European Commission logo
English English
CORDIS - EU research results
Content archived on 2024-05-27

Raising the alert about critical feedbacks between climate and long-term land use change in the Amazon

Final Report Summary - AMAZALERT (Raising the alert about critical feedbacks between climate and long-term land use change in the Amazon)

Executive Summary:

(see also AMAZALERT fact sheet number 7, attached as pfd)

The Amazon is under threat through the combined effects of unsustainable regional development and climate change. As summarised in the IPCC reports of 2007 and 2014, studies in the past ten years have indicated that these effects can lead to deforestation, regional disturbance of temperatures and the water cycle, as well as loss of carbon stocks and biodiversity. In turn, these changes can lead to forest loss, droughts, low river levels, floods, loss of hydropower energy and many other ecosystem services and even enhanced risk of diseases and loss of agricultural productivity.
The AMAZALERT project (2011-2014) has: i) addressed and quantified uncertainties in future changes in the Amazon as a consequence of climate change and deforestation; ii) improved the projections of the impacts of these changes; iii) identified potential effective regional and global policies and iv) analysed the possibilities for warning and prevention of large-scale loss of ecosystem services.
The forests of the Amazon are essential in the hydrological cycle to maintain rain, control floods and droughts, store CO2 from the atmosphere, and protect many other ecosystem services. Stakeholders pointed out the importance of the practical services ‘near to the people’ to local populations such as food security, river transportation, hydropower and disease control.

The overall conclusion of AMAZALERT has been summarised as follows:

AMAZALERT brought together a range of global climate predictions from the CMIP5 studies, improved several atmospheric and land surface models and combined them with new scenarios for regional land-use change to assess the likely impact on vegetation and water in the Amazon, in the 21st century. Scenarios were developed through a participatory process, also evaluating policy options, and essential new data have been collected on drought- and temperature dependence. A proposal has been made for an early-warning system for Amazon degradation.

The current generation of climate models (CMIP5) simulates warming of up to 5 - 6°C over Amazonia, by 2100. Although projections of annual rainfall changes are mixed, >80% of models project drier and longer dry seasons, especially in the south and east. Dry season length has a strong relationship with forest area, and the region with a long dry season is projected to expand in the future.

For deforestation, two opposite scenarios were transformed to explicit land-use models. The resulting land-use maps were used to explore the interactions between deforestation with the dynamics of the vegetation, hydrology and climate, using various Earth system models. Imposing high deforestation rates on coupled models resulted in reductions in evapotranspiration and precipitation in Amazonia. Results show that biomass increases in the northern Amazon but in the vulnerable south-east it declines, even in intact forests. Further, the combined effects of land-use change, climate change and fire were investigated. Results show that impacts of climate change including higher temperatures and increased dry season length are enhanced by including land-use change and fire. Results also show a clear impact of land-use change on the water cycle in the entire Amazon basin. However, the magnitude and spatial pattern of the simulated impact is model dependent, which means that there is still substantial uncertainty.

Coupled climate-vegetation models show that if deforestation is low, widespread die-back from climate change alone by 2100 seems unlikely . However, rapid decline cannot be ruled out, because uncertainties remain regarding the sensitivity of Amazon forests to climate and land use change, particularly related to CO2 fertilisation, fire dynamics, incidence of drought and socio-economic developments.

The stakeholder processes carried out within AMAZALERT indicate that compared to other countries importing Amazon goods, and compared to domestic consumption by Amazon nations, in particular Brazil, Europe has a significant but limited direct impact on Amazon deforestation. Particularly, the importance of the EU’s involvement in international initiatives, through influencing trade and domestic processes in Amazon countries, has been highlighted.

In Brazil, stakeholders indicated that actions are needed in the environmental, social, and economic sectors.. In particular, a range of current policies needs to be maintained and enforced, including protecting Conservation Units, PPCDAm, and the Forest Code. Also, valuing forests (PES), diversifying the local economy, and education were singled out as important elements.

AMAZALERT has shown that severe degradation of the Amazon is likely to occur when the climate changes severely and deforestation progresses at the same time. However the type of change can vary strongly and the onset of change can be difficult to anticipate, because it may come too late. Early warning of such change will therefore have to be approached from a broad perspective, combining new and existing networks. Thresholds should be defined that account for society’s coping capacity as well as with the uncertainty in prediction of degradation.

Project Context and Objectives:

Amazonia is under threat through the combined effects of unsustainable regional development and climate change. As summarised in the IPCC reports of 2007 and 2014 studies in the past ten years have indicated that these changes can lead to deforestation, regional disturbance of temperatures and the water cycle, as well as loss of carbon stocks and biodiversity. In turn, these changes can lead to forest loss, droughts, low river levels, floods, loss of hydropower energy and many other ecosystem services and even enhanced risk of diseases and loss of agricultural productivity.

We are only beginning to understand and quantify these threats individually. For example we know that forests are bio diverse, carbon rich and sensitive to climate change and climate extremes, that fire incidence rises substantially in the presence of drought and land use, and that forest loss might affect regional-scale evaporation rates, potentially affecting water supply within Amazonia and beyond its limits, to the bread-basket of South America, the La Plata Basin. We also know that the changing variability of climate can lead to floods as well as droughts in the region, affecting hydropower and agricultural productivity. However, what has been missing until recently has been the capability to combine these different areas of knowledge within one or a few modelling frameworks to understand in an integrated way how the system might respond to coupled global change drivers, and to focus new measurements and new knowledge-gathering processes on our key gaps in biophysical and socio-economic understanding. By combining and developing such model frameworks and data-providing networks, we can develop capacity to predict and identify the possible circumstances leading to loss of Amazonian ecosystem service provision, thereby creating the basis for an ‘early warning system’ that increases societal resilience and improves our capability to avoid large scale or rapid decline of one of the world’s key biomes, the forests of Amazonia.

The AMAZALERT project (2011-2014, a consortium of 14 institutes in Europe and south-America, funded by EU-FP7 and national funds) has been pioneering the multi-level integrated analysis needed to address the environmental and societal risks and aspirations described above. The project has i) addressed the uncertainties and provided novel insight into the likelihood of future change in vegetation, water and carbon cycles in the forests of Amazonia as a consequence of climate change and deforestation; ii) developed new land use scenarios and discussed regional and global policies for the region; and iii) analysed the possibilities for identification and prevention of the large-scale loss of ecosystem services.

More specifically, AMAZALERT aimed to enable raising the alert about critical feedbacks between climate, society, land-use change, vegetation change, water availability and policies in Amazonia:
• To analyse and improve coupled models of global climate and Amazon land use, vegetation and socio-economic drivers to quantify anthropogenic and climate induced land-use and land cover change and non-linear, irreversible feedbacks among these components
• To assess the potential role of regional and global policies and societal responses in the Amazon region for altering the trajectory of land-use change in the face of climate change and other anthropogenic factors, and finally
• To propose i) an Early Warning System (EWS) for detecting any imminent irreversible loss of Amazon ecosystem services, and ii) policy response strategies to prevent such loss.

The notion of ‘climate-induced Amazon die-back’ has attracted more and more attention. Even popular press articles addressed it, e.g. Economist, Sept 23, 2010. Possible ‘die-back’ was strongly emphasised in the IPPC AR4 report (IPCC 2007, WG II, Ch.4) while the so-called ‘Climate gate’ affairs highlighted the associated uncertainties. Meanwhile, the IPCC AR5 WGII report presented mixed evidence, on one hand of low risk of die-back shown by the ensemble of climate model forecasts, while also showing evidence of degradation of the Amazon already happening. Also the risk of uncontrollable die-back plays a prominent role in the market-based approach to a UNFCCC REDD+ mechanism in the Amazon because it could severely reduce the feasibility of using REDD+ credits to meet the EU’s ambitious climate goals (EU 2020 strategy; Gumpelberger et al, 2010).
It has become increasingly clear that global climate change could potentially negatively affect Amazon land cover over the next 30-50 years (Sampaio et al., 2007; Nepstad et al, 2008). The emerging picture until 2010 was that deforestation of more than 40% could significantly reduce rainfall, and global climate change, even at 2°C warming (Jones et al, 2009) has the potential to lead to irreversible degradation of the Amazon basin’s current ecosystems to a more savannah- type ecosystem (Cox et al. 2000, Scholze et al. 2006, Malhi et al, 2008). Such die-back would not only affect regional water resources, biodiversity and livelihoods but has been predicted to also affect global climate as far as Europe and North America through alternations in energy exchange and circulation (Werth and Avissar, 2002; Marengo, 2005) as well as through substantial emissions of greenhouse gases associated with forest loss (Phillips et al., 2009; Cox et al., 2004; Poulter et al, 2010). However, predictions of die-back remained highly uncertain (Nobre and Borma, 2010).

Climate and land-use impacts in Amazonia are highly uncertain. We lack understanding of the role of human-driven changes of Amazonian land-use patterns in the climate system and vice versa (Nobre and Borma, 2010). During the year 2010, Amazonia was witnessing an extremely dry year again, as it did in 2005, but neither the cause of these droughts nor their effects on the forests is well-understood (NatureNews, Oct. 2010). One of the members of the present AMAZALERT consortium recently edited a special feature of the scientific journal New Phytologist devoted to the uncertainties of vegetation dynamics responding to drought, arguing that “our ability to understand the true vulnerability to loss of the region’s forests has been hampered, first, by insufficient field data and, second, by a weak understanding of the predictive skill of the models themselves.” (Meir and Woodward, 2010). Drought, temperature and CO2 response of vegetation are highly uncertain, while the effect of CO2 can ‘make or break’ climate-induced die-back (Rammig et al, 2010; Galbraith et al, 2010). The interaction between land-use change, fires and climate probably further enhances vulnerability, and climate models themselves differ widely in their predictions (Friedlingstein et al., 2006). A primary focus of AMAZALERT has been to improve modelling of key dynamics involved.

After decades of increasing forest conversion deforestation rates are now at an all-time low now and still decreasing (INPE, 2010), despite sustained economic growth in Brazil. Clearly, society and policies have an effect on land-use changes, both in a positive and a negative sense. It is possible, however, to assume that society, including regional, national and international communities, could ultimately provide resilience to the degradation process through policies and trade pressures, and there is uncertainty as to how effective or responsive future socio-economic systems and/or policies can be to limit regional and global change (Raskin et al, 1989). National legislation, development plans, regional economic growth, as well as international trade in agricultural products and biofuels, and international treaties and conventions on climate change and biodiversity, all affect what happens in the Amazon region. AMAZALERT aimed to translate these drivers of land-use change into scenarios linking to quantitatively modelled land-use change.

Whether Amazon forest ‘die-back’ is realistic or not, there is little doubt that the Amazon basin is under threat from climate and society, and it is of great importance to enable South-American society, the EU and the international community to inform itself of any trends and the risk of major loss of Amazonian ‘ecosystem services’ (e.g. carbon storage). While systems are in place to monitor land-use change and rivers (e.g. ANA, PRODES-INPE 2010), there is little else operational to systematically monitor the state of the climate, the biodiversity or even the regional economy. When, as stated above, we know more about the risks and mechanisms of loss of ecosystem services, it should be possible to design an operational, basin-wide system that enables us to raise the alert with a clear evidence base, before it is too late to intervene.


AMAZALERT consisted of seven work packages (WP1-WP7), of which WP1-5 are RTD, WP6 was devoted entirely to dissemination and WP7 dealt with consortium management. AMAZALERT analysed and improved models of climate, land use, and vegetation, while furthering knowledge on (climate) policies and societal responses for the Amazon region in the face of climate change and land-use pressures; and proposed policy response strategies as well as an Early Warning System (EWS) for irreversible loss of Amazon ecosystem services. Fig A.1 shows the project logic and information flow, as depicted in the AMAZALERT project logo.

