Geospatial services for security are traditionally focused on analysing isolated scenarios with potential risks for citizens’ safety. Geospatial Intelligence (GEOINT) applications as the ones for Critical Infrastructure Monitoring, Disaster Monitoring or Humanitarian Aid are widely adopted by security stakeholders, and the use of EO and collateral data is already operationalized in the associated decision chains.

In a more global context, security is an intricate subject and potential scenarios can be triggered by causes of different nature. The links between events happening in the domain of security and other domains (e.g. climate, hazards, health) are highlighted in the most relevant global policies (e.g. Sustainable Development Agenda, Sendai Framework, Paris Agreement, EU Green Deal) as well as in the work programmes of several key entities such as the Group on Earth Observations (GEO). Thus, the change from the traditional paradigm of security as isolated domain to a broader new concept of security is ongoing. One of the most relevant examples of this new security paradigm is the so called Climate Security, which refers to how climate change related events amplify existing risks in society, endangering the safety of citizens, key infrastructures, economies or ecosystems.

To make the right decisions, undertake the appropriate measures and work towards a sustainable future on Earth, it is essential to understand the link between climate and other domains, as well as its consequences, and there is where EO data can provide high-value information. Several initiatives carried out within R&I projects or cooperative frameworks by the European Union Satellite Centre (SatCen) Research, Technology Development and Innovation (RTDI) Unit are hereafter described as examples of Geospatial Intelligence applications to address Climate Security issues.

- Flood events are increasing worldwide, having an impact on the safety of citizen and related activities. In the frame of the H2020 E-SHAPE project [1], the FRIEND pilot of the Disaster Showcase on Flood Risk Assessment is being implemented. The aim is to develop services based on Sentinel-1 / Sentinel-2 data and other ancillary data (e.g. climate and statistical data) to generate time-series and automatically detect changes. The output will provide both citizens and experts with a Flood Risk & Impact Assessment tool based on indicators, time-series charts and forecast maps.

- Climate related events are having an impact on population displacement and conflicts. In the frame of the H2020 GEM project [2], a Conflict Pre-Warning Map (CPW) is being developed with a dashboard considering: 1) meteorological data; 2) statistical data on demographics, socio-economic variables and political stability; and 3) relevant ancillary datasets on conflicts and migration. The outcome will show in a single entry point the correlation between these different datasets.

- The increasing scarcity of fresh water can lead to security issues. In the frame of the GEO SPACE-SECURITY Community Activity (SSCA) [3], a pilot is being developed addressing the vulnerability of an identified region and the effects on the safety and security of population (e.g. water access and crop yield, conflict and migration). The pilot is based on Sentinel-1 interferometric products to address underground water extraction. The output will consist in maps describing the vulnerability of such areas and the risks for population and infrastructures.


[1] https://e-shape.eu/
[2] https://www.globalearthmonitor.eu/
[3] https://earthobservations.org/

Session E2.02 Climate Security: The key role of Research and Innovation, Earth Observation and co-operation to address global threats


Climate change as a direct and indirect multiplier of international crisis and conflict, and the role of Earth observation

Prof. S.A. Briggs & Prof. P. N. Cornish

At all levels – individual, local, national and international – the consequences of climate change for human security are often presented in the language of ‘risk’ and ‘threat’, framed in a mindset that is traditionally state-centric in character and instinctively defensive, reactive and military in response. As Paul Collier has noted, politics is essentially spatial and territorial – the politics of the consequences of climate change are unlikely to contradict Collier’s definition entirely.

Rather than imagine a different politics altogether, one that might allow a more perfect response to the challenges posed by climate change, here we make a grounded and pragmatic argument for doing more, and better, with the limited means already available.

Using terms derived from military strategy and operations, the consequences of climate change are often described as ‘crisis multipliers’ or ‘threat multipliers’. Thus, when climate change results in desertification and famine, conditions can be created in which known violent extremist groups, such as Boko Haram, might flourish. Similarly, the retreat of the Arctic ice cap, enabling competition for oil and gas resources and for control of navigable sea routes, might result in increased tension between NATO members on the one hand and their competitors in Russia and China on the other. These two examples suggest that the consequences of climate change can have an indirect multiplying effect on existing tensions over territory and resources.

