Invited review: Intergovernmental Panel on Climate Change, agriculture, and food—A case of shifting cultivation and history

Since 1990, the Intergovernmental Panel on Climate Change (IPCC) has produced five Assessment Reports (ARs), in which agriculture as the production of food for humans via crops and livestock have featured in one form or another. A constructed database of the ca. 2,100 cited experiments and simulations in the five ARs was analyzed with respect to impacts on yields via crop type, region, and whether adaptation was included. Quantitative data on impacts and adaptation in livestock farming have been extremely scarce in the ARs. The main conclusions from impact and adaptation are that crop yields will decline, but that responses have large statistical variation. Mitigation assessments in the ARs have used both bottom‐up and top‐down methods but need better to link emissions and their mitigation with food production and security. Relevant policy options have become broader in later ARs and included more of the social and nonproduction aspects of food security. Our overall conclusion is that agriculture and food security, which are two of the most central, critical, and imminent issues in climate change, have been dealt with an unfocussed and inconsistent manner between the IPCC five ARs. This is partly a result of not only agriculture spanning two IPCC working groups but also the very strong focus on projections from computer crop simulation modeling. For the future, we suggest a need to examine interactions between themes such as crop resource use efficiencies and to include all production and nonproduction aspects of food security in future roles for integrated assessment models.

reads "Recognizing the fundamental priority of safeguarding food security… and the vulnerabilities of food production systems to the adverse impacts of climate change" acknowledging the central role of food security regionally and globally. Important interdisciplinary departures in the food security chapter of the IPCC (Porter et al., 2014) were recognition of factors other than food production in food security: such factors include food distribution and social and economic access to food, which all stand to be affected by climate change and which have possibilities for adaptation. Food security and agriculture have not always had such a clear or prominent position in IPCC ARs-with food security only specified in AR5 and with agriculture often rolled in with forestry and forest products (AR1 and AR4) or general ecosystem services (AR3). AR2 did examine impacts and adaptation of agriculture.
We regard the evolution of a food system perspective in IPCC AR5 as a very positive development that we hope will be amplified in AR6 and future IPCC Special Reports.
This review aims to develop further, and in more detail, the recent paper by Porter, Howden, and Smith (2017) on the link between the five IPCC ARs (AR1-AR5) and agriculture. Space constraints in that article prevented the presentation of topics such as regional differences in assessments of impacts, adaptation, and mitigation linked to agriculture; the balance between assessment of climate change and crops versus livestock; the methods used and how and why assessments might develop in the future. Post AR5 (Porter et al., 2014), the Royal Society of London (Royal Society, 2017) published an update on climate change effects on food production. Their conclusion was that post-AR5 studies have confirmed conclusions in AR5, but new studies "point strongly to the importance of accounting for how land use and cropping intensity might change." Our review addresses the above gaps and addresses potentially policy-relevant information that has become available since the AR5.
In addition, as this is an invited review, we have allowed ourselves the licence to include two issues which we think are of importance for future ARs dealing with food, agriculture, and climate change. Section 6.16.1 presents ideas to improve the robustness of crop models that have been the "work horses" of many, perhaps too many, climate change assessments. Models should check that they simulate accurately all crop and soil responses underlying responses to climate and we suggest a way for doing this. Secondly, models should be able to simulate accurately the interactions between resources such as radiation, water, and nutrients with and without changes in CO 2 level, an issue that rarely has been investigated (cf. Teixera et al., 2014). Simulating such interactions correctly is particularly important when it comes to examining adaptation options for crops. We show an example where this examination has been done for a well-known and well-used dataset. The second issue we raise is the integrated assessment of adaptation and mitigation. While crop models are generally responsive to climate, the range of crops that can be simulated is not sufficiently broad for a full assessment of food security-this is clear from the data presented in this paper.
Furthermore, disparities between integrated assessment models (IAMs) and crop models in spatial scale, treatment of uncertainty, data demand, and representation of agricultural management all limit the extent of crop model integration into IAMs that is currently possible. We think that more than one approach is needed if we are to capture the range of trade-offs and synergies that are important to food systems and relevant to policy design and development. We need to recognize that emissions occur across the full range of activities that deliver food security, not only agricultural production but also with a focus on climate-smart food systems.