At the outset, ecosystem functions have been prioritised, using existing biophysical analyses and consulting various stakeholder groups (WP1). Amazon land cover and vegetation models (DGVMs) as well as regional models of river discharges have been used, improved and benchmarked against data sets (WP2). Improvements in the DGVMs were aimed to constrain the land surface response in coupled models (WP3). At the same time, existing global and regional simulations by coupled climate-land surface models have been analysed and improved to find critical indicators of irreversible change (WP3). Combining the DGVM results of WP2 and the coupled climate models of WP3 enabled us to quantify and understand the likelihood of changes in and irreversible losses of the prioritized ecosystem functions.
Land-use change models have been linked to the DGVMs in WP2, driven by socio-economical scenarios and related policy options that have been developed in WP4 based on a selected set of national and international policies and guided by stakeholder groups. WP4 then analysed the implications of the qualitative scenarios for policy instruments, but, more importantly, also the impact of (modelled) changes in the Amazon, both gradual and irreversible, on the effectiveness of policies and regional development plans. In collaboration with WP2, WP4 also identified which socio-economic factors can serve as (quantitative) indicators of irreversible change in the Amazon, for use in WP5.

Finally, in WP5, a ‘blueprint’ for an Amazon-wide Early Warning System has been developed. To achieve this, we used the theory on the non-linear dynamics of (eco)systems near ‘tipping points’ and combined this with the insights on priority ecosystem functions of WP1, the critical indicators of irreversible loss from WP3 and WP2, the socio-economic indicators from WP4, as well as insights from existing monitoring systems. At this stage, we again interacted with stakeholder groups, now in their role as potential users of an EWS, to generate a blueprint that will be as effective as possible.

Dissemination has been organised in WP6 and covered a range of activities, importantly the science-policy interaction. First, as the project engaged Amazonian stakeholder groups right from the beginning, these groups were at the core of dissemination activities throughout the project’s lifetime. Yet, a much wider audience was actively engaged towards the end of the project, for example through interaction with the public media, as this is particularly effective in South-America, or through effectively interacting with the wider scientific community. Dissemination outside of the Amazon focused on two groups in particular: entities engaged in REDD+ and EU policy makers with relevant mandates. These target groups were engaged through a meeting in Brussels. To ensure successful collaboration between partners and WPs, the key partners from Europe (ALTERRA) and South America (INPE) both had a role in WP7 (Management and Coordination). Thus, regional matters, meetings, workshops and communication as well as modelling activities have been effectively managed.

Project Results:

The main results of AMAZALERT have been summarised in a summary for policy makers (see appendix 2, attached as pdf, and deliverable D6.7). In this section, we first reflect this summary, and then report systematically on the results per itemised project task.



This scenario approach was participatory, involving representatives of civil society, the productive sector and government. The workshops also addressed possible (policy) solutions within Brazil and possible strategies from Europe. One of the objectives of the new scenarios was to consider a broader sustainability discussion, while being aligned to global scenarios. The scenarios vary from Low to High Social Development and High to Low Environmental Development, and are aligned to the new IPCC AR5 Socio-Economic pathways. The methods combined exploratory and normative approaches, addressing policy options to build a trajectory towards sustainability, synthesised in Box 1. Policy options for reducing deforestation even in the worst scenarios were also discussed in some of the workshops (see result 8).

These qualitative scenarios were transformed to explicit land-use models using the open-source platform LuccME ( The resulting annual land use maps (2005-2100), were used to explore the interactions of deforestation with the dynamics of the vegetation, hydrology and climate, using various Earth system models.

Box 1 – Synthesis of actions towards a sustainable future (AMAZALERT stakeholder scenario workshop results)
(a) MONITORING SYSTEMS: continuation and enhancement of the satellite based monitoring systems initiated at PPCDAM, considered as the key aspect to control deforestation. This includes the development of new systems (based on new sensors, for instance), and expansion to other biomes, to avoid leakages.
(b) INTEGRATED TERRITORIAL PLANNING: consolidation and enhancement of multiple instruments for territorial and land use planning, in order to concomitantly regulate pressure for land, create sustainable economic alternatives and integrate social programs at a territorial basis. This includes private and public lands (such as conservation units, indigenous lands, settlements), rural and urban areas.
(c) CITIES RESTRUCTURING: Strengthening of cities to create an interconnected network of medium-sized cities, with infrastructure, proper network of services and education to meet the demands of sustainability.
(d) LARGE INVESTMENTS PLANNING: Planning for the implementation of large projects (including infrastructure and mining) combined to the integrated territorial planning (item B), avoiding the boom-bust economies of the cities. In the case of infrastructure, planning geared both to the needs of the local population (river transport, for example), as well as market demands (commodities production flow through waterways).
(e) LEGAL FRAMEWORK PROTECTION: enforcement and enhancement of the legislation governing the access to natural resources and land use, creating mechanisms to balance the influence of macroeconomic interests in modifying legal marks at the expense of regional, social and environmental aspects.


The forests of the Amazon and their biomass are essential to maintain rains, control floods and droughts, store CO2 from the atmosphere, and many other ecosystem services. Stakeholders in particular point out the importance of the practical services ‘near to the people’ such as food safety, river transportation, hydropower or disease control. Degradation of biomass will lead to degradation of these services.

AMAZALERT has brought together a range of global climate predictions from the CMIP5 studies, improved several atmospheric and land surface models and combined them with new scenarios for regional land-use change to assess the likely impact on vegetation and water in the Amazon, in the 21st century. These model simulations using scenarios of deforestation developed under AMAZALERT have been carried out to investigate the effects of alternative deforestation futures. In coupled models, the effect of imposing the high land use scenario in the Amazon basin results in reductions in evapotranspiration and precipitation in Amazonia compared to the standard CMIP5 projection (RCP8.5).


By combining experimental data and simulations using large ensembles of models, AMAZALERT has found that when deforestation is kept to a minimum, projected changes in climate alone seem unlikely to bring about large-scale forest die-back. We investigated the forest response to climate change by 2100 according to three greenhouse gas concentration scenarios as simulated by an ensemble of the HadCM3C climate-carbon cycle model. Although there is a range of responses, risk of forest loss increases with the emissions scenario severity. On average the forest is projected to remain largely unchanged in extent in the first two scenarios. In the third – and strongest – scenario (RCP8.5) the mean average reduction in forest cover remains marginal but the risk of substantial loss is significant, whilst the probability of forest gain is minimal; in this last scenario, however there is much greater uncertainty in the extent of possible forest loss, ranging from almost no loss to 50% or more.


Despite its relative resilience if deforestations stays low, AMAZALERT has shown that severe degradation of Amazonia is possible when severe climate change and deforestation progress simultaneously. However the type of change can vary strongly and can be difficult to predict, because signals of change may come only after a biophysical threshold has been passed when decline will already be rapid or irreversible. Early warning for such change will therefore also have to be approached from a broad perspective. Here, basin-wide monitoring of climate change and the frequency of climate extremes, moisture availability, biomass and carbon exchange metrics, will need to make combined use of new and existing networks, both in situ and using remote observation (airborne or space borne), to detect large-scale risk of the loss of ecosystem stability, and to help predict likely high-risk scenarios. Thresholds should be defined that account for society’s coping capacity as well as with the uncertainty in prediction of natural ecosystem degradation or instability. In this envisioned early warning system (EWS), new scientific insight and technical capability should be constantly adopted and tested for effectiveness.


Novel field data collection in AMAZALERT has focused on quantifying the effects of increased temperature on photosynthesis and on understanding the processes governing differential species mortality during drought. Leaf warming experiments carried out in the canopy of experimentally droughted and non-droughted forest did not suggest that the maximum rate of photosynthesis changed significantly following warming, and that current vegetation models are generally overestimating temperature sensitivity. In relation to drought sensitivity, our data from experimentally droughted forest show that some species, and large trees, are particularly vulnerable to drought, potentially risking significant biomass losses. These mortality differences appear to be related to increases in respiration (CO2 emission to the atmosphere) in drought-sensitive species. Overall, these data suggest some resistance to climate stress in carbon uptake, but more sensitivity in respiration. The net effect means that a long-term change in the species composition and structure of the forest under extended drought and warming is likely (if no fire occurs), together with an increase in the amount of dead and decomposing wood.


We defined three extreme land use scenarios (low and high deforestation, also including high biofuel targets) linking different land use policies to possible impacts on the provision of ecosystem services in Amazonia. The resulting land cover maps were used to estimate the impacts in the provision of key ecosystem services in the region, using four dynamic vegetation models: INLAND (Brazilian), ORCHIDEE (French), LPJml (German) and JULES (British). There is a clear impact of land use change on the water cycle in the entire Amazon basin, with increasing removal of forest area leading to significant decreases in evaporative flux, increased variability in river flow and lower potential for hydropower. However, the magnitude and spatial pattern of the simulated impact depends on the model used, which means there is still substantial uncertainty about the precise magnitude and location of these impacts.


The work undertaken for AMAZALERT indicates that the southern and eastern Amazon Basin is more vulnerable to changes than the north and northwest. In the coupled model HadGEM2-ES, it was found that underlying conditions can affect the drying impacts of deforestation. In areas with a long dry season – the south and east – deforestation caused a bigger reduction in evapotranspiration than areas where the dry season is short – the north-west. The range of deforestation scenarios along with three climate scenarios were used to drive four Dynamic Global Vegetation Models to 2100. Results show that biomass increases in the northern Amazon but in the vulnerable south-east it declines, even in intact forests. Further, the combined effects of land use change, climate change and fire were investigated in a land surface model (the Brazilian Earth System Model, BESM). Results show that impacts of climate change including higher temperatures and increased dry season length are enhanced by including land use change and fire.

The current generation of climate models (CMIP5) simulates annual warming in Amazonia relative to late 20th century levels of between 0.8°C and 7.3°C, capturing the range from a low-sensitivity model under a low emissions scenario (RCP2.6) to a high-sensitivity model under a high emissions scenario (RCP8.5). Although projections of annual rainfall changes are mixed, under RCP8.5 >80% of models project drier and longer dry seasons, especially in the south and east. Changes in dry season length have already been observed in Amazonia, with earlier onset and later demise seen in the last decades, such as during the devastating drought year of 2010. Dry season length has a strong relationship with forest area, and the region with long dry season is projected to expand in the future. Drought severely impacts the occurrence of fires.


Beside the scenario participatory process (Research Result 1), interviews and other workshops were held in the scope of AMAZALERT to discuss specifically drivers and actions to decrease deforestation, in any of the giving scenarios.

Currently, the three most important drivers of deforestation are livestock, mechanised agriculture and associated logging, together with underlying causes related to land tenure. These are closely connected to national and international trade of Amazon products. Drivers are projected to change towards 2050, with important drivers now including large infrastructure programmes and mining for oil and gas. However, as before these are strongly related to domestic consumption and interregional trade.

Besides the influence of trade on deforestation, investments such as in hydropower and mining were mentioned by stakeholders as being important. Stakeholders also perceived the strengthening of the civil society and social cohesion as important given its potentially major role in reducing future deforestation in different development scenarios. Domestic actions within the Amazon nations were stated to be of highest priority due to their proven success in reducing deforestation and the importance of measures coming from within a country (see Research Result 1).

Stakeholders during the European workshop indicated that Europe could potentially have a limited, but significant, impact on deforestation in the Amazon. They suggested that all efforts together could result in a 25% decrease of deforestation rates. Yet, as long as demands for agricultural products and energy from other countries, such as China, continue to increase export might simply shift from Europe to elsewhere. In any case, EU impact is less than from other importing regions and domestic consumption. Stakeholders highlighted the importance of the EU’s involvement in international initiatives. These should include the establishment and strengthening of trade standards and certification. It was suggested that EU induced deforestation drivers could be addressed by reduced import through increased efficiency at the demand-side and of production within the EU. At the same time, stakeholders suggested that enhancing demand for products that meet high environmental standards may reduce negative implications and will generate momentum towards an overall increase in environmental standards. One example that was discussed to have shown a positive impact is the soy moratorium.


AMAZALERT has been organised along five research work packages (WP1-WP5) , one dissemination work package (WP6) and one co-ordination work package (WP7). Here, we systematically report on the committed work package tasks.