More direct effects should also be considered, particularly when it comes to access to fresh water. Human security in northeast Syria is being affected by declining water flow in the Euphrates from Turkey, and the previous drought in Eastern Syria (Voss et al, 2013)) has already been strongly argued to have exacerbated if not directly caused increased tension in population Centres in Syria, leading to further unrest, civil war and the migration of over 1.5 million refugees to Europe. There is hence a credible link between Syrian drought and subsequent discontent in Europe, leading to the rise of authoritarian leaderships in several countries and to the conditions and sentiments that propelled the UK via a narrow majority into Brexit. The possibility of conflict between Egypt and Ethiopia over the effects of the Grand Ethiopian Renaissance Dam is openly discussed; with Egypt’s population growing by c.1.5 million per annum, by some accounts Egypt’s supply of fresh water could fall below the ‘absolute scarcity’ limit of 500m3 by 2025. Rising sea levels threaten the lives and livelihoods of c.150m people around the world and might result in resource-competitive population displacements. It is sometimes said that these and similar scenarios could have a direct multiplying effect on crisis and conflict in the context of Collier’s sparse description of politics as spatial and territorial.

Furthermore, the Brookings Institution listing of failed states shows that sub-Saharan Africa across the continent from East to West is populated almost exclusively by countries with the least robust political systems. We have there a confluence of potentially the most severe impacts of food insecurity coupled with an absence of political resilience. This could lead to migration of one or two orders of magnitude greater than that seen from Syria into Europe in the decade after 2010.

We do not argue that the state-centric and military mindsets should (or could) be discarded altogether. Neither are likely to disappear in the near future, and both can contribute to the management of the consequences of climate change. Instead, we argue for a more informed and effective response from politics, both national and international. We argue that, within the means available, the most effective response to crisis and threat multipliers is to ‘multiply’ our response to these challenges. A strategic response to the challenges of climate change would see the adoption of cross-governmental approaches in which the security and defence branches are but one component, working with others including development, health, transport, policing etc. A new area of policy might also be developed, known as ‘resource diplomacy’ – a hybrid of foreign, economic, development and security policies. The next step would be to internationalise this cross-governmental, integrated approach to ensure that best use is made of the inputs available, including EO data, information and analysis, and to ensure that the response is more anticipatory than reactive.

The UK could provide a test case for the super-informed, agency-multiplying approach of the sort we envisage. The UK Government’s Integrated Review published in March 2021 speaks of a need for ‘global resilience’ and, where climate change is concerned specifically, ‘to increase our collective resilience to the damage that has already been done, in particular supporting the most vulnerable worldwide in adapting to climate effects and nature loss.’ There is hence an increasing recognition among governments for the need for a more integrated approach to societal security and resilience capitalising on the wealth of geospatial information now becoming available through satellite data sources.

The European Union (EU) strategy on adaptation to climate change recently adopted by the European Commission (EC) addresses the importance of improved adaptation and resilience to climate change. The impact of climate change is already visible on economies, communities and ecosystems worldwide with consequences on citizens’ wellbeing as well as the natural and built environment. In this context, the EU Green Deal sets achieving climate neutrality by 2050 as key goal. Thus, the need for climate change adaptation measures and processes has become urgent and necessary. In particular, a sustainable and environmentally-friendly economy requires changes in decision making processes, practices and systems to enable the prevention of hazards and natural disasters, as well as to limit potential damages. To this aim, the EC has developed a roadmap which is based on steps such as: i) the development of national adaptation strategies; ii) the mobilisation of regions towards the identification of adaptation measures and their deployment; and iii) the identification of financing and economic models to enlarge the potential solutions. The roadmap emphasizes the need to reduce the gap between what can be achieved using proven adaptation solutions, and what is needed to achieve a rapid and far-reaching change. This is indeed a challenging issue, as it entails the implementation of transformative processes encompassing societal, environmental, administrative and financial policies and actions. Τo achieve successful climate adaptation and mitigation, alignment between climate adaptation measures and economic recovery is crucial and for that societal consensus is sine qua non. Climate adaptation and mitigation rely on innovative physical solutions, as well as deep societal changes that can only be achieved by increasing social participation in decisions related to climate adaptation and the creation of innovative region-specific regulatory, governance and bio-physical strategies.