| PROJEC TED IMPAC TS
To get an overview of the assessment of the projected impacts of climate change on crop yield across the five IPCC ARs, across the different global regions, and for the major global crops, we complied all data (2,116 entries) on projected crop yield with and without adaptation from AR1 to AR5. We constructed a database with information about the AR volume, crop type, global region, and projected mean change and variation in yields with and without adaptation. In this context, adaptation refers to all adaptation measures investigated in the scenarios throughout AR1-AR5, including but not limited to altering sowing times, crop cultivars, and species, adjusting irrigation and fertilizer application, reducing tillage, and implementing technical measures to more effectively capture rainwater and reduce soil erosion. Subsequently, the average mean change in yield with and without adaptation was calculated for each IPCC AR, each global region, and each major global crop (Tables 1-3). By reviewing the constructed dataset, it quickly becomes evident that the number of cases increases almost exponentially (except from AR2 to AR3), thereby increasing the confidence level of the results of each subsequent report. A striking omission across the five ARs is the almost complete lack of quantitative data of the effects of climate change on livestock; no quantitative data were presented from AR1 to AR3 and only 18 cases were reported in AR4 and AR5 combined (Rivera-Ferre et al., 2016).
All IPCC ARs, except AR1, have projected a crop yield reduction without adaptation (Table 1). The largest projected yield reduction was in AR2 with −13.8% followed by −9.9% in AR5. When climate adaptations were included in the analysis, most ARs also projected a yield reduction except for AR1 with a 9% yield increase and AR4 with a 3.6% yield increase. However, the standard deviations in the projections are large, ranging from 11.5% to 33%.
The standard deviation is also large for the mean change in yield for different global regions (Table 2). Without adaptation, Central Asia had yield change of −19.2%, followed by North Asia with −14.0%, Central and South America with −12.1%, and South Asia with −11.7% are the regions with the largest projected yield decreases. With adaptation, Southeast Asia, North Asia, and Australasia have the largest yield increase with +10.4%, +8.9%, and +6.9%, respectively.
For major global crops (Table 3), it is evident that the crops most severely affected by climate change without adaptation are soybean and maize with yield reductions of −16.7% and −10.8%, respectively.
The yield reduction for beans is even larger but only based on a single observation. For protein crops, this yield reduction is particularly alarming given their potential to replace meat-based protein with both health and greenhouse gas emissions benefits (Tilman & Clark, 2014). Also, besides maize, some of the other major staple crops for the southern hemisphere are projected to have significant yield reductions without adaptation, for example, −10.8% for maize, −9.3% for millet, −9.1% for sorghum, and −16.9% for soybean. Even with adaptation, large yield reductions are projected for maize (−5.6%), millet (−27.0%), sorghum (−23.5%), and soybean (12.8%). Considering that these three crops cover 60% of the area cultivated with cereal crops in Africa and provide 67% of the cereal yield on the continent (Macauley, 2015), a yield reduction of this magnitude would have severe consequences. Overall, adaptation is not projected to have a very large effect on reducing or even reversing yield reductions for the major global crops. Large yield increases can be seen for beans, groundnut, and grass, but these results are only based on few observations. Based on this analysis, rice and wheat, with yield increases of +3.4% and +1.9%, seem to be the only major global crops to benefit from adaptation efforts. It is worth noting that only 15 crops are included in the IPCC ARs. It is evident that a more accurate assessment of food security would require that a much larger number of crops are investigated.

| ADAP TATION
From the first IPCC Assessment onward, a systems approach has been applied to the analysis of climate impacts and adaptation relating to agriculture, food production, and, more recently, food systems. However, rainfall) toward integration of downscaled general circulation models data in grid-based or multisite, regional assessments. Nevertheless, the focus of the IPCC remained on mean yield change and it was only in AR3 was there inclusion of a focus on changes in yield variability and, in AR5, the nutritional quality of crops. While there are regional and global crop production studies, there have been few impact studies which have used a value chain or a food system's perspective.
Developing country studies remain relatively underrepresented in terms of population (Table 2), even though developing countries were identified as early as AR2 that they were likely to be the most negatively affected. Similarly, even though AR2 concluded that elevated atmospheric CO 2 concentrations would have beneficial impacts on crop production, there remained active debate in AR5 about the degree to which this may affect crop yields and quality.
As noted above, there has been relatively little quantitative treatment of livestock (Rivera-Ferre et al., 2016), other field crops, horticulture, and viticulture across IPCC reports with coverage being largely restricted to either generic, system-level responses, or sitespecific cases, largely because of the relative lack of studies using somewhat comparable modeling or other analysis methods in contrast to the mechanistic and other crop models, which have enabled meta-analysis, cross-model comparison, and assessment of uncertainties (Rosenzweig et al., 2014). The treatment of weeds, pests, and disease impacts is also inconsistently dealt with across the reports for the same reasons.
The aggregation of climate change impacts on food production systems to broadscale economic and food price impact was also initiated in the AR2 with results reported from two economic models (Reilly, Hohmann, & Kane, 1994;Rosenzweig & Parry, 1994 address global and regional impacts of climate trade on prices, production, and trade but also the uncertainties in model results and the reasons behind these (e.g., Nelson, Valin, & Sands, 2014).
Adaptation to the sorts of climate change impacts noted above is a fundamental part of risk management. Agriculture and food producers as well as value chain managers, consumers, and policy makers have shown considerable ability to adapt to climate changes both currently and going back into history; for instance, the establishment of grapevines in England in Roman times or the settlement of Greenland in medieval times. The expectation that adaptation of food systems is likely to be both feasible and attractive has resulted in coverage from AR1 onward. However, the framing, scope, likely effectiveness, and analytical methods used in IPCC reports have changed significantly since then (Table 4). There remain many gaps TA B L E 4 The framing, scope, and analysis methods used to address climate adaptation in agricultural and food systems in successive IPCC ARs IPCC assessment Framing, scope, and analysis methods used