WP1 - Identifying priority Amazon ecosystem functions and services and their drivers of change

This work package had two objectives:
1.To select the various stakeholder groups and manage their contributions during the project
This responsibility mainly served the tasks of WP4, and hence much of this work has been conducted and is reported with WP4
2.To identify and prioritize the most important large-scale Amazonian ecosystem services and drivers of change.
The identification of key ecosystem services has been carried out during the first phases of the project

Task 1.1 Identification and management of stakeholders through the project
Stakeholders and other experts were interviewed to prioritize their ecosystem services. This was compiled in the deliverable 1.3 in the first reporting period. Stakeholder groups were then identified and respective workshop organized in close collaboration with WP4. These workshops were conducted to identify qualitative storylines for future land-use scenarios through the stakeholder dialogue. These scenarios are important to quantify future changes in the provision of ecosystem services in the Amazon. The outcome of the co-organized workshops was described in Deliverable 4.2 for example, which describes how the stakeholder dialogue was used to develop land-use scenarios for the Amazon and what the important drivers are. The results of the stakeholder dialogue, and first indications for the impacts on ecosystem services, were presented to the experts at the final stakeholder meeting in Belem, Pará, on 6th Oct 2014. Stakeholders gave their final feedback on the process and discussed what the AMAZALERT results mean for their own work.

Task 1.2: Exploring ecosystem services and drivers of change.
At the first project and stakeholder workshop in October 2011, seven ecosystem services were selected as the most relevant in the prioritization (consumptive use, carbon storage in intact forests and soils, maintenance of favourable climate, subsistence agriculture, fishing, providing living space to wild plants and animals, and protection of biodiversity, see Figure A.2). These ecosystem services are well known for the project partners considering that they are in the research area and they know the benefits for the climate regulation. On the other hand, if the questionnaire would have been filled by local stakeholders, maybe they would not have identified the carbon storage and the maintenance of favourable climate as high priority due to the fact that their local reality and needs are more urgent (e.g. food, fuel, water, etc.)

The main drivers of change for the seven ecosystem services identified were also analysed. The drivers of change were split in two categories, deforestation (or land use change drivers) and climate change drivers. The key drivers of change identified that threaten the ecosystem services were large scale agriculture production and infrastructure (river dams, roads, settlements, expansion of cities, etc.), following by slash and burn, wood industry (legal and illegal) and cattle ranching.

As part of this task, a conceptual review on Amazon ecosystem functions and services was done (D.1.4 Review paper on Amazon ecosystem functions and services and their drives of change). The results from this review showed that the ecosystem services concept has been widely used in the last years; however, there is still a wide scientific debate about its interpretations, definition, classification systems, framework, and use at different scales. With this review we identified and contrasted ecosystem services definitions and how these have been used, interpreted, or differentiated. In addition, we aim to distinguish between the provisioning and regulating ecosystem services and the benefits they provide at different scales and how they are perceived by local and global stakeholders. We highlight the main ecosystem services in the Amazon and the potential impacts of the drivers of change (mainly deforestation and climate change) on these ecosystems. Furthermore, the role of protected areas and indigenous territories in the preservation of the Amazon ecosystem is highlighted.

WP2 - The response of land use change, vegetation and water to anthropogenic drivers and climate change

This work package had the following objectives:
1. To assess the uncertainty on an ensemble of climate/land use scenario runs for the 4 different state of the art process-based vegetation models and scenarios to set a baseline.
2. To provide new land use scenarios from an improved land use model
3. To harmonize existing data and collect targeted new data
4. To identify model deficiencies by comparing the baseline runs with data.
5. To improve and expand dynamic vegetation modelling capacity on key processes, in order to use these improved models in ensemble runs (WP2) and earth system models (WP3)
6.a To assess impacts of land use scenarios and basin scale hydrology responses.
6.b To assess the uncertainty on an ensemble of climate/land use scenario runs for the different models using improved vegetation models and improved scenarios; and combining these ensemble runs with large scale datasets to improve our understanding of the processes through which climate and land use change affect the Amazon.

Task 2.1. Scenario runs with state of the art DGVMs and scenarios .
Following an initial planning meeting, January 2012 in Leeds, on how to develop the DGVM comparisons, an agreement was made between AMAZALERT and the Andes-Amazon Initiative (AAI) project (Moore Foundation) led by Paul Moorcroft (Harvard) to share protocol, driver datasets and validation datasets used by the AAI project. A consolidated dataset of historical baseline model runs is available for 4 DGVMs (INLAND, LPJml, ORCHIDEE, JULES) on This dataset is described in detail in deliverable report D2.1.

Two important analyses have been performed specifically on the historical models runs. These runs were first presented and discussed at a second WP2 workshop in Gent, in January 2014. A first analysis, led by Michelle Johnson (ULeeds), studies the spatial variability in biomass, woody NPP and mortality rates in the four DGVMs in comparison with new kriged maps based on plot network data (Johnson et al. in prep.). The main conclusion from this analysis is that biomass and vegetation carbon is more related to mortality than to NPP, and this is the reason that most DGVMs, which usually contain rather sophisticated NPP simulations but make crude and uniform assumptions on mortality, simulate biomass badly in comparison to data.

A second analysis of the historical model runs is focusing on the responses of the vegetation carbon cycle to historical drought events in the period 1970-2008. In this study, led by Hannes De Deurwaerder (UGent), we investigated multiple datasets (flux tower data, plot network inventory data) on their ability to capture the impacts of the historical drought events (De Deurwaerder et al.; in prep.).

Future DGVM runs

To simulate the future vegetation of the Amazon, it was decided to use the climate forcing’s of the IAA protocol as well, in order to keep the future runs consistent with the historical runs. Those scenarios were combined with the land use scenarios that were constructed by INPE in task 2.2. This resulted in an AMAZALERT specific modelling protocol, available on and described in deliverable report 2.5. The first results of the coupled land-use DGVM simulations were presented on the AMAZALERT meeting in Alter de Chao in October 2014. In deliverable reports 2.4 and 2.5 the results of the four DGVMs (INLAND, LPJml and ORCHIDEE) that performed the future mode runs are presented. This analysis is further presented under task 2.6 and 2.7.

Task 2.2. Land use change/cover modeling .
The three spatially explicit land use scenarios (1997-2050) were developed by INPE as an initial input to WP3 and WP2 model runs. The scenarios are built using the LuccME generic modelling framework, also developed at INPE, as described under task 4.2. We also present the LuccME/BRAmazonia model parameters, the current scenario premises and results. After WP4 stakeholder workshops, new LuccME quantitative scenarios have been generated. The resulting land-use maps used in modelling are illustrated in figure A.3

Task 2.3. Harmonization of existing key datasets and targeted new data collection

Harmonized existing datasets
Several datasets have been made available by various AMAZALERT partners. These datasets are centrally available for model-data comparison on (password protected). These data include: biomass data from the plot network, rainfall exclusion experiment data, hydrology data and flux tower data. These datasets have been intensively used for model-data comparison (see other tasks and deliverable reports).

New data collection
Two coordinated new data collection efforts were performed to investigate leaf temperature response and mechanisms responsible for differences previously observed in drought vulnerability across tree species, respectively.

In October 2012, and May and October 2013 (covering the wet and dry season) CO2 and light response curves were carried out at five different temperatures (25-45 °C) at long term heated leaves and non-heated leaves, to derive photosynthetic parameters (e.g. Jmax and Vcmax) under different temperature levels. Measurements were taken in three individuals located in control and drought plots in the long-term moisture manipulation experiment carried out in the Caxiuana National Forest, Brazil, which allowed us to investigate the combination of drought and leaf heating on gas exchange. Leaves were accessed from the two towers, limiting the representativity of sampling. Additional measurements included stomatal conductance (porometer) and leaf water potential; we also sampled leaves for chemical analyses. To understand gas exchange response to long-term heating (acclimation), leaf heaters were installed in a set of canopy and subcanopy leaves in both control and drought plots. The heaters are powered by solar panels and their heating efficiency is monitored by thermocouples. The heaters have persisted well despite harsh field conditions. Resulting temperature dependences of Vcmax and Jmax were used to fit parameters in continuous temperature response functions used in various models.

The resulting dataset has been archived and organized in data base form, and a methodological report, detailing all procedures, has been written (D2.2). Resulting curves have been analyzed, and a publication is in preparation (Kruijt et al., 2015). In figure A.4 normalized average temperature responses of Vcmax, Jmax and Rd in the light are shown. Vcmax in the non-heated leaves show an increase with temperature, but with a dip in the mid-30 degree range. For the heated leaves there is an optimum for Vcmax at 38 °C for the control plot and at 35 °C for the drought plot. Even though Vcmax may not, or hardly show a temperature optimum, re-calculated photosynthesis (carboxylation rate) itself does. The optimum temperature is clearly higher than in non-heated leaves. Therefore we can conclude that there is some evidence that photosynthesis acclimates with environmental temperatures. The temperature dependence equations that are used in 5 DGVMS were fitted to the experimental data, generating curves that are close to the measured ones (right top panel). Figure A.4 also shows that in case of the Jules and ORCHIDEE models there is a large difference between the default parameterization of these models and the fitted curves using the new parameter values. The recalculated carboxylation rates (bottom right panel) also show big differences in temperature dependence between the fitted curves and the default parameterization of the models. Vc for Jules, ORCHIDEE and LPJmL showed higher values and Inland showed lower values using the fitted Vcmax. CLM showed minor effect.

A second, targeted new set of measurements was designed to assess the mortality risk in two groups of trees, drought-resistant and drought-vulnerable trees. Substantial field work has been dedicated to this. The study monitors: 1) carbon storage and fluxes, using chamber measurements of respiration and photosynthesis, and measurement of non-structural carbohydrates, 2) hydraulic vulnerability using pressure volume curves on branches and leaves, leaf water potential and sap-flux measurements, 3) leaf herbivory. We are using this comprehensive set of measurements to piece together why some trees are more prone to mortality via drought than others. Initial results show no difference in the photosynthetic parameters (Vcmax and Jmax) between the drought and control forest, however there is a significant increase in leaf and stem respiration in the droughted forest, which is driven by elevated respiration in trees which are vulnerable to drought. Our hydraulics data are partially consistent with our grouping into drought resistant and vulnerable species. The implications these new data have for understanding and modelling the response by rain forest to drought are under review, have been presented at the 2014 AGU in California, and will be published in the near future.

Task 2.4. Multiple dataset evaluation to identify DGVM deficiencies:
For this task important results have been achieved. In the first place we refer to the studies on biomass maps (Johnson et al. in prep) and on drought responses (De Deurwaerder et al. in prep), which we discussed above under Task 2.1. In these studies all 4 DGVMs are compared with data. In addition deliverable D2.4 includes important model-data comparison focusing on basin hydrology.

Several studies initiated by LBA-MIP (de Goncalves et al. 2013) and Moore AAI were completed over the last 2 years by AMAZALERT researchers. In these studies the DGVMs (including AMAZALERT models ORCHIDEE, JULES, LPJ and IBIS/INLAND) are compared at site level with flux tower and other field data. This includes a study on the effect of water supply and demand on ET seasonality (Christoffersen et al. 2014), a study to compare different formulations of root water uptake in models and an evaluation of their simulated soil moisture dynamics with data from experimental drought manipulations (Christoffersen et al. in prep), a study to compare the combined effects of temperature and drought on simulated ecosystem productivity (Rowland et al. 2014), and a study on inter annual variability of carbon and water fluxes (von Randow et al. 2013).

In addition to the mentioned multi-model analyses, several single-model studies were performed by the individual modelling groups that contribute strongly to the identification of model deficiencies, including:
• A study where two soil models for ORCHIDEE were compared with multiple hydrological datasets over the Amazon basin (Guimberteau et al. 2014).
• Joetzjer E, C. Delire, H. Douville, P. Ciais, B. Decharme, R. Fisher,B. Christoffersen, J.C. Calvet, A.C.L. da Costa, L.V. Ferreira, P. Meir. Predicting the response of the Amazon rainforest to persistent drought conditions under current and future climates: a major challenge for global land surface models. Geoscientific model development (accepted).

