect the impact of anthropogenic forcing on precipitation and surface pressure are far weaker than on temperature.
(5) In river basins recharged by precipitation and affected by human activities, the studies on D&A of anthropogenic forcing are difficult due to the impacts of multiple drivers such as land use change, urbanization, reservoir control, water consumption and the significant natural variability of hydrological variables. Currently the D&A of elements of the hydrological cycle focus mainly on analytical approaches which link observed changes to changes in temperature, precipitation and non-climatic factors without the emphasis on causation by natural or anthropogenic components. For this kind of approach, long-term climatic and non-climatic observation data, advanced statistical techniques and hydrological models on a profound physical basis are vital for the correct causation process.
(6) The optimal fingerprinting method based on climate models will achieve great progress in terms of more observed data, the improvement of regional climate models, the interactions between recognised anthropogenic and natural forcing, and increasing information of decision maker’s towards natural, anthropogenic and environmental stress. In river basin with long-term climatic and non-climatic observed data where hydrological variables are sensitive to climate change, the optimal fingerprinting method will be valuable in D&A studies.
The D&A of climate change impacts on the observed climate and elements of the hydrological cycle have made great progress since the IPCC FAR in 1990. Based on climate model simulation, the optimal fingerprinting methods have been used to detect and attribute the responses of observed change to GHG emissions from other external forcing at large spatial scales (from global, hemispheric, marine to continental, and 7 sub-continents). At present, the detection of anthropogenic influence is not yet possible for all climate variables. It is still difficult to attribute observed changes in climate or variables of interest on a spatial scale less than 5,000 km and temporal scales of less than 50 years. For river basins recharged by precipitation and influenced by strong human activities, D&A studies mainly focused on analytical approaches to link physical impacts to changes in temperature or precipitation as driver (natural or anthropogenic). For river basins with better observational data and more sensitivity towards climate change, the use of formal D&A methods to identify the pattern responses of the hydrological cycle to external forcing and drivers is a valuable and promising area of research.
This study was supported by the National Basic Research Program of China (2010CB428406) and the Ministry of Water Resources Commonwealth Project (200801001).