AR1 1990
Framing: Three adaptation domains-physiological adaptation, farm level management "adjustment," and responses arising from policy at regional, national, and international levels. These were expressed in terms of enabling farming systems to reach a new equilibrium in response to altered climates. Scope: Farm level production focus not food systems. Analysis: Generally, adaptations were described qualitatively using historical analogs or first principle approaches rather than quantified responses.

AR2 1995
Framing: Spontaneous or planned adaptation, in response to or anticipation of climate change.
Scoping: Farm-level production system focus not food systems with brief reference to global economic analyses of producer surplus, which included with and without adaptation. Analysis: Few quantitative adaptation studies although most adaptation options were raised based on a systems view. However, these were mostly incremental such as agronomic adjustments although there were some systemic adaptations (sensu Rickards & Howden, 2012) such as the introduction of new species. There was a recognition that successful adaptation depends on technological advances, institutional arrangements, availability of financing and information exchange as well as adaptive capacity and alignment of the options with farmer needs so as to enhance adoption paths. Additionally, there was recognition of the possibility of policy maladaptation.

AR3 2001
Framing: No specific framing focused on farm-level agronomic changes. Scope: Farm-level production focus not food systems, with examples of integrated regional economic analyses of impacts and adaptation. Analysis: As well as qualitative discussion of options such as crop breeding to adjust to elevated CO 2 and temperatures, there were more quantitative analyses of cropping system adaptations allowing both tabular and figure summaries of the modeled effectiveness of adaptation. However, there was a critique that methodologically, there had been little progress since the previous IPCC assessment with the adaptation strategies being modeled limited to a small subset of the possible options and unrealistic assumptions regarding the degree and effectiveness of farmer adoption. There was recognition of adaptation costs including transition costs, dislocation costs, and capital and operational costs. There was, however, limited coverage of livestock adaptation with discussion of a range of management adaptations to reduce the effects of heat waves but few quantified or modeled analyses to draw from.

AR4 2007
Framing: Autonomous and planned adaptation modes. Scope: Recognition of the importance of a food system's approach but the focus remained on agricultural production. Analysis: Discussion of a broader range of possible adaptation options for both cropping and livestock using a more structured approach, particularly drawing off the burgeoning literature on cropping system impacts and adaptations. This allowed more geographically explicit analyses as well as a meta-analysis of impacts and adaptation as a function of temperature increase. However, most adaptation options addressed were still incremental in nature, reflecting in part limitations of the modeling approaches being used. There was a critique of the failure to provide generalized knowledge of adaptive capacity, of adoption pathways and barriers to these, and of a more comprehensive range of adaptation strategies, especially beyond simple, single agronomic changes. There was still limited evaluation of the costs of adaptation or of consequences of adaptation in relation to the environment and the natural resource base.