Task 2.5. Development of DGVM to improve representation of relevant processes
An extensive review of the state-of-the art of vegetation models and needed model improvements has been described in deliverable D2.3. Several specific model improvements have been performed (see previous periodic report) and are included in the future model runs (task 2.6 and 2.7). While multiple important possible improvements remain to be included in the models based on what we have learned from the project outcomes (such as improved temperature response, drought response, mortality functions, and nutrient dependence), some processes have been fully implemented in smaller-scale models for testing and evaluation prior to their inclusion in the DGVMs. As an example, a novel full representation of plant hydraulics has been developed and implemented in an individual tree model and it captures key aspects of known hydraulic constraints and risks associated with drought-induced mortality, including tree height-dependent hydraulic path length and trait-linked variation (via wood density) in xylem vulnerability. This has provided valuable insight into the relevant formulations necessary for inclusion into the JULES DGVM whose novel development in this crucial area has now significantly advanced, as well as its wider inclusion into DGVMs generally.

Task 2.6. Synthesis/integration of improved ensemble model runs
For consistency with the historical model runs, all future runs were performed based on the protocol of the Moore Foundation Andes-Amazon Initiative (AAI) project with the supplementation of some extra outputs (mainly hydrology related) required for AMAZALERT. Each model performed 12 future runs for a time period of 91 years, covering 2009 till 2099. For initialization, the models used a fixed Land Use map taken from the year 2008 of the land use scenario’s. This represents a historical run without land use, following the Moore protocol, but using that single constant land use map run from 1970-2008. The result for 2008 was used as a starting point for the future predictions. The models use, for all subsequent years the INPE LUC maps (task 2.2) and CO2 dataset under IPCC SRES A2 scenario. An example of the model results is presented in figure A.5

One noticeable finding is that the maps and timeline analyses generated suggest a potential biomass growth in the western and northern regions of the Amazon, while even in intact forests, a decline might be expected in the more vulnerable southern and eastern ecosystems, independent from - however strengthened by - more intense land use practices. Secondly, findings indicate a potential alteration of the water cycling, enhanced by deforestation. Additionally, resulting from the HadCM3 future forcings, we can conclude that carbon fluxes and stocks seem sensitive to sudden, unpredictable and intense climate extremes and variability, which strengthens the importance of studying the vegetation responses to these hazardous events. So although some regions of the Amazon appear quite resilient to gradual climate change, potential dieback due to more unpredictable climate patterns is still conceivable. Finally, insights resulting from both the historical and future runs strongly encourage future research on inter-model differences, model set-up improvement (i.e. by including mechanistic models on mortality) and vegetation dynamics on climate extremes.

Task 2.7. Simulating impacts of climate, land use and policy on river runoff
This task consisted of different research activities performed in AMAZALERT Work Package 2 related to analyses of possible impacts of climate change and land use / land cover change (LUCC) on components of the hydrologic cycle of Amazonia (evapotranspiration and river discharges) and possible implications on the provision of ecosystem services such as water yield for hydropower generation.

Four different land surface models and a distributed hydrological model were run following a standard protocol, forced by scenarios of climate change and three different land use scenarios projected by a land use model: a sustainable, environmental development scenario and two levels of intensive deforestation scenarios, one with a mid-level land demand and one with a strong land demand that emulates strong biofuel targets.

Results obtained in these analyses reflect the collaborative aspect of the project. Partner CNRS with collaborations from UGENT, PIK and INPE analysed the relationship between climate, land-use patterns and evapotranspiration patterns. Partner INPE focused on the smaller scale, using its distributed hydrological model MHD-INPE, and studied impacts on hydropower plants at three sub-basins (tributaries of the Amazon). Partners PIK and ALTERRA also collaborated on analyses of hydropower dams but at the Amazon basin scale. Additionally, Partner UNAL with collaborations from INPE, UGENT, CNRS and PIK analyse average and extreme flows, using basin-wide runoff–climate statistical relationships with outputs from the hydrological model outputs and observations. A small workshop was held at 24 November, 2014, to co-ordinate future joint publications of this group and on these themes.

The consequences of climate and land-use change for the Amazon regions’ water resources are of the most palpable for the regional society. Extremes, low river levels such as in the extremely dry years of 2005 and 2010 (Marengo et al. 2011) are of direct effect to local livelihoods.

It is likely that various policies and development plans for the Amazon region are highly sensitive to disturbances of water resources, and conversely, large-scale hydrology plans such as dams and waterways are likely to affect the river discharge itself. The standard wisdom in landscape hydrology is that deforestation leads to lower evapotranspiration (as surfaces are smoother) and increased runoff, i.e. to increase river discharge. If this were the case for the Amazon, deforestation should lead to increased flooding, more extreme low discharges, and in general should reduce the downwind moisture transport, and recycling in the basin. Compilation of data over the last decades, however, and small-scale catchment studies, seem to show that this effect is absent (Rodriguez et al 2010). Explanations are not clear, but could be related to enhanced energy redistribution, or mesoscale circulation enhancement, between fragmented forests. During AMAZALERT, a detailed study of the impacts of climate and land-use change on the hydrological cycle and functionality of hydropower plants in the Amazon was performed. A standardized model intercomparison protocol was used, with uniform forcing data and land use / land cover projections that served as inputs to drive 4 different large scale land surface models with dynamic vegetation, of which 2 were also explicitly simulating river discharges and one additional detailed hydrological model calibrated to key sub-basins in the Amazon region. The underlying scientific question is: to what extent do deforestation and spatial patterns thereof, and dams, affect rainfall-runoff ratios and how does climate change modify this through changed rainfall and evaporative demand?

These questions were then addressed at two scales:
1) At the amazon-wide scale, using large-scale river routing schemes in ORCHIDEE and LPJml models, as well as a statistical scaling approach (HydroSIG, Poveda et al, 2007).
2) At the scale of three major South Amazon tributaries: the Madeira/Mamore, the Tapajos and the Tocantins river basins. These catchments are selected as test study cases that are both located in dynamic frontiers since include savanna-rainforest transitions and are both part of Brazilian agriculture frontier, under strong pressure for land expansion, and have important hydropower plants currently installed and with plans for new developments.

Within Amazalert WP2, possible effects of climate and LUCC on river discharges and functionality of existing and planned hydropower plants were studied with focus on sub-basin scales (tributaries of the Amazon river), using INPE’s distributed hydrological model MHD (Tomasella and Rodriguez, 2014; Rodriguez, 2011), and on wider scale of the whole basin, the global vegetation and hydrology model LPJmL (Rost et al, 2008).

In the Madeira basin, climate change projections by most models produced a reduction of discharge along the whole year in future time slices, with a consequent reduction in potential energy production, but this reduction was largely counterbalanced by LUCC effects. In some models, however, (MR5, CSR, IPS and HD2), wet and dry season discharges changes due to CC have opposite signals, which resulted in lower than the baseline annual discharges and, then, in the energy potential. Under LUCC effects, both maximum and minimum discharges increased, resulting in a more consistent change in the energy potential in those projections.

For the TOCANTINS basin, the MHD model was run including LUCCME simulations that include the secondary forests, which provided an opportunity to analyse how the recovery of forest cover can also influence discharge and hydropower generation. The changes caused by CC and deforestation were found similar to that obtained for Madeira basin, with a lower decrease in the discharge caused by the CC when LUCC of scenario C1 is taken into account, mainly because of deforestation. However, the decrease in the energy potential in the basin may again become pronounced when the low-impact scenario A is considered. This happens due to the deceleration of the deforestation added to a conversion of pasture into secondary forest, since the primary forest changes little in this scenario.

Considering that the results presented in this work do not include scenarios for water consumption, it follows that the reduction found on the potential energy production can be further exaggerated. Despite the high variability among the model integrations, the annual energy production is likely to decrease until the end of the century for the projections analysed under the current HPP design. Although the IPSL scenario in Tapajós points to an increase in rainfall, this is not reflected in the annual production because the constraint is given in terms of plant capacity.

We analysed hydropower dams at the Amazon basin scale using LPJmL simulations created by forcing LPJmL with observed climate (Sheffield et al, 2006), under potential natural vegetation, performed by WP2 partners PIK and ALTERRA. The objective of this study is therefore twofold. The first is to assess the impact of climate change and land use change on the production capacity of operational and planned hydropower plants and the second to evaluate the potential impacts of operational and planned hydropower plants on downstream ecosystems and ecosystem services.

We used the global vegetation and hydrology model LPJmL (Rost et al, 2008) to calculate the effects of climate change and land use change on the future hydrology of the Amazon. Subsequently, the model was driven with climate scenarios from 3 global climate models for the AR4 SRES A2 scenario (HadCM3, PCM and CCSM) combined with 3 land-use scenarios, representing different levels of deforestation (Aguiar et al, 2012).

We observe that two contrasting patterns are projected for the Northern and the South-Eastern parts of the basin (Figure A.6). The major part of the Northern tributaries will experience an increase in average annual streamflow, however decreases in streamflow are expected in the South Eastern part of the basin, where many new hydropower plants are planned. Still, according to these simulations, changes in land use from forest to agricultural land will only have small effects on mean annual discharge.

We analysed projected changes in mean annual flow at the locations of operational and planned hydropower generation plants. Most of the planned hydropower plants will be located in regions where increases in mean annual streamflow can be expected, although over 20.000 MW installed capacity is planned on locations with projections of decreasing streamflow.

One could conclude that most of those hydropower plants will be able to achieve the planned energy production levels, and eventually even more because of possible increases in discharge. However, more detailed analysis would be necessary, as the achievable energy production will depend on the type of hydropower plant (run-of-the-river or dam with reservoir) in combination with the (changes in) intra- and interannual variability in streamflow.

Energy production in run-of-the-river hydropower plants - that are typically built at locations with high elevation differences but little streamflow - are vulnerable during low flows when minimum levels are required in order to divert the water and keep turbines working. Further, any streamflow above the maximum capacity of the turbines will not be used.

WP3 - The role of vegetation-climate non-linear feedbacks in the impacts of land-use and climate change

The objectives of this work packages were:
1. To quantify and understand the uncertainties in simulated changes in Amazon regional climate, vegetation and hydrology. Identify possible global impacts such as teleconnections and the relative importance of regional climate response to global warming vs. biophysical feedbacks
2. To assess the robustness of these coupled ESM results in the context of results from other models
3. To separate, quantify and understand the feedbacks in the Amazonian regional climate in ESMs arising from vegetation physiological responses to CO2 , possible biome shifts, anthropogenic deforestation, and climate change
4. To quantify and understand the impacts on regional and global climate arising from interactions between land use and climate change through fire activity and the resulting emissions of CO2 and aerosols and changes in the physical properties of the land surface
5. To quantify and understand the impacts of improved land use scenarios developed in WP4 for the Amazonian regional climate and global climate, as a result of changes in the physical properties of the land surface.
6. To identify key thresholds and indicators relating to critical damage to Amazon system
7. To estimate the likelihood of specific trajectories

Task 3.1. Analyse IPCC projections for Amazon region.
Deliverable D3.1 reports on the new simulations from the fifth phase of the Coupled Model Intercomparison Project (CMIP5) projections of climate change in the Amazon basin (e.g. fig A.7). The centennial simulations have been carried out according to different scenarios of greenhouse gas (GHG) concentrations, and include land use change consistent with development pathways and policy decisions. Thus, the implications of IPCC GHGs and land use according to the Representative Concentration Pathways (RCPs) on the changes in Amazonia can be explored in the CMIP5 multi-model ensemble. It presents an update to the last major phase (CMIP3) model projections of change that were reported in the IPCC AR4 (IPCC 2007).

Data from model simulations run under historical and RCP2.6 4.5 and 8.5 conditions were selected for analysis to span the uncertainties in greenhouse gas concentration and land use change that are encompassed by the RCP suite. Radiative forcing is greatest in RCP8.5 and lowest in RCP2.6 the latter of which is a mitigation scenario. RCP2.6 and RCP8.5 both have relatively large land use change in the global tropics, while in RCP4.5 there is a general global increase in forested area. Refer to the Special Issue of Climatic Change (2011, 109(1-2)) for a more complete description of the RCPs.

The CMIP5 models simulate reasonably well some aspects of the current climate of Amazonia and the wider region, such as the timing of the transitions in the seasonal cycle, and the mean temperatures in the region. Many of the models capture some characteristics of the important observed relationships between rainfall and sea surface temperature anomalies in the tropical Pacific and Atlantic. However, as a whole, the ensemble simulates conditions that are too dry in the Amazon basin throughout the year, and in many models substantially so. The known biases present in the CMIP5 ensemble for Amazonia should be taken into account in the interpretation of the projections of climate change, the development of Amazon ecosystem-relevant climate indicators, and in using model output to run offline impacts models.