Remote sensing can play a key role in this context. Multimodal remote sensing in particular, shows significant potential, leveraging an ever increasing diversification of available datasets in ever increasing spatial, spectral, temporal, and radiometric resolutions. Multimodal remote sensing enhances our understanding of physical phenomena by combining records acquired by different remote sensing devices and platforms and allowing higher granularity of information that can be extracted on the physical-chemical processes occurring on the ground. The ability to provide a synoptic view of large areas at regular intervals makes multimodal remote sensing fundamental in obtaining a precise characterization of nature, extent, and yields the potential of analyzing dynamic phenomena such as those affected by climate change. The proliferation of remote sensing data also gives rise to larger diversity and higher dimensionality of related datasets. This development offers the opportunity for better monitoring and more precise characterization of key environmental parameters, such as: biophysical parameters assessment; natural resources use, potentials and limits; water quality assessment; atmospheric pollution estimation; monitoring natural disasters and catastrophic events.

In this context, the H2020 Green Deal project IMPETUS aims to integrate remote sensing datasets and products into a coherent multi-scale, multi-level, cross-sectoral climate change adaptation framework to accelerate the transition towards a climate-neutral and sustainable economy. This goal will be achieved through the development and validation of Resilience Knowledge Boosters (RKBs). The RKBs are open knowledge spaces customized for different regions, within which stakeholders will be able to design, monitor and evaluate climate adaptation measures using available data on climate change impacts on the environment, society (including traditions and cultural values), economy and infrastructure. These architectures can be implemented at multiple scales, e.g., at different governance levels (local-local, region-region) or at multiple administrative levels (local-regional-international). In the context of RKBs, remote sensing data will be complemented by additional data collected on the ground and assessment methods to support decision and policy making within a process of co-creation with local stakeholder communities. The result of this approach will be the co-creation of regional Adaptation Pathways, based on the exploration and sequencing of sets of possible actions aiming at optimizing adaptation and mitigation approaches to climate change in a specific region based on the needs of local communities. This is expected to lead to increased community empowerment in terms of adopting and deploying Innovation Packages as part of coherent Adaptation Pathways. By integrating remote sensing with societal, financial, administrative, economical, and environmental inputs, the RKBs will become open federated spaces for sharing data, knowledge and experiences. The RKB solution will be deployed and validated in all seven EU biogeographical regions (Continental, Coastal, Mediterranean, Atlantic, Arctic, Boreal, Mountainous) covering key community systems, climate threats, and multi-level governance regimes.

As the climate system warms, the frequency, duration, and intensity of different types of extreme weather events have been increasing. For example, climate change leads to more evaporation that may exacerbate droughts and increase the frequency of heavy rainfall and snowfall events. That directly impacts various sectors such as agriculture, water management, energy, and logistics, which traditionally rely on seasonal forecasts of climate conditions for planning their operations.

In this context, stochastic weather generators are often used to provide a set of plausible climatic scenarios, which are then fed into impact models for resilience planning and risk mitigation. A variety of weather generation techniques have been developed over the last decades. However, they are often unable to generate realistic extreme weather scenarios, including severe rainfall, wind storms, and droughts.

Recently, different works proposed exploring deep generative models in weather generation and most explored generative adversarial networks (GAN). [1] proposed to use generative adversarial networks to learn single-site precipitation patterns from different locations. [2] proposed a GAN-based approach to generate realistic extreme precipitation samples using extreme value theory for modeling the extreme tails of distributions. [3] presented an approach to reconstruct the missing information in passive microwave precipitation data with conditional information. [4] proposed a GAN-based approach for generating spatio-temporal weather patterns conditioned on detected extreme events.

While GANs are very popular for synthesis in different applications, they do not explicitly learn the training data distribution and therefore depend on auxiliary variables for conditioning and controlling the synthesis. Variational Autoencoders (VAEs) are an encoder-decoder generative model alternative that explicitly learns the training set distribution and enables stochastic synthesis by regularizing the latent space to a known distribution. Even if one can also trivially control VAEs synthesis using conditioning variables, such models also enable synthesis control from merely inspecting the latent space distribution to map where to sample to achieve synthesis with known characteristics.

This work explores VAEs for controlling weather field data synthesis towards more extreme scenarios. We propose to train a VAE model using a normal distribution for the latent space regularization. Then, assuming extreme events in historical data are rare, we control the synthesis towards more extreme events by sampling from normal distribution tails, which should hold less common data samples. We report compelling results, showing that controlling the sampling space from a normal distribution implements an effective tool for controlling weather field data synthesis towards more extreme weather scenarios.