AR5 2014
Framing: incremental to transformational adaptation. Scope: Food system's approach although much of the literature able to be synthesized was on food production only. Analysis: Discussion of a broad range of possible adaptation options and their adoption paths for both cropping and livestock using a consistent framing. The further increase in the literature on cropping system impacts and adaptations allowed (a) an improved meta-analyses of impacts and adaptation as a function of temperature providing finer grained information across the major crops, by broad region and disaggregating results to allow assessment of the effectiveness of different agronomic adaptation options; and (b) a meta-analysis of the possible increase in crop yield variability over time. Livestock adaptations were not able to be dealt with as comprehensively as cropping systems due to limitations in the literature. There was increased recognition of the importance of institutional limits and adoption barriers but some other issues identified as shortcomings in prior IPCC Assessments remain largely unaddressed (e.g., adaptation costs, lack of methodological innovation, and diversity in adaptation analysis).
Abbreviations: AR, Assessment Report; IPCC, Intergovernmental Panel on Climate Change.
in terms of adaptation of food systems including, but not limited to, the need to include assessment of more transformative adaptations, adaptation of value chains and of regional food systems. Other important issues for the future include how to address the multitudinous barriers to adaptation, developing the pathways to not only build adaptive capacity but to also move this into adaptation actions, developing policies and programs to establish effective monitoring, evaluation, and attribution of adaptation and assessments, and to more effectively address net greenhouse gas emission reduction within adaptation strategies. This latter point is starting to be addressed in the IPCC AR6 cycle, being covered by two Special Reports (www.ipcc.ch/reports) as well as within the main ARs. Taking an approach based on the sectors from which emissions are reported is logical, but does not encourage food system's approaches to addressing emission reduction goals. Future assessments will need to take a more holistic view of the food system, and go beyond the accounting/reporting sectors considered to date.

| MITI G ATI ON
Another persistent issue across IPCC ARs arises from the structure in which assessments are conducted, with Working Group 1 focussing on the physical science basis of climate change, Working Group 2 focussing on impacts of climate change and adaptation, and Working Group 3 focussing on mitigation. The chapters dealing with agriculture and land in each AR are written by different authors and appear in different volumes, corresponding to each Working Group.
While efforts are made to encourage cross-working group/crossvolume collaboration and consistency, results have been uneven, with a number of disconnects in emphasis across the volumes.
The IPCC Special Report on Climate Change and Land, under production as part of the AR6 cycle and due in 2019, offers an opportunity to address some of the issues raised above. Firstly, it is a joint action across the three Working Groups, thereby including experts from more disciplines than usually found within Working Groups.
Secondly, it considers a wide range of land and climate change-related issues, including mitigation, adaptation, desertification, land degradation, sustainable land management, and food security. With an emphasis on integrated response options to address all of these challenges, considering synergies and trade-offs, it necessarily takes a broader view of land, agriculture, food systems, and the interventions available to address the considerable challenges facing humanity now and in the future. While examining all of these factors together is extremely challenging, due to the complexity of the sectors involved, the importance of food and agriculture and climate change for the future of humanity means it is a challenge that must be met. Future IPCC ARs could learn from the experience of producing this Special Report-to take a broader view of the issues facing land and agriculture and to facilitate cross Working Group integration.

| P OLI C Y
The policy elements of climate adaptation and mitigation in relation to agriculture and food systems have been addressed unevenly and incompletely over the various ARs. AR1 acknowledged the importance of a range of policies (listing food price, land use, forest resources, extension, and water transfers) but required more information in relation to potential responses. AR2 expanded the list to include research, land-use planning, water pricing and allocation, disaster vulnerability assessment, transport and trade policy and policies countries use to encourage or control production, limit food prices, and manage resource inputs to agriculture. There was a brief critical analysis of how policies may discourage adaptation strategies and acknowledgement of the political, economic, and cultural factors at play but overall very little concrete guidance in relation to policy design and development. In contrast, the AR3 and AR4 provided few linkages to policy and it was not until AR5 that more policy-relevant suggestions were developed. These included inter alia capacity building across the food system via support of monitoring and communication, systems analysis, extension capacity, and industry and regional networks that develop social capital and share information, supporting community partnerships in developing food and forage banks, enhancing investment in irrigation infrastructure and efficient water use technologies, revising land tenure arrangements (including attention to well-defined property rights), establishment of accessible, efficiently functioning markets for inputs and outputs (seed, fertilizer, labor, water, products, greenhouse gases emissions, etc.) and for financial services, including insurance. There was also introduction of ideas relating to modes of operation such as policy "mainstreaming" and policy analysis methodologies such as the need for multilevel assessment. Importantly, these policy inclusions in AR5 were consistent with moving away from the previous "agricultural production" focus to a more "food systems" focus but nevertheless did not substantially progress the integrated treatment of climate adaptation and mitigation.