The broad patterns of climate change projected by the CMIP5 ensemble are similar to those of CMIP3, and show that impacts increase under higher concentration scenarios. Temperature is projected to increase over the 21st century in Amazonia.

The changes in rainfall projected by the ensemble are mixed over the Amazon basin, and vary by season (fig A.7). However, there is generally more agreement on drying in the eastern basin, particularly in the June to November period, with wetter conditions projected by the majority of models in the western basin particularly in December to May. However, there is a spread in the model projections that spans zero, and over the Amazon basin itself, there is no clear scenario dependency apart from an increase in spread in RCP8.5 over 4.5 and 2.6.

Experiments complementary to the CMIP5 centennial RCP simulations were carried out in order to isolate the impacts of land use change from the other drivers of change. RCP land use change in Amazonia over the 21st century is small, and does not have a discernible effect on climate. A comparison between observation-based estimates of historical deforestation rates and those in the RCP with largest change (RCP2.6) reveals that the RCP rates are substantially lower. Furthermore, they are optimistic in comparison with some previously developed bottom-up scenarios. It is suggested the historical land cover is not sufficiently accurate at the regional scale and also that RCP land use scenarios are unlikely to realistically represent changes at regional scales. The land use changes implemented under the RCPs are found to be insufficient for investigating impacts of land use change in Amazonia in the future, and would benefit from improved region-specific scenarios. The new scenarios of land use change developed by INPE through AMAZALERT WP4 help to address this requirement.

Towards the assessment of likelihood of dieback (Deliverable D3.4) and whole-project outputs (link to Task 3.4 and forest resilience), further work was carried out by IPSL that used observations to constrain the IPCC CMIP5 projections of precipitation for the Amazon. Highlights from this work include the finding that the constrained projections of change show a lengthening of the dry season in southern Amazonia, leading to an expanded region where, using dry season length as an indicator of tropical biomes, the tropical forest may not be supported in the future. This work suggests that the ‘model democracy’ view of CMIP5 projections may underestimate regional impacts. The research has been submitted as a paper to Nature Climate Change (Juan P. Boisier, Philippe Ciais, Agnès Ducharne and Matthieu Guimberteau: ‘Projected strengthening of Amazonian dry season by constrained climate model simulations’).

Task 3.2. Understand and quantify coupled processes using ESMs.
Partner VU has been using a zero-dimensional model to test and isolate potentially important mechanisms for forest stability. Results show that the vegetation in this model is more sensitive to moisture stress than temperature stress, and that soil moisture is of prime importance in determining forest health. More detailed results are presented in AMAZALERT Deliverable D3.4 and in a forthcoming paper (Meesters et al, in prep).

Analysis of experiments aimed at understanding and quantifying the physiological versus the radiative forcing of CO2. This was done by IPSL for the subset of CMIP5 models for which appropriate experiments were available, and also in more depth by the Met Office for HadGEM2-ES. Results are presented in AMAZALERT Deliverable D3.2. Results from the set of CMIP5 models demonstrate the important effects of CO2 physiological forcing on carbon and hydrological cycles. In the absence of large-scale warming, the Amazon gains a globally-significant quantity of biomass through biochemical effects. However, the climate response to CO2 compromises the uptake capacity of the tropical forest. With regards to CO2 physiological effects on the water cycle, productivity and leaf area index (LAI) increase, but stomatal conductance decreases as water use efficiency increases. This latter effect dominates, and the overall impact is a reduction in evapotranspiration. This leads to some reductions in precipitation across the basin, amplifying the drying impacts induced by the large-scale climate perturbation (radiatively-driven) at the end of the dry season. Comparisons have been made between the simulation where both physiological and radiative effects of CO2 are operating (‘ALL’), and the sum of two parallel experiments that each has only one or the other (‘BGC+ RAD’). Differences between the two indicate the possible action of feedbacks or nonlinearities. In HadGEM2-ES, although evapotranspiration reductions are greatest in the eastern basin in the ‘ALL’ simulation, the difference with ‘BGC+RAD’ – and hence the potential for nonlinearities – is greater in the west.

FAN identified serious deficiencies in model representation of vegetation in Bolivia, which makes regional climate impacts assessments of limited use. To address this problem it adapted a dynamic vegetation model (LPJ-Guess) specifically for Bolivia and validated model outputs against biomass measurements and remote sensing data. This regionally adapted model was then used to assess the impacts of climate change by forcing the model with contrasting climate change projections from two statistically downscaled CMIP5 climate models. Changes in carbon stocks and fluxes were evaluated, factoring out the individual contributions of atmospheric CO2, temperature and precipitation. Special attention was paid to the effect of rising temperatures on photosynthesis, respiration, and the atmospheric demand for transpiration. This work led to two articles: Seiler, C., et al. (2014), Modelling forest dynamics along climate gradients in Bolivia, J. Geophys. Res. Biogeosci., 119, 758–775, doi:10.1002/2013JG002509; and Seiler et al., The sensitivity of wet and dry tropical forests to climate change in Bolivia, submitted to Geophys. Res. Biogeosci.

Task 3.3. Quantify and understand impacts of new land use (LU) scenarios; modelling fire processes.
LU scenarios from WP4 (detailed in Deliverable D4.2 Aguiar et al. 2014) were used in the implementation of experiments at WP3 project partners: Met Office, IPSL, INPE. At INPE, these scenarios were implemented in a land surface model in conjunction with a new fire model. The impacts of the pessimistic ‘fragmentation’ scenario (refer to D4.2 Aguiar et al., 2014, for details) was explored in the Earth System Models. This is a scenario of high deforestation, and provided a contrast with the low deforestation levels seen in the RCPs. Outside the Amazon basin, the standard RCP8.5 land use was used. Results were presented in Deliverables D3.3 and D3.4. Highlights of this research indicate that the additional deforestation has a drying effect in Amazonia in comparison with the standard RCP. Evapotranspiration, precipitation and runoff all decline relative to the global climate change-driven effects. We demonstrate the effect of the new LU scenario on the hydrological cycle (precipitation, P; evapotranspiration, ET; runoff/drainage, RD). In experiments run at the Met Office, the underlying state was found to affect the magnitude of the deforestation-induced climate change. Southern and eastern Amazonia was highlighted as having a greater response (reduction of ET) to forest loss than in other parts of the basin, particularly during the dry season.

A major accomplishment by INPE has been the implementation of fire modules into HadGEM2-ES and the land-surface component (IBIS-INLAND) of BESM (Figure A.8). For HadGEM2-ES, the fire module estimates burned area for each vegetation class. As yet, it does not represent feedbacks between fire occurrence and vegetation dynamics, so at this point it does not simulate plant mortality or changes in nutrients dynamics due to fires. For IBIS-INLAND, the implementation of the current version of the fire module allows the simulation of fire probability and effects on vegetation dynamics. Direct relationships between fire and deforestation have not been modelled, but deforestation processes have been included via an increase in the disturbance rates for the vegetation. The implementation of the fire modules into the models, and results of simulations including land use and fire are fully reported in AMAZALERT Deliverable D3.3.

Task 3.4. Stability and resilience analysis.
Analysis was performed of a large (56-member) ensemble of variants of the HadCM3C coupled climate-carbon cycle model. This was done by AMAZALERT researchers at the Met Office working in conjunction with a group from the University of Exeter. Results are presented in AMAZALERT Deliverable D3.4. The results of this work were important as simulations from fully coupled climate-carbon cycle models have been very few, and so there can be a risk of over-emphasis on sparse results. Furthermore, this is an ensemble of versions the same model that produced the original ‘dieback’ result. It was found that the original result of the Hadley model in 2000 was atypical in the context of this ensemble, although higher uncertainty in forest future was found for the highest RCP (8.5).

A risk framework methodology was used at the Met Office to capture information contained within extreme events and their effects on forest health, and to try to identify thresholds in regional meteorological and vegetation indicators. This methodology together with a worked example was presented in Deliverable D3.4. It was presented as a framework through which more or improved climate-vegetation indicators could be used as observations and process understanding increase, and improved indicators have subsequently been applied to it.

AMAZALERT researchers at the Met Office working in conjunction with a group from the University of Exeter carried out an ‘expert elicitation’ of the likelihood for climate-drive forest collapse. This provided a means by which several sources of information could be combined, and could take into account some of the uncertainties and missing processes from the models in a subjective manner. This approach provides an alternative view on likelihood and forms an update to a previous exercise carried out in 2006 (published in 2009). Presented in D3.4.

Detailed synthesis of different types of information, including a review of literature and methods along with new research carried out under AMAZALERT, was made by all WP3 partners, to assess the likelihood of irreversible collapse, presented in Deliverable 3.4.

VU has also conducted work on the use of fluctuations for early warnings, and a research highlight is that fluctuation analysis appears in practice to be unsuited for early warning systems. This work is being prepared for submission to the peer-reviewed literature.

A workshop on thresholds, tipping points and early warnings was held at the Met Office, UK, 6-7 June 2013. This brought together AMAZALERT researchers from Work Packages 3, 2 and 5, and experts external to the project from Exeter University and the Met Office. All workshop documents, including agenda, presentations and workshop report, are available here:

WP4 - Implications of climate and land-use scenarios for national and international policies

The aim of this work package has been formulated as follows:
To evaluate the effectiveness of a set of national and international policies by linking them to a set of land use, land cover, and water availability (LULCWA) scenarios. This enables an analysis of the robustness of policies over a range of future socio-economic and climate conditions.

Specific objectives:
• Develop a set of qualitative and quantitative LULCWA scenarios using participative methods.
• Select a set of sub-national to international policies relevant for land use change.
• Evaluate the implications of scenarios for policy and other instruments relevant to provision of Amazon ecosystem services, focusing both mitigation and adaptation strategies.
• Provide guidance for the development of the early warning blueprint

Task 4.1.1 Describe current and foreseeable key climate and policy instruments
Although developments in the Amazon Basin will be most directly impacted by national and sub-national policies and programs, a limited set of international policies and programs are also likely to influence land use and land cover. In particular, bioenergy programs in, e.g. Europe, the United States, and China and REDD+ initiatives are expected to play a role. Important demand for agricultural products other than for bioenergy and related certification standards can also be expected to be of importance. Examples are the high Chinese demand for soybean imports and standards and certification programs for agricultural and biofuel products. On the Brazilian level, the new Forest Code is of major importance, next to the Action Plan for Prevention and Control of the Legal Amazon Deforestation and land titling. Major international and non-Amazonian national as well as Brazilian policies and programs were briefly described. It is important to highlight that these different policies and programs may affect land use in the Amazon due to direct conversion of forest or other land use transitions that may occur in other parts of Brazil. For example, it is not expected that sugar cane plantations will expand in the Amazon, due to biophysical reasons but also because the land zoning for that product does not allow its plantation in the Amazon and the Pantanal Biome. However, if sugar cane plantations expand in southeastern Brazil, it is likely that cattle ranching will expand even more than it did already towards the Amazon borders. Thus, the connections of different forces and policies driving land use changes in the Amazon must be considered. Relevant policies were discussed in more detail in D4.1 and have been used in the subsequent discussions with stakeholders and development of scenarios.

Task 4.2.1 and 4.1.2 Scenario method development, and review and selection of relevant set
The work started with an evaluation of existing scenarios that could bear relevance for AMAZALERT. It was foreseen that a couple sets would emerge that could be used to kick-start the process of developing scenarios for Brazil and the Legal Amazon. Indeed, two sets of scenarios were identified that were highly relevant to the Brazilian context, the goals of the AMAZALERT project and the scenario process that was being envisioned. Those sets are:

1. The new IPCC-guided scenarios that are being developed for the Fifth Assessment Report (AR5). These scenarios consist of a set of four climate change scenarios (Representative Concentration Pathways – RCPs) and a set of 5 socio-economic scenarios (Shared Socio-economic Pathways – SSPs), and will be accompanied by a set of policy scenarios (Shared Policy Assumptions – SPAs). A crucial aspect is that these are scenarios that are very recent, have a specific component to include regional studies, and that are on the same topic as AMAZALERT. The release of the final version of the stories has been delayed, but an earlier draft of the full stories served as starting point for the work in AMAZALERT.
2. A set of scenarios (‘visions’) for Brazil, developed by an expert panel, composed of mainly invited researchers from INPE's Earth Science System Center (CCST, our INPE-AMAZALERT scientist are in this center). This is a set of initially 2 scenarios, which was later expanded to 4. The development of the qualitative stories was finalised during the course of 2012. A crucial aspect is that these are scenarios that are for the same region, and that are partly developed by the same researchers that are included in AMAZALERT.