We used the Climate Hazards Group Infrared Precipitation with Stations v2.0 (CHIRPS) dataset \cite{funk2015climate}, a global interpolated dataset of daily precipitation providing a spatial resolution of 0.05 degrees. The data ranges from the period 1981 to the present. We experimented with a one-degree by one-degree bounding boxes around the spatial region characterized by the latitude and longitude coordinates (20,75) and (21,76) that geographically corresponds to Palghar, India. We used daily precipitation data from 01/01/1981 to 31/12/2009 as the training data set and the data from 01/01/2010 to 31/12/2019 as the test set.

In our dataset histogram, one can also visually inspect that India's monsoon period begins around day 150 in the year and goes on to around day 300. We considered sequences of 32 days in this time range and 16 bounding boxes randomly picked around the coordinate center and selected 18.000 random samples for composing our final training (14.400) and testing (3.600) sets.

Our tested architecture consists of the encoder and decoder networks. The encoder architecture consists of two convolution blocks followed by a bottleneck dense layer and two dense layers for optimizing the means and standard deviations that hold the latent space distribution that is sampled to derive the latent array following a standard normal distribution. We employed two down-sampling stages (one for each convolutional block) to reduce the spatial input dimension by four before submitting the outcome to the bottleneck dense layer. After the convolutional and dense layers, we also applied ReLU activation functions. The decoder receives a latent array as input with the size of the latent space dimension that is ingested to a dense layer to be reshaped into 256 activation maps of size 8x8x8. These maps serve as input to consecutive transposed convolution layers that up-sampling the data up to the original size. A final convolution using one filter is applied to deliver the final outcome.

For running the experiments, we used the Adam optimizer with a learning rate of 0.0001, beta1 as 0.9, and beta2 as 0.999. We implemented a warm-up period of 10 epochs before considering the regularization term in the loss, which was weighted using the beta-VAE criteria. We trained the models for 100 epochs, with 32 data samples per batch, and monitored the total loss to apply early stop. All experiments were carried out using V100 GPUs.

We evaluated our results using quantile-quantile (QQ) plots, a probability plot used to compare two probability distributions. In QQ plots, quantiles of two distributions are plotted against each other; therefore, a point on the plot corresponds to one of the quantiles of a given distribution plotted against the same quantile of another distribution. In our cause, one distribution is computed from input samples pixel values, and the other from the reconstructed samples pixel values. If these two distributions are similar, the plot points will approximately lie on the line where axis-x is equal to axis-y. If the distributions are linearly related, the points will approximately lie on a line, but not necessarily on the line where axis-x is equal to axis-y. We also plot the historical data distribution to be used as a reference in the test set plots.

First, we observed that the historical data is not precisely defining the test data distribution. That indicates the distribution of precipitation values suffered a shift from the training time interval (1980 to 2010) to the testing (2010 to 2020), especially for higher quantiles, meaning that higher precipitation values became more common than in historical data. We also observed that our trained vanilla VAE model synthesized data that matched the testing distribution up to around 70mm/day but then failed to match the quantiles for higher precipitation values (considering we trained the model using historical data, that is somewhat expected).

Concerning synthesis control, we created reference extreme weather data, considering the 10% and 30% samples with the greatest and lowest average precipitation values, and then evaluated if our sampling schema can control the synthesis towards those samples by simply varying the quantile threshold. We synthesized samples that have distributions coherent with those selected as references for more or less extreme weather field data.

We also observed that samples from the average standard deviation sampling are similar to those drawn from real data, as expected since they are more likely to happen. The samples synthesized using smaller standard deviation values depict weather fields with lower precipitation values, and the ones using larger standard deviation seem to show higher precipitation patterns.

Our work explored the efficient use of variational autoencoders as a tool for controlling the synthesis of weather fields considering more extreme scenarios. An essential aspect of weather generators is controlling the synthesis for different weather scenarios considering climate change. We reported that controlling the sampling from the known latent distribution is effectively related to synthesizing samples with more extreme scenarios in the precipitation dataset experimented in our tests. Further research expects to explore models that enable multiple distributions for more refined synthesis control and tackle data with multiple weather system distributions.