| Assessing crop growth models to predict interactions between resource use efficiencies
The main types of models used in IPCC impact assessments on crop production fall into the category of crop simulation models that attempt to predict yields based on bioclimatic inputs and are mostly site based; statistical relationships have also been used (Porter et al., 2014). Such models are only just being used to examine CO 2 and other effects on yield and its protein concentration (Asseng et al., 2019) even though this topic has been a persistent theme in the ARs.
Thus, as suggestions, we wish to highlight the need to analyze the interactions between resource use efficiencies to change the consistency of crop models and better understand cropping systems response to climate change as a topic, focused on modeling. We think that this is an important topic for future assessment of climate impacts, adaptation, and mitigation within the land sector and agriculture and their position and role in climate change.
Bennetzen et al. (2016) showed via a historical deconstruction analysis, using a modified Kaya identity analysis (Kaya & Yokoburi, 1997), that greenhouse gas emissions from agriculture have decoupled from food production since 1970, and give grounds for optimism that agriculture can make a substantial contribution to reducing global emissions as well as helping to store carbon in land. A reduction in emissions per unit product means that the utilization efficiency of the principle inputs into food production, namely water and fertilizer, has increased. At the same time, crop simulation models have been used extensively to project the impacts of changes in CO 2 , temperature, rainfall, and other factors for global and regional productivity of crops (e.g., Ruane et al., 2017). Resource utilization efficiencies do not operate in isolation; that is to say that there are interactions between, for example, a crop's utilization efficiency of water, nitrogen, and photosynthetically active shortwave radiation. How far these interactions of resource utilization efficiencies are incorporated into crop models is unclear and needs testing, together with a critical need to design and make experiments to test the models. Models should not get the "right" answers for the "wrong" reasons such as via cancellation of errors Martre et al., 2015).
To this end, we propose a methodology based on mathematical identities (Porter & Christensen, 2013) that decomposes water and nitrogen utilization efficiencies and portrays their interactions or trade-offs with water utilization efficiency (WUE). The ideas stem originally from the work of CT de Wit and his colleagues at Wageningen, the Netherlands and have been developed by others (Sadras et al., 2016;Teixera et al., 2014) but has seemingly not as yet penetrated crop modeling as an issue for climate change impacts (Ruane et al., 2017). The identity for WUE and its graphical portrayal while only water uptake was measured.