Task 4.2.2 Qualitative scenarios
A range of participatory activities took place that aimed at qualitative scenario development. These included:
• Two workshops in Brazil (Belem and Brasilia; ws1 and ws2), during which two existing scenarios were discussed and detailed, both in terms of scenario components and related policies and other actions.
• One workshop in Brussels (ws3) during which two scenarios were developed and possible policies discussed, related to the role of Europe in reducing deforestation in the Amazon (fig A.9)
• A set of about 20 in-depth stakeholder interviews that broadened the stakeholder base represented in the two workshops, and likewise aimed at discussing future scenarios and policies.

The results and analysis of the three activities are documented in Deliverable 1.2. The different types of results span a large range of scales. In short, the recently completed IPCC-guided global Shared Socioeconomic Pathways (SSPs) were linked to scenarios for Brazil and the Brazilian Amazon. Those served as starting point both in the first workshop in Brazil (ws1) and the workshop in Europe (ws3). The land use scenarios resulting from ws1 served as input in a land use change model (LuccME) that in turn provided spatially explicit maps with patterns of deforestation in the Amazon (next section and WP2). These were an important input to the vegetation models. The Brazilian and European (land use) scenarios and model results were used to discuss actions, policies, and strategies needed to reduce deforestation in the Brazilian Amazon (ws2 and ws3). Through the use of multiple scenarios, we drafted a list of potential “no-regret” policies for the EU, importantly related to strengthening civic society and social cohesion; investment and import standards, and a better embedding of the EU in international agreements and policies. Also, there should be an increased involvement of the EU in strengthening environmental standards in trade on a global level would enhance the impact of other international countries and organisations that have a more important influence on the Amazonia.

Task 4.2.3 Quantifying scenarios
The two workshops in Brazil were directly aiming at developing qualitative land use scenarios. The output of the scenarios was used to develop quantitative land use scenarios that were used as an input to LuccME, a spatially explicit land use change model. The method and results are described in Deliverable 4.2.

A few observations related to these scenarios stand out:
• There is a large uncertainty on future land use change in the Amazon. As much as 50% of current forests might be deforested by 2100, but it could also be as little as 10-15%.
• In the most positive scenario, deforestation hot-spots are concentrated along the edges of the Amazon; in the most negative scenario the “heart of the Amazon” is also opened and partly deforested.
• Also in the most positive scenario, deforestation will continue, even if it is at a very slow rate.

Task 4.2.4 Consistent integrated scenarios

Consistency of qualitative and quantitative scenarios:
The overall Story-And-Simulation approach that was followed to link qualitative and quantitative scenarios includes a specific step of iterating between the stakeholder-generated qualitative stories and model-based quantitative scenarios. Iteration increases consistency between the assumptions of the stakeholders and of the modellers, thus leading to more consistent products. With this in mind, the two workshops were timed such that the land use change modelling results could be presented and discussed during the second workshop, and subsequently be used during the further process of that second workshop. Maps of model output (figure A.10) were available during the group discussions and played a role in further elaboration of the scenarios. Thus, the input of the model runs was based on stakeholder-generated stories, while the output was fed back into the workshop process. The resulting qualitative and quantitative scenarios can thus be considered highly consistent.

Consistency across scale:
The availability of interviews across the Brazilian Amazon and results from a European workshop allowed us to discuss the scalability of the scenarios that were produced. Some preliminary conclusions include:
• An utopian scenario resulted from all three stakeholder engagement activities. Although this was partly by design, there is a high degree of consistency between “Scenario A” (ws1,2), “Sustainability First” (ws3), and desirable futures as mentioned during interviews. These all also match the global starting point, SSP1.
• A dystopian scenario resulted likewise from all stakeholder engagement activities, again partly by design. There is a high degree of consistency between “Scenario C” (ws1,2), “Security First” (ws3), and the long-term future outlooks of some interviewed stakeholders.
• Similar components were identified. Importantly, the drivers of deforestation overlapped to a (very) large extent, as did proposed solutions (see Task 4.3).

Overall, it can be concluded that the design of developing nested scenarios based on existing global (SSPs) and Brazilian scenarios (CCST) ensured that resulting scenarios are highly consistent across scale, and highly consistent between qualitative and quantitative products. A more detailed analysis is needed to see to what extent the scenarios map onto each other.

Task 4.3. Implications for policy instruments.
The scenario development has value in its own right and particularly the quantification in terms of land use change and deforestation will contribute to the discussion where hot-spots of deforestation could be expected. Yet, the main value of scenarios in AMAZALERT was to contextualise the discussions on which policies and other actions might be most effective in combatting deforestation. Although methods of the three stakeholder-engagement activities to yield these results differed, all three activities produced results on implications for policy instruments. Results of the Brazilian and the EU activities are highly complementary as they aim at identifying priorities for action either within the Amazon region/Brazil on the one hand and by the EU on the other hand. An important link between the two is given when it comes to the EU (or other regions) to directly support efforts in the Amazon region. Note that because the majority of the activities were carried out in Brazil, there is a strong focus on national policies addressing the question “what can Brazil do to slow deforestation?”. Policy recommendations were published in a ‘summary for policy makers’, as cited in section 1 of this report and in the next part of this report (the ‘Impacts’ part).

Task 4.4. Input for the Early Warning System.
The input for the Early Warning System was presented in Deliverable 4.3.
There are, many potential future social and economic tipping elements in a list that is very preliminary and subject to change. It is close to impossible to assess tipping elements in terms of critical values. Much depends on (temporal, spatial, and thematic) system boundaries. It is also evident that many potential elements act over (very) long time scales and might be very difficult to reverse once critical values are exceeded. Despite many unknown and the preliminary character of this analysis, it is clear that social and economic aspects need to specifically be considered when designing an Early Warning System. Further details can be found under results for WP5.

WP5 - Integration towards a blueprint for an Early Warning System

The objectives for this work package are as follows:
• Integrate the understanding from previous work packages on how the Amazon system will behave when its critical services approach irreversible collapse.
• to determine the critical indicators of change in the Amazon region, in a spatially explicit way, and identify the relevant and required monitoring systems
• to develop the blueprint for an Early Warning System, based on observation, modelling and practical needs and possibilities

Task 5.1 Behaviour of the system near tipping points
NOTE: reference to ‘D5’ here means the joint deliverable D5.1-D5.3. The text from this deliverable is under development to generate a scientific publication on ‘The scientific basis of Early Warning for critical transitions in Amazonia’ (Kruijt et al, in prep) .
Integrate the likelihood of tipping points in the Amazon and the potential behaviour of the Amazon biophysical and societal system near or around critical transitions from previous work packages

Deliverables D2.5 D3.4 D4.2 and D4.3 are the most informative on this issue, and have been integrated into several key messages (see D6.7) and one summarising message (see fact sheet 7 in D6.3): “Most of the Amazon is not likely to degrade severely as a result of climate change this century, if deforestation is kept low. The south-east is more vulnerable. Uncertainties about the effects of CO2 and temperature increases, drought and policies are high and Amazon forests need to be monitored to ensure early prediction of degradation.“ It has lasted until the final stages of the project until we could formulate such summary. Uncertainty on the likelihood of tipping points has delayed our work to investigate the behaviour of the system near such tipping points, and forced us to change focus to more broad transition processes. See: D5, sections 2,3 and 8.1.

Confront this behaviour with the priority ecosystem services defined in WP1, and determine their relative importance

Deliverable D1.2 presents an investigation of the most important services (ES), while D3.4 shows some of the biophysical mechanisms behind these. In particular, the water cycle is regarded a key ES, linking to many other services (such as hydropower, illustrated in D2.4). D3.4 shows how important the forests are to maintain this water cycle. This is also illustrated in our key messages (D6.7). Together with the more obvious ES, Carbon storage, these two ES have from early on guided our analysis: what characterises breakdown of these ES and how can we monitor them. Suggested by one of the members of our Scientific Advisory Board, Prof. James Shuttleworth, we detected one promising technique for future monitoring of soil moisture. See: D5 sections 3.1,4,7.2 and 8.1.

Explore how early warning system elements might be incorporated into a market-based REDD+ mechanism and implications, if any, for design of such a mechanism.

Policy analysis during the first project half (D4.1) as well as further stakeholder interactions, show that REDD+ is at best a fairly disputed policy mechanism in Brazil, our main focus relating to policy. Discussion on policy focused much more on national legislation on forestry, land tenure and social development. This was also confirmed in the EU-stakeholder workshop (see D4.2). Therefore, REDD+ has not receive much attention in the project. Nevertheless any REDD+ project will have to deal with risks of climate-induced forest die-back, and as such implications of REDD would justify more intensive monitoring and warning systems. As such we analysed it briefly. See: D5 section 8.1.

Explore how early warning system elements might be used in conjunction with national agriculture, forestry, soil and water protection policies.

As described under the previous subtask, these kind of policies have attracted more attention in AMAZALERT (D4.1 D4.1) however, in this WP (D5) we have limited the links of an EWS to policy to more general principles. For general principles see D5, section 5 and for more practical principles see D5 section 8.1.

Task 5.2 Identify potential variables for detecting imminent transitions
Identify the critical quantities and variables from task 5.1 and formulate observational capability.

In D5, section 4, we describe in a systematic way, and linked to priority ES, the most important options for monitoring and potential future developments. Some exciting new possibilities include novel techniques to for high-resolution biomass mapping, large scale monitoring of vegetation and soil moisture using cosmic rays, and a comprehensive list of socio-economic indicators. See D5, section 4.

Review existing monitoring data sets and systems in the Amazon, and planned or potential extensions of this.

In D5, section 6, we investigate existing systems and future planning. From this, it becomes clear that several monitoring systems are already in place in Amazonia. However, the majority of these is focused on monitoring forest cover and biomass, with the notable exception of river discharge monitoring, which is well in place through the Brazilian national water agency ANA. Fire monitoring and warning systems are actively being developed, but usually with a more local and short-term focus than an EWS. There is a surprising scarcity of climatological monitoring. What becomes abundantly clear from this, and also from the interaction with stakeholders (D5 section 7), is that the power of a monitoring system will lie in integrating existing and new systems, and bringing them to a common interface. See D5, section 6 and 7.

Confront the requirements with existing and planned systems and identify priorities.

As stated with the previous, the real strength will lie in integrating many systems rather than prioritising, however, from AMAZALERT science it is clear that monitoring vegetation carbon and (atmospheric, vegetation and soil) moisture will be key to detecting degradation. The latter would therefore deserve priority interest, because the spare availability of data and monitoring to this respect. See D5, section 4 and 8.1.

Engage selected stakeholders to define priorities for monitoring in the Amazon from different perspectives, looking into the future decennia.

Stakeholders were involved in various ways. First of all, a discussion was set up during the second Brazilian workshop in Brasilia, November 2013. Following that event, selected stakeholders were sent a questionnaire. Results of this questionnaire can be found in D5, section 7 and accompanying tables. In addition, a visit to the Brazilian ministry of environment, yielded fruitful and encouraging feedback and a broader group of stakeholders was consulted at the final project meeting. This resulted in 1) a strengthening of our vision into the main focus needed and 2) forced us to change focus on potential set-up and institutional embedding of an EWS (see task 5.3-2). See D5, section 7.

Task 5.3 Formulate blueprint for an early warning system
Bring together monitoring needs from practical perspectives with the monitoring needs from the perspective of early warning for irreversible changes, write an argued case for a network of monitoring sites/instruments, including initial cost assessment.