[1] Zadrozny, B., Watson, C. D., Szwarcman, D., Civitarese, D., Oliveira, D., Rodrigues, E., and Guevara, J. A modular framework for extreme weather generation, arxiv.org/abs/2102.04534, 2021.
[2] Bhatia, S., Jain, A., and Hooi, B. Exgan: Adversarial generation of extreme samples, arxiv.org/abs/2009.08454, 2020
[3] Wang, C., Tang, G., and Gentine, P. Precipgan: Merging microwave and infrared data for satellite precipitation estimation using generative adversarial network. Geophysical Research Letter, 2021.
[4] Klemmer, K., Saha, S., Kahl, M., Xu, T., and Zhu, X. X. Generative modeling of spatio-temporal weather patterns with extreme event conditioning, arxiv.org/abs/2104.12469, 2021

In recent years, climate change is posing increasing threats at global level: developed and developing countries are affected by natural hazards, drought, floods, sea level rise just to name a few. The more a country is developed, the better it reacts to climate pressure thus, even though climate change is a worldwide challenge, less developed countries are those paying the higher price. On the other side, strong efforts are devoted to monitor the Earth-Atmosphere system: new satellites, new models and environmental services (among which the role of the Copernicus program is prominent) provide continuous and accurate measurements of climate status and trends.
In this work we focus on prominent climate security issues, namely natural hazards, factors affecting food availability and climate-related health threats.
One of the main effects of evolving climate is change precipitation and temperature regimes, causing floods, drougth, evolution of the natural and anthropogenic landscape. There is a need of indicators to quantify climate risk related to different threats, contextualised on a geographic domain (specific site, region, country). In the framework of the Earth Observation for Sustainable Development (EO4SD) Climate Resilience cluster (https://eo4sd-climate.gmv.com/), a set of climate variables and indicators have been computed, using as baseline data from the Copernicus Climate Change Services (C3S, https://climate.copernicus.eu/), ESA Climate Change Initiative (ESA CCI, https://climate.esa.int/en/) and various other data sources to make available more than 30 indicators for climate screening, climate risk assessment and climate adaptation activities. Indicators have been made available to various entities: a relevant consumer is World Bank Climate Change Knowledge Portal (CCKP, https://climateknowledgeportal.worldbank.org/) that integrates spatially and temporally-integrated climate indicators within the clutry profiles. Another relevant example is the TRENgthening resilience of Cultural Heritage at risk in a changing environment through proactive transnational cooperation (STRENCH, https://www.interreg-central.eu/Content.Node/STRENCH.html) that allows managers of natural and cultural heritage sites to assess climate risk and define mitigation actions through the use of a dedicated webGIS tool fed by a large pool of climate indicators computed from models and satellite data.
The correlation between climate conditions and the effects on the animals and human being health is well known as demonstrated in several studies, while the quantification of this correlation is still under investigation especially in remote areas where the meteorological and climate information are hard to be collected. EO data, associated with epidemiological data about diseases, outbreak and other kind of social-health data, are relevant to analyse the impact and the cause-effect dynamics linking meteo-climate parameters to human and animal health. Furthermore, thanks to the large amount of climate data made available by EO systems, it is possible to predict the evolving risk for those regions where the climate conditions become favourable to disease vectors development and diffusion. In this work we show the results of a study that combines meteorological and climate parameters from heterogeneous data sources (satellite, in-situ, model) with health information related to distribution of plasmodium Falciparum as a proxy for Anophele mosquitoes diffusion in Tanzania. Open source data on plasmodium Falciparum parasite rate available from the Malaria Atlas Project database (https://malariaatlas.org/) containing monthly geo-referenced data on the number of positives and the number of tested subjects. The statistical model chosen for the analysis is based on the Bayesian approach where the outcome variable represents the posterior probability for a human subject to be affected by plasmodium falciparum and the likelihood is assumed to follow a binomial distribution: longitude, latitude and altitude are assumed to have a linear effect, while rainfalls and temperature are assumed to have a non linear effect (i.e. second order random walk).
Food security in certain areas of the world is strongly correlated to insects’ infestation: locusts have always represented a plague for populations living mainly in Africa and Asia. Despite the large use of pesticides to try keeping their diffusion under control, extension of agriculture and climate change have driven an increasing impact of locusts on human life. NGOs are facing challenges in planning support activities in areas where locusts are destroying most of the crop causing famine. Locust swarms are moving every day on long areas, estimating the precise movement could be difficult. Within the ESA EO4YEMEN project, the use of long climatological series, satellite observations and locations reported by the FAO Locust watch portal (http://www.fao.org/ag/locusts/en/info/info/index.html) have been used to develop and train AI-based modules with the scope of forecasting occurrence / apparition of a new swarms. To demonstrate the impact of climate change on locusts´ behaviour, the presence of hoppers recorded on the FAO database has been aggregated by latitude (10 degrees interval from 20W to 80E), over four periods: 1985-2004, 2004-2018, 2019 and 2020-2021 (see Figure 1): the shift from west to east of the maximum occurrence demonstrates how the patterns has changes over time. The correlation of locusts´ occurrences and climate data and indicators has been modelled with the finals scope of forecasting hooper’s presence. Four environmental parameters are used: temperature, precipitation, soil moisture and vegetation tenure. The model is trained over the 10 days before the event occurrence, and provides a warning for the length of the forecast data used to run the model . Four neural network approaches have been implemented: Fully Convolutional Neural Network (FCNN), Long short-term memory (LSTM), Convolutional LSTM (ConvLSTM), and Support Vector Machine (SVM). ConvLSTM provides best performances among the different experiments performed (accuracy: 0.96, macro average: 0.49). The main issue faced is the unbalanced database: while the FAO locust watch database provides locusts detections, a dedicated strategy has been implemented to obtain a similar amount of points with “no detection”.