| Impacts, adaptation, and mitigation in integrated assessment studies
Our second point to improve future impacts and assessment analyses concerns how impacts, adaptation, and mitigation have historically been assessed by different communities using different methods.
This history is reflected in the structure of the IPCC reports, with each of these, especially mitigation, being treated separately. This separation has also been reflected in many policy domains. Recent progress and trends have helped to break down these silos. One example of this is climate-smart agriculture. This idea was borne from the need to integrate climate adaptation and mitigation. Early progress in climate-smart agriculture came through intellectual and political leadership (Lipper et al., 2014), with the evidence base supporting the identification of specific climate-smart agriculture practices coming later (e.g., Rosenzweig et al., 2016). Similarly, introduction of carbon taxes, carbon prices, or greenhouse gas footprint labeling and similar programs necessitates reevaluation of risk and returns in all components of food systems which could include addressing the implications of increasingly frequent disruptions from climate extremes (Lim-Camacho et al., 2017).
Integrated assessment models are one way of assessing the integration of adaptation and mitigation. Efforts to include agriculture in IAMs are relatively new and a number of challenges need to be addressed . While crop models are generally responsive to climate, the range of crops that can be simulated is not sufficiently broad for a full assessment of food security. Furthermore, disparities between IAMs and crop models in spatial scale, treatment of uncertainty, data demand, and representation of agricultural management all limit the extent of crop model integration into IAMs that is currently possible. While significant progress is being made with F I G U R E 1 Decomposition of water use efficiency (WUE). (a) Identity showing the relationship between WUE and radiation use efficiency (RUE) and water, nitrogen and light utilization trade-offs. (b) Four quadrants' visual representation of the identity shown in (a). In quadrants 1 and 2, the thick lines are the upper limits of RUE and WUE, respectively. In quadrant 1, the plateau is the potential grain yield defined as the grain yield that can be attained by current cultivars grown in an environment to which it is adapted with water, nutrients, and other abiotic and biotic factors controlled effectively (Evans & Fisher, 1999). In quadrants 3 and 4, the thick lines are critical N uptake, defined as the minimum N uptake for achieving maximum aboveground biomass at the upper limits WUE and RUE, respectively these challenges (Ruane et al., 2017), it is likely that more than one approach is needed if we are to capture the range of trade-offs and synergies that are important to food systems (Vermeulen et al., 2013) and relevant to policy design and development in the huge variety of contexts that exist globally. One particularly important challenge, for any holistic approach to food systems and climate change, is to develop a framing for research that recognizes that emissions occur across the full range of activities that deliver food security, not only agricultural production (Whitfield, Challinor, & Rees, 2018).
Thus, the idea of climate-smart food systems has emerged as way to take a more comprehensive look at how climate, food, and human activities are interrelated.
Progress in climate-smart food systems can be expected to come from a number of promising avenues. IAMs have the potential to be an important tool for allowing a broader and more complete view of agricultural impacts, adaptation, and mitigation but as argued earlier, can be limited in their ability to include locally important factors. Risk assessment methods provide another set of approaches (Challinor, Müller, et al., 2018). Working with stakeholders and using multiple methods to identify the timing of key risks is one approach that has been shown to work within constrained systems (Challinor, Koehler, Ramirez-Villegas, Whitfield, & Das, 2016) but is not without its costs and risks (Cvitanovic et al., 2019). The review of Challinor, Adger, et al. (2018) found increasing transparency and intercomparability in risk assessments to be an important aspect to future work.
While studies often address uncertainty, the nature of the treatment and the assumptions underlying that analysis are often unclear.  Figure 1a. A free-air CO 2 enrichment experiment conducted over a 4-year period with a spring wheat cultivar at Maricopa, AZ, USA (Kimball et al., 2017) was simulated with the wheat simulation model SiriusQuality (Martre & Dambreville, 2018;Martre et al., 2006). In the first 2 years, wheat crops were grown with high (38.9 g N/m 2 ) and low (7.6 g N/m 2 ) nitrogen supply under ambient (370 ppm; aCO 2 ) and elevated (550 ppm; eCO 2 ) atmospheric CO 2 concentration. In the following 2 years, fully irrigated (665 mm) and water deficit (330 mm) treatments were factorized with the same two CO 2 treatments. In (a) and (b), black dashed lines are upper limits of grain yield calculated with potential radiation use efficiency (2.93 g aboveground DM/MJ photosynthetically active radiation (PAR); Sinclair & Muchow, 1999), harvest index (0.6; Foulkes et al., 2011), and water use efficiency (2.2 g grain DM m −2 mm −1 ; Sadras & Angus, 2006) for wheat, and orange dashed lines are RUE and WUE isopleths calculated with measured data for the control treatment, respectively. In (c) and (d), dashed lines are critical crop N uptake defined as the minimum N uptake for achieving maximum aboveground biomass calculated using the RUE and WUE shown in (a) and (b) and the N dilution curve for wheat (Justes, Mary, & Meynard, 1994). The solid lines between eCO 2 and aCO 2 are drawn to improve the reading of the figure

| CON CLUS ION
The IPCC ARs have evolved over 34 years since AR1. During this time, several themes have become apparent, which we have tried to identify in this review. There has been a plethora of modeling studies on the impacts, with and without adaptation, on a wide range of crops and in many regions. Results from these more than 2,100 studies show consistently both the adverse effect of climate change on a basic element of food security, namely food production and the significant potential value of adaptation in reducing these impacts.
Over the IPCC cycles, an increasing array of mitigation approaches has been treated by both top-down and bottom-up approaches and the range of adaptation options considered has both become more nuanced and broader. These are positive evolutions in the synthesis and evaluation of research that is the role of the IPCC authors and reviewers. However, there are large remaining gaps-particularly with respect to impacts, adaptation, and mitigation in the livestock sector. The lack of quantitative data on livestock in the five ARs was a shock for us as "historical" reviewers; this needs addressing as does an increased attention to nonproduction aspects of food systems. We also suggest a couple of "closer to now" issues on the interactions between resource use efficiencies and the future role of IAMs that may become important in the context of climate change assessment in the near term. In the longer term, future directions for research in agriculture and food will be to ask as much about efficiency and food demand issues as the past has been concerned with adequacy of food supply and environmental outcomes. Thus, issues such as human nutrition and health, diet and obesity, food waste, and circular and local food systems could become dominant themes for food system's research and thereby the foci for future IPCC ARs. F I G U R E 3 The range of approaches that are needed, including expert judgment, interactive scenario building, global systems science, innovative use of climate and integrated assessment models, and methods to understand, project societal responses to climate risks