This subtask was carried out towards the end of the project, when everything came together. The ‘argued case’ is describe in D5, section 8.1 and includes the accounting for different approaches to early warning depending on the nature of the transitions expected and priority monitoring. We arrive at a general sub-regional distinction of priorities, distinguishing a NE, NW, SE and SW sector that vary depending on the sub-regional risks, uncertainties and current state of rain forests and society (figure A.11). Because of the growing insight that an EWS should not depend on newly planned systems but should rather bring together existing initiatives and new initiatives from third parties, it was deemed not very informative and too complex to formulate an ‘initial cost assessment’. See: D5, section 8.1.

Write a plan for institutional management and infrastructure to support such a network.

Stakeholders were clear, and especially among Brazilian colleagues the realisation also grew, that an EWS as envisaged, should not reside at one single institution or service. Rather it should be a community effort, co-ordinated by a small group of scientists. This is described in D5, section 8.2 including an assessment of personnel and materials needed for such a set-up.

Formulate tools and guidelines for interpretation of data in such a network, especially how to identify imminent irreversible transitions.

Tools and guidelines for analysis should not, as originally envisaged, be focused mainly at statistical methods to detect early warning signals from variability and ‘critical slowing down’ in indicator values. Thus a broader, more general system is needed, based on modelling and observations, trend analysis and projections. This is described in D5, sections 1-3.

Formulate initial guidance on appropriate action to be taken if and when transitions are apparent from the system.

Communication, refined monitoring and policy actions (mitigation, adaptation) are also discussed in the plan. More general principles linking an EWS to action are presented as well. See: D5, section 5 and 8.2.

Evaluate the plan on practical feasibility together with the selected stakeholders.
The final stakeholder meeting also entailed an evaluative aspect. Comments were variable. Most stakeholders strongly encourage the notion of an EWS for critical degradation in Amazonia. Several pointed out the existence of several systems. Some were critical about the plans and progress presented at that moment. But the underlying and overall opinion was, as stated above, that there should not be one institution running such an EWS, and that too little attention had been paid to existing systems.


Aguiar, A., Tejada, G., Assis, T and Dalla-Nora, 2014. AMAZALERT, Deliverable 4.2. Set of land-use scenarios for Brazil, linked to implications for policies.

Boisier, J.P. Ciais, P., Ducharne, A. and Guimberteau, M. Projected strengthening of Amazonian dry season by constrained climate model simulations. Submitted to Nature Climate Change.

Christoffersen B.O. Restrepo-Coupe, N., Altah Arain, M., Baker, I.T. Cestaro, B.P. Ciais, P., Fisher, J., Galbraith, D., Guan, D., Gulden, L., van den Hurk, B., Ichii, K., Imbuzeiro, H., Jani, A., Levine, N., Miguez-Macho, G., Pulter, B., Roberti, D., Sakaguchi, K., Sahoo, A., Schaefer, K., Shi, M., Verbeeck, H., Yang, Z., Araujo, A., Kruijt, B., Manzi, A., da Rocha, H., von Randow, C., Muza, M.N. and Borak, J., 2014. Mechanisms of water supply and vegetation demand govern the seasonality and magnitude of evapotranspiration in Amazonia and Cerrado. Doi: 10.1016/j.agrformet.2014.02.008.

Gonçalves, L.G.G. Borak, J.S. Costa, M.H. Saleska, S.R. Baker, I., Restrepo-Coupe, N. and Muza, M.N. 2013. Overview of the Large-Scale Biosphere-Atmosphere Experiment in Amazônia Data Model Intercomparison Project (LBA-DMIP). Agricultural and Forest Meteorology, in press.

Guimberteau et al. 2014. Influence of rainfall rate on the surface hydrology and vegetation of Mato Grosso, Amazonia. Submitted to Environ. Res. Lett.

Joetzjer, E., C. Delire, H. Douville, P. Ciais, B. Decharme, R. Fisher,B. Christoffersen, J.C. Calvet, A.C.L. da Costa, L.V. Ferreira and P. Meir, 2014. Predicting the response of the Amazon rainforest to persistent drought conditions under current and future climates: a major challenge for global land surface models. Geoscientific model development, 7, 2933-2950. Doi: 10.5194/gmdd-7-5295-2014.

Kruijt, B., Jans, W.W.P. Vasconcelos, S., Almeida Melém, E., Tribuzy, E.S. Felsembourgh, C., Rowland, L., Lola da Costa, A., Meir, P., Chambers, J. and Camargo, P., 2015. Temperature response of leaf photosynthesis of Amazon forest trees and its acclimation to mean temperature change. In Preparation.

Marengo, J., Chou, S. C., Kay, G., Alves, L.M. Pesquero, J.F. Soares, W.R. Santos, D.C. Lyra, A.A. Sueiro, G., Betts, R., Chagas, D.J. Gomes, J.L. Bustamante, J.F. Tavares, P., 2011. Development of regional future climate change scenarios in South America using the Eta CPTEC/HadCM3 climate change projections: climatology and regional analyses for the Amazon, São Francisco and the Paraná River basins. Climate Dynamics, 38:1829–1848. DOI 10.1007/s00382-011-1155-5.

Poveda, G., Vélez, J.I. Mesa, O.J. Cuartas, A., Barco, J., Mantilla, R.I. Mejía, J.F. Hoyos, C.D. Ramírez, J.M. Ceballos, L.I. Zuluaga, M.D. Arias, P.A. Botero, B.A. Montoya, M.I. Giraldo, J.D. and Quevedo, D.E. 2007. Linking long-term water balances and statistical scaling to estimate river flows along the drainage network of Colombia. Journal of Hydrologic Engineering, 12(1), 4-13. DOI: 10.1061/_ASCE_1084-0699_2007_12:1_4_

Rodriguez, D. A., 2011. Impactos dos padrões espaciais da vegetação nas variáveis atmosférica e terrestre do ciclo hidrológico, em bacia de floresta amazônica. 208 f. Tese (Doutorado em Meteorologia), Instituto Nacional de Pesquisas Espaciais, São José dos Campos, 2011. available on:

Rodriguez, D.A. Tomasella, J. and inhares, C., 2010. Is the forest conversion to pasture affecting the hydrological response of Amazonian catchments? Signals in the Ji-Paraná Basin. Hydrological Process. 24, 1254–1269 . DOI: 10.1002/hyp.7586.

Rost, S., D. Gerten, A. Bondeau, W. Luncht, J. Rohwer, and S. Schaphoff, 2008., Agricultural green and blue water consumption and its influence on the global water system, Water Resour. Res., 44, W09405, doi:10.1029/2007WR006331.

Rowland L., Malhi, Y., Silva-Espejo, J.E. Farfan-Amezquita, F., Halladay, K., Doughty, C.E. Meir, P., Phillips, O.L. et al., 2014. The sensitivity of wood production to seasonal and interannual variations in climate in a lowland Amazonian rainforest. Oecologia, 2014(174), 295-306. Doi: 10.1007/s00442-013-2766-9.

Rowland L., Charles Hill, T., Stahl, C., Siebicke, L., Burban, B., Zaragoza-Castells, J., Ponton, S., Bonal, D., Meir, P. and Williams, M., 2014. Evidence for strong seasonality in the carbon storage and carbon use efficiency of an Amazonian forest. Global Change Biology, 2014(20), 979-991. Doi: 10.1111/gcb.12375.

Sheffield, J., G. Goteti, and E. F. Wood, 2006. Development of a 50-yr high-resolution global dataset of meteorological forcings for land surface modeling, J. Climate, 19 (13), 3088-3111

Seiler, C., Hutjes, R.W.A. Kruijt, B., Quispe, J., Añez, S., Arora, V.K. Melton, J.R. Hickler, T. and Kabat, P., 2014. Modelling forest dynamics along climate gradients in Bolivia, J. Geophys. Res. Biogeosci., 119, 758–775. doi:10.1002/2013JG002509.

Seiler et al., The sensitivity of wet and dry tropical forests to climate change in Bolivia. Geophys. Res. Biogeosci. In review.

Tomasella, J., and Rodriguez, D.A. A numerical approach for modelling sub-grid variability in tropical catchment. Hydrological Processes - 2014 - Under Review.

Von Randow, C., Biemans, H., Guimbertau, M., langerwisch, F., Garofolo Lopes, L, Poverda, G., Rammig, A, de Cassia S. von Randopw, R., Andres Rodrigues, D., Samprogna Mohor, G., Lazaro Siquera , J., Thonicke, K., Tomasella, J. and Verbeeck, H., 2014, Report on the likely response of the Amazon basin hydrology and river discharge to land use and climate change. AMAZALERT deliverable D2.4

Potential Impact:


Amazonia is under threat through the combined effects of unsustainable regional development and climate change. As summarised in the IPCC reports of 2007 and 2014 studies in the past ten years have indicated that these changes can lead to deforestation, regional disturbance of temperatures and the water cycle, as well as loss of carbon stocks and biodiversity. In turn, these changes can lead to forest loss, droughts, low river levels, floods, loss of hydropower energy and many other ecosystem services and even enhanced risk of diseases and loss of agricultural productivity.
We are only beginning to understand and quantify these threats individually. For example we know that forests are bio diverse, carbon rich and sensitive to climate change and climate extremes, that fire incidence rises substantially in the presence of drought and land use, and that forest loss might affect regional-scale evaporation rates, potentially affecting water supply within Amazonia and beyond its limits, to the bread-basket of South America, the La Plata Basin. We also know that the changing variability of climate can lead to floods as well as droughts in the region, affecting hydropower and agricultural productivity. However, what has been missing until recently has been the capability to combine these different areas of knowledge within one or a few modelling frameworks to understand in an integrated way how the system might respond to coupled global change drivers, and to focus new measurements and new knowledge-gathering processes on our key gaps in biophysical and socio-economic understanding. By combining and developing such model frameworks and data-providing networks, we can develop capacity to predict and identify the possible circumstances leading to loss of Amazonian ecosystem service provision, thereby creating the basis for an ‘early warning system’ that increases societal resilience and improves our capability to avoid large scale or rapid decline of one of the world’s key biomes, the forests of Amazonia.
The AMAZALERT project (2011-2014, a consortium of 14 institutes in Europe and south-America, funded by EU-FP7 and national funds) has been pioneering the multi-level integrated analysis needed to address the environmental and societal risks and aspirations described above.


The research addressed one of the main global ‘hot spots’ of climate change impacts. Public and policy attention for possible ‘die-back’ of the region has been strong since reports in the early 2000s that climate change might lead to ‘Savannisation’ of the amazon. the fourth IPPC assessment report has highlighted this same risk, where the fifth IPPC report reported both on the wide range of uncertainty and on the already on-going degradation in parts of the region.


Although of course resilience and vulnerability are complimentary, and although many of our results do indicate important risk for Amazonia, it makes a difference whether emphasis is on the value and strengths of the remaining forest rather than focus on how vulnerable the region is. Our results suggest that in several ways, the region is not as vulnerable as previously assumed because the mere presence of forests maintains the climatic conditions for forest health, which highlights the importance to conserve and protect the forests. We expressed this through the message that ‘without further increase of deforestation, dieback of Amazonian forests caused by climate change alone is not very likely’. We also explored, through our scenario development workshops, the necessary policy conditions to realise this, and accounted for already successful policy which lead to recent reclines in deforestation. Ensuing recommendations are listed below. Finally, we quantified the importance of Amazonian rivers for hydropower, as well as the likely change of this potential with a changing climate.


One of the main products of AMAZALERT is the analysis and ‘blue print’ for an early warning system. We focused on the potential to early observation of change, based on what our model results told us about the likelihood of such change. For this we studied necessary analysis methods as well as existing and required new monitoring systems.


Even though uncertainties are still high (and recommendations to improve on these were developed, see below), we formulated answers to the main question whether die-back is likely or not.