Climate service centers translate scientific data generated by the scientific community into locally-relevant information that helps with decision-making and policy setting. These translations are important across all disciplines and require various levels of external data integration and fusion. Climate service centers combine future climate projections from modelling centers, remotely sensed data from satellite instruments, and ground measurements from observation networks to create climate products and services tailored to decision-makers' needs. Although the tailoring process is specific to local contexts, climate service centers face similar challenges and with more data and more platforms serving information for climate security the integration challenges are growing. This calls for extending the concept of FAIR data to FAIR climate services, with FAIR being Findable, Accessible, Interoperable, Reusable. Thus, not only data, but full climate service information systems should adhere to the FAIR principles, to be realized by R&I and cooperation to address the global threats of climate change. It requires agreements on metadata aspects for discovery, application programming interfaces (APIs) and resource models for interaction with the climate service information system. Here, Integration experts such as OGC can help because of the combination of standards setting with best practice development and experimentation. It starts with discovery, where many open issues are already addressed by R&I like accessibility which can enhance quickly thanks to RESTful APIs. With GEMET and INSPIRE registry there are promising starting points to achieve semantic interoperability, though linked data principles need to be further explored how data was captured, produced, processed, and fused to make sure that we stay on top of data complexity and integration.
On an operational level two possible approaches can be considered, responding to the global threats of climate change: First, reducing and stabilizing the levels of heat-trapping greenhouse gases in the atmosphere (“mitigation”) and/or adapting to the climate change already in the pipeline (“adaptation”). Climate services that fuse climate change scenario data with local data such as topography, economy, infrastructure, or demographics provide a solid base for adaptation modeling. Thus, these services allow us to understand how to live and adapt to the unavoidable climate change. At the same time, there is an urgent need for actions to address the mitigation side. Here, monitoring of essential variables based on Earth Observation data has still to catch-up with the many approaches based on assumptions and rough estimates.
Implemented in the UNFCCC policy frames, research and systematic observation plays an important role to foster climate information production and provision. International initiatives like the ESA climate change initiative or the European atmospheric monitoring services are built based on Satellite data production. For example a critical issue is to enhance the certainty of atmospheric trace gas monitoring like CO2 or Methane. Due to the increasing quality of affordable sensors, ground truthing analytics are implementable into the Climate Service Information Systems and related Spatial Digital Infrastructures. Data fusion of this IoT and sensor data with the Satellite images are an upcoming opportunity to enhance the precession of Earth Observation based climate services. Further more in bridging EO to the numerical model data of climate change future projection scenarios or shorter time scales like the 6Month Seasonal forecasts there is a key role of R&I and cooperation to address global threats like Hurricane prediction where new data and information products can be provided.
We will present ongoing efforts to meet these objectives, discuss experiences with operational infrastructures, and propose actions to establish climate information systems capable of answering the leading questions of a broad user community to achieve climate security and keep climate change in check.