Real Time Collaboration Environments For Regional Global Climate Forecasts Tim P. Barnett Climate Research Division, Scripps Institution of Oceanography, University of California at San Diego, MC 0224.......619-534-3223 9500 Gilman Drive 619-534-8561 (fax) La Jolla, CA 92093 tbarnett@ucsd.edu Nicholas E. Graham Climate Research Division, Scripps Institution of Oceanography, University of California at San Diego, MC 0224.......619-534-8088 9500 Gilman Drive 619-534-8561 (fax) La Jolla, CA 92093 graham@grace.ucsd.edu Dr. Sidney Karin Director, Center For Advanced Computational Science & Engineering San Diego Supercomputer Center University of California at San Diego 619-534-5075 9500 Gilman Drive 619-534-5056 (fax) La Jolla, CA 92093-0505 skarin@ucsd.edu Bilal Chinoy Staff Scientist, San Diego Supercomputer Center 10100 Hopkins Drive, La Jolla, CA 92093-0505 619-534-8305 bac@sdsc.edu 619-534-5077 (fax) Introduction A new global climate forecast system has been developed by the Scripps Institution of Oceanography (SIO) as part of its contribution to the NOAA-funded International Research Institute for Climate Prediction (IRICP). The SIO system uses the latest in numerical ocean and atmospheric models to predict climate globally for periods up to a year in advance. The predictions from these global models of air temperature and precipitation are then projected via a high resolution atmospheric model onto a regional grid (say, 20 km by 20 km) anywhere in the world. These high resolution forecasts are used to force regional crop and hydrological models, and this information, in turn, is provided to regional managers to assist them in their daily and long term planning operations. This 'end to end' concept is being implemented in California, Queensland (Australia), southern Brazil and the Zambezi River drainage area of Africa, among others. The potential financial and social impact of the system described above is huge, exceeding billions of dollars US. The key to maximizing its effectiveness is in shared decision making. Therein lies the problem addressed here. The Problem Practical use or 'application' of the forecast system requires simultaneous input and consultation by a wide range of experienced individuals, e.g. climate scientists, local meteorologists and hydrologists, regional crop experts and state/government decision makers. This group needs to collectively share their knowledge in 'give and take' sessions to arrive at an action plan based on the forecasts and their implications. Needless to say, such a collection of talent from different interest lines and different countries will be quite difficult to bring together for the decision making process. Another problem is that all the individuals and state/government entities mentioned above need to be operating from a common information base. This means they need joint access to all forecast data, current regional climate conditions, current crop/water supply data, etc. While each group may have part of this information, there now exists no system where it is held for common use. The Solution We envision a system of shared resources linked by high-performance networks that will solve the above problems. Essentially, we will need to implement a system that has 'every one on the same page' so that the most informed decisions can be made about future climate impacts. This system will have a number of innovative features: Mirror data base systems: Individuals data holdings on forecasts, current conditions of weather, crops, water supply, etc. will be combined into a single shared data base that is resident on computers in all application areas. Thus, the information available on SIO computers will also be identical to that on, say, the California Dept. Water Resources machines or a computer sitting in the offices of the Zambezi River Authority. All data bases will be updated on a regular basis. High-Performance Networks: Individual participants will need to interactively discuss the forecasts and their implications based on both a 'global view' of expected conditions and a 'regional expert' view based on long term familiarity with climate conditions/impacts in the area of interest. Thus, experts and/or decision makers will have the same information on the screen in front of them as they consider actions appropriate to the expected climate situation (e.g. drought). System commonality: The computer, communications and networking systems to be used in the program will be constructed by a single group so they will be 'plug' and 'software' compatible. Their installation in either United States or other international sites will consist of largely of 'plug in and go' and communications network hook ups. The systems will be serviced from a single site for global updates and regional sites within country for local information. Education and Training: The above system will have two educational/training components. The first consists of training local personnel in the operation and maintenance of the system hardware and software. The second set of training efforts will develop the means and methods by which regional data normally collected by locals can find its way into the data base. Networking Requirements The solution we propose above imposes a number of unique requirements for the communications substrate linking the resource facilities. Some of these are: Data and Information Availability: Users of the system will require access to data that is stored in distributed locations. In fact, given the nature of global climate interactions, it is expected that the various databases of interest will reside in systems spread across large distances. The key challenge is providing transparent access to data and information such that users can retrieve selected datasets without regard to their physical locations. We expect that our proposed system will require a multi-level caching architecture. The information stores themselves will have to be mirrored appropriately, and caching servers that service regional areas of affinity will be required to minimize retrieval latency and jitter. Furthermore, we will require observation and measurement of network traffic to ensure intelligent cache management and placement. Quality Of Service (QOS) : True Location Independence Prototype systems have demonstrated the viability of interpersonal interactions over a packet-switched network. The Internet MBONE overlay network has shown that audio, video and shared white board applications are feasibly carried over the Internet. However, we have also learnt that some applications require a "Quality Of Service" (QOS) guarantee from the underlying networks. For our applications, we expect to have a mixture of scenarios wherein groups may operate in "browse" mode which may not require stringent real-time communications guarantees, but at other "decision" times will require the network to deliver on the application requirements with a degree of confidence. Thus, we are keenly interested in the ongoing research into protocols such as RSVP, which would enable receivers to specify a desired QOS to the network and get, if available, a upper-bound on network delay, jitter etc. Data Security: We envision teams of scientists working together by viewing data and sharing models. One of our most important requirements for a working system is the ability to share such data and models in a secure fashion. End-to-end measures such as data encryption are essential to our application, but we must be able to assume that the network itself is protected and robust. Recent attacks such as the "TCP sequence number spoofing" attack must be guarded against. Another area of concern are "routing" attacks, which could redirect valuable information to unauthorized sources. Summary The global climate forecast system that we have developed at the Scripps Institution of Oceanography (SIO) can be simultaneously used by a number scientists at disparate geographic locations. We require that the communications networks supporting such a collaboration must be reliable, high-bandwidth, secure and incorporate a scalable distributed information architecture. Philomene Offen Scripps Institution of Oceanography, 0224 La Jolla, CA 92093 Tel: 619-534-3224 Fax: 619-534-8561