We collaborated with several other research initiatives on climate, land-use and Amazonia. For example, coupled model analysis made use of the extensive CMIP5 model exercise, our dynamic vegetation models were run on the basis of forcing and scenarios developed by the Moore foundation’s Amazon-Andes initiative and earlier work by the LBA-MIP (Model Intercomparison Project). This collaboration was also expressed through the joint organisation of a session at the December 2014 Fall meeting of the American Geophysical Union. Stakeholder interactions and scenario development was organised in collaboration with and on basis of work carried out under Brazilian GEOMA. Our final meeting was co-organised with the EU-FP7 ROBIN project, which has a wider geographical (i.e. Latin-American) scope and which studies links between climate mitigation and biodiversity. A joint stakeholder and science meeting was held in Belem and Alter-Do-Chao, Para, Brazil, while internal project meetings were held side-by-side, fostering new contacts and potential collaboration in both Europe and Latin-America.


AMAZALERT has boosted the development of both land-use change modelling (the LUCCme framework) and the development of coupled climate-vegetation-fire-land-use models (the BESM - Brazilian Earth System Model- framework). Both are important assets for the Brazilian science community to advance its position in global research on global change impacts and national climate adaptation. Furthermore, through the organisation of workshops and small training sessions, we fostered the science community to advance interdisciplinary work. Importantly, a big new initiative, AMAZON – FACE, to set up a full elevated CO2 experiment north of Manaus was supported and promoted. This addresses the prime uncertainty in climate change impact on Amazon forests. One of the triggers to set this up came from AMAZALERT and several of its scientists (e.g. Kruijt, Nobre, Meir, von Randow).


The results by AMAZALERT have generated a number of recommendations in the fields of both policy and science. These are:

AMAZALERT policy recommendations
1. Continued protection of standing primary forests is essential. It is important to maintain the buffering capacity of forests against the impact of climate change on regional water cycles, the carbon cycle, and to conserve the many ecosystem services that depend on the integrity of Amazon forests.
2. The effectiveness of policies needs to be enhanced in multiple ways. By improving spatial planning using coherent conservation strategies, formalising land tenure and ensuring effective law enforcement, that overall aims to maximise the ecosystem functions of Amazon forests such as water recycling, carbon storage and as a store of biodiversity.
3. Long-term monitoring systems need to be sustained and improved. Due to the uncertainty of the future trajectory of forest degradation, sustained long-term monitoring of the environmental conditions in, and state of, the forest is of crucial importance to detect the possible onset of its instability in a timely manner.
4. The EU should support regional initiatives. The most effective policy action available to the EU is to support efforts in Amazonia to towards a sustainable future, including actions to reduce deforestation because of their high likelihood of success and good fit within the national context.
5. European investments and laws have an impact on Amazonia. Environmental and social impacts on Amazonia should be considered carefully in European investments and corresponding legal frameworks.
6. Certification should be stimulated. Enhancing demand for products from the region that meet high environmental standards may reduce direct negative impacts and will generate momentum towards an overall increase in environmental standards.
7. The EU can play an important role in the international context. The EU is most effective in their involvement in strengthening environmental standards in trade on a global level would enhance the impact of other countries and organisations that have a more important influence on Amazonia.

AMAZALERT research recommendations
1. Investigating ecosystem functioning is still crucial. Future research on understanding the risks of Amazon forest degradation should focus on the dependence of forest stability on the balance and interactions of CO2 fertilisation, temperature increase, drought and fire dynamics.
2. Extensive monitoring of the state of Amazonia should be stimulated. Furthermore, basin-wide research should emphasise the understanding of regional variability in vulnerability; building of comprehensive time series and databases on the condition of the forests, regional climate and socio-economic indicators; and identifying actions that maximise the resilience to climate change of the ecosystem services provided by forests.
3. Scientific results need to be shared with policy in an efficient way. Research is needed on improving the science-policy interface and the uptake of state-of-the-art science in policy making. Novel methods are needed to involve decision makers.
4. Understanding of land-use change needs to be refined. The complexity of interactions between multiple drivers acting at multiple scales needs to be studied so as to better understand and model the dynamics of land-use change.

Project summary and recommendations have been published as a 'summary to policy makers' as well as in summary fact sheets, see:

In this section we summarise some of the more detailed impact of AMAZALERT through dissemination

Task 6.1. Produce printed material for distribution.
We produced an initial project brochure in English, later translated into Portuguese and Spanish. The brochure draft was open for discussion to all project partners. The brochure provides an overview of the project, including its motivation; its goals, the scientific questions and policy needs to be addressed; and how it will proceed, including involvement of stakeholders. This initial brochure was established in web and print electronic versions that were sent to all partners and posted on the public website.

In spring 2012, a revised brochure was again established in web and print electronic versions and posted on the website. The brochure text in all three languages served to update the website pages. A number of hard copies was sent to all partners in late summer 2012. Additionally, each partner may duplicate the brochure according to its needs, using the electronic print version. The brochure will be distributed at general meetings, at workshops and to stakeholders.

Task 6.2. Establish and maintain public web site.
The website was established in Autumn 2011, and has been continuously adapted and filled with updated information since then.

The website provides an English, Portuguese and Spanish version. The menu of the website is structured in (i) a project specific section, presenting the idea and aims of the project, work packages and participants, (ii) a publications section, where all folders, fact sheets, articles, presentations, event reports and main deliverables that have been produced by the project partners can be found. The results of this project are provided in a separated menu section called Results, where the project results, policy and research recommendations can be found. If not fully available at the webpage, links to all scientific publications are provided. (iii) A news section joins media work done (press releases, newspaper articles) and media echo on AMAZALERT (print media, broadcast) and includes subsections for the newsletters and archive of articles. A specific (iv) link section provides links to other sources, networks, scientific projects on similar topics. These links have been integrated continuously as they emerged. At the end of the project, a specific section was created (v) to present the outputs of the important final project and stakeholder meeting in Brazil. Finally, the website offers also an Intranet with share point function for project partners.

Also dissemination paths via social media have been developed, i.e. a facebook page (30 members) and a twitter account (75 followers, still growing in late 2014).

Task 6.3. Work with public media in Amazonia nations to publicize project.
From the beginning of the project, actions to disseminate the activities of the project in Amazonia were taken as follows:

• Press releases from INPE highlighting the start of the project were sent to major newspapers, magazines, TV and radio stations. At least 4 spots in major news sites appeared about the activities of the project, and these appearances were mirrored in many science blogs
• INPE's researcher Dr Celso von Randow gave two interviews, one for a news edition on TV and one for a local radio station. The TV spot can be watched at:

Later, eight more contributions on AMAZALERT have been prepared by partners and presented by TV/radio or print media, seven of them in Brazil media. A professional journalist was integrated into the EMBRAPA team. Especially at the kick-off and at the final meeting media attention was high. The AMAZALERT website provides direct links to these contributions

As another example of successful achievement of this task, an important TV report was presented in Brazilian biggest national TV station, with AMAZALERT researchers. The TV spot highlighted the importance of possible effects of deforestation in Amazonia in degrading ecosystem services, and the video can be seen at:

As a consequence of available data volume, analysis and publication of interim results increased in the second reporting period. Some of the numerous publications and presentations are made accessible via the project website and

Seven AMAZALERT Fact Sheets, in three languages, on key themes of research were issued by project partners, all of them during the second reporting period. In a condensed yet attractive format, recent scientific findings were presented, discussed and made available to interested scientific and non-scientific project stakeholders via the project webpage and via direct distribution.

The fact sheets were dedicated to the following themes:
Fact sheet1: Land‐use change in the Brazilian Amazon: products, policies and initiatives
Fact sheet2: Baseline model runs performed with 4 Dynamic Global Vegetation Models (DGVMs)
Fact sheet3: New data on the temperature response of photosynthesis in the Amazon forests: first results
Fact sheet4: New multi‐model projections of climate in Amazonia
Fact sheet5: Earth System Modelling
Fact sheet6: Amazonian ecosystem functions and services and their drivers of change
Fact sheet7: Impacts of Climate and Land-Use on Tropical Forests in Amazonia: Summary of Amazalert results

Task 6.5. Offer training sessions on project methods.
A total of five training sessions have been organised, aimed at different audiences:

•International workshop on environmental modelling in amazonia/ Side event,Manaus, BR,November 2013, 3 days aimed at Latin-American young scientists. Contents: Keynote talks and invited student talks, workshop on paper writing (Prof. J. Grace) and on hydrology modelling (A. Nobre)
•Tipping points analysis workshop at Project Midterm Meeting, Wageningen, NL, March 2013, 1/2 day, aimed at Wageningen Students and AMAZALERT scientists. Guest lectures by specialists on tipping points (Ben Booth (Exeter) and Valerie Livina (University of East Anglia) Contents: training on tools and approaches to detect tipping points in data sets
•Fifth INLAND Workshop (Integrated Model of Land Surface Processes) Cachoeira Paulista, SP, BR 27 - 28 November 2012, 2 days, aimed at Latin-American young scientsis. Contents:Training session on the use of the Brazilian INLAND surface model and aspects of biosphere modeling in Amazonia. Further info(in portuguese):
•Caxiuana workshop on Photosynthesis, Belem, PA, Br and Caxiuana field station, PA,BR, September 2012, 1 day, Students and scientists from Para, Brazil. Contents:Training on photosynthesis and response curves, on-site training on ecophysiological instrument usage
•Seventeenth Brazilian Congress of Meteorology / side event, Gramado, BR, September 2012, 1/2 day,Latin-American young scientists. Contents: Round table about Biosphere-Atmosphere Interactions, with special emphasis on training of the use of the INLAND surface model

Task 6.6 Promote implementation and further development of project results

A final stakeholder discussion event was organised and held in Brazil (Belem) on 6th October 2014, followed by a joint research meeting with the Robin project on 7th and 8th October and the final AMAZALERT project meeting from 9th-10th October 2014 in Alter-do Chao. The first day was dedicated to informing stakeholders and policy makers of the major results and products of AMAZALERT, to receive feedback, discuss results and the way ahead. The second day focused on the exchange on detailed scientific results in the wider scientific context. Many presentations and posters served to provide this information to stakeholders. A field trip on day three demonstrated the importance of local business and knowledge in the Amazon. Day four and five (AMAZALERT Project meeting) concentrated on final discussions at the Project level.

According to the general title of the meetings “Impacts of climate and land-use on rainforest biomes and their services in Latin America”, Latin American policy makers, civil society, administration, business and international scientists discussed AMAZALERT Project results, their application and their potential implementation in policy making. Besides AMAZALERT research results the Early Warning System was discussed in terms of its design, role and compatibility with existing institutions and monitoring systems.

As the AMAZALERT Project meeting was combined with a ROBIN EU-Project meeting, the event benefitted from the presence and knowledge of an impressive number of internationally renowned scientists. A high number of Brazilian stakeholders participated in the AMAZALERT stakeholder workshop.

A summary of major project results was prepared and distributed among participants and via the website as a basis for these meetings. The summary was updated after the event, based on discussions and work undertaken during the meetings.

The combined event benefitted from a remarkable media attention that is documented on the AMAZALERT website All documents related to this combined event may be found on the Project website

As all results obtained so far had indicated that the influence of Brazilian policy making on Amazon deforestation was dominant compared to the EU influence, it was decided to give priority to organise one sound and large event in Brazil, attracting key stakeholders.

Instead of a single EU event, a series of ample discussions with stakeholder groups on interim findings and their relevance for policy making took place during the reporting period. EU stakeholders representing policy makers, science, NGOs and businesses provided their inputs on the occasion of the European Workshop on the future of the Amazon in Brussels 2013 as well as EU science stakeholders did at the Brazil meeting 2014 (see above).
EU stakeholders’ inputs have been integrated into final Project documents, such as the Summary for policy makers or Fact Sheet No.7. This output has been also discussed with EC project officer during its elaboration.

SUMMARIES FOR POLICY MAKERS, specified for EU and Latin America policy makers have been elaborated on the basis of project key findings of AMAZALERT. Policy recommendations, based on these key findings, have been formulated. As for an approach to an Early warning system, discussions stated that detecting alarming trends would have to be approached from a broad perspective, integrating basin-wide monitoring of climate change and weather extremes, moisture indicators, biomass and carbon exchange, combining new and existing networks, and defining thresholds that account for society’s coping capacity as well as with the uncertainty in prediction of degradation.

List of Websites:

ALTERRA, Bart Kruijt, P.O.Box 47, 6700 AA Wageningen, Netherlands
INPE/CCST, Sao José dos Campos, SP, Brazil