UPDATE: David Sheets is scheduled to speak at the next San Francisco WebGL Developer’s Meetup in May.

We have just released our first version of gloc, a toolkit for WebGL and OpenGL ES2 GLSL shaders. The purpose of gloc is to make it much easier to build – and much easier to share – shaders. As a result, we hope gloc will help to further the creative potential of 3D graphics on the web and mobile platforms. (gloc links our creativity!)

With gloc, shaders and shader fragments, for numerical and graphics functions, become portable first-class web resources.

What gloc enables:

• Libraries of shader functions
• Capability-based shader component fall-backs
• Collaborative shading effect mash-ups
• LD_PRELOAD-like functionality
• Compression by factorization of common shader functions

What gloc makes much easier:

• Semantically-linked shaders in visualization systems
• Cascading shaders for selective overriding
• User Interface effects and compositing
• and much more!

Where gloc lives:

The source code for the gloc suite version 1.0.0 has been released under the BSD 3-Clause license on GitHub. A Web interface is immediately available.

How gloc works:

gloc pipeline: gloc transforms ES SL into glo object files which are subsequently linked by glol with other glo resources to produce complete WebGL shaders.

gloc is a multi-stage compiler that accurately parses and reifies the lexical structure of the language tokens, the lexical preprocessor, and the shading language proper. Every token’s position is tracked to the file, line and column, and comments are tagged to their nearest token. gloc’s preprocessor is first-class, open-scoped, and partial, which allows staged evaluation of preprocessor directives as environmental parameters become known.

After the parsing and evaluation of the preprocessor expressions, gloc analyzes the potential shading language token streams for validity (grammatical only in this release, type checking will be added in a future release). gloc’s shading language parser is also open-scoped, allowing for undeclared variables, functions, and types. Scope analysis is then performed on the token streams; and finally, required and exported macros and symbols are inferred.

The result of this scope analysis and a minimally modified reserialization of the source is then packaged into a simple JSON format, glo. The glo format we have designed makes accurate linking as easy as possible with a concise linking algorithm, glol. glo includes support for link maps indicating the location of each linked source fragment to aid debugging. Also included is optional hyperlinked authorship metadata declaring the copyright holder, author, license, library, version, and build. We recommend transporting glo shader libraries in the glom manifest format which provides an ordered, named, hierarchical JSON structure around each glo object. The glom format and glol algorithm are flexible enough to provide storage and transport for arbitrary application-specific JSON objects in a well-defined order.

Finally, the generated glo files may be linked back into valid shader programs by gloc or a simple JavaScript linker, glol.js, included with the distribution. The linking algorithm takes a list of symbols to be satisfied (['main'] by default) and a search path of glo resources to use in shader construction.

As always, we welcome comments, contributions, and discussions on OCaml, WebGL and pretty much anything cool!

Subscribe to the RSS feed of this blog or join the gloc-users or gloc-devel mailing lists at OCaml Forge

Happy hacking!

As with the WRF model that we also run at Ashima Research, the core MITgcm is primarily used for study of the Earth’s climate addressing crucial questions of atmospheric and oceanic circulation, regional and global climate, and climate change.

The figure, below, shows Martian topography and albedo (reflectivity) on the native cube-sphere grid.  The full globe can be constructed by “folding” the edges of the domain together.  The major advantage over standard grid-point models is the absence of two “convergence” points at the poles.

Data Assimilation (DA) is an enabling piece of your daily weather forecast.  In this case, observations are used to drive a skillful (meaning it gets predictions of the current climate right) model of Earth’s atmosphere towards simulation of a specific day.  What does this mean?  A skillful model of Earth’s climate will get things like the seasonal temperature cycles at a location right, and will even get things like the ‘storminess’ of a given location right, but if you look at output for “August” from the model, what you get is a typical August, not August 2011.  And you’re vanishingly unlikely to get the atmospheric state exactly as it was on 21 August 2011.  For climate or atmospheric science, this doesn’t matter.  But the way you get the weather forecast for 22 August 2011 is to take a global atmospheric state representing 21 August 2011 and integrate it forward by a day.  In that case, you really need the specific global state in the model to represent 21 August 2011.  The way this is done is by the ‘ingestion’ of data into the model with DA – basically the DA system pushes the model towards observations.  The biggest errors for the ‘typical’ vs. ‘specific’ August day tend to be associated with the ‘phase’ of weather systems (is the eastward-moving low pressure storm system centered east of the UK over the Atlantic at this time or is it already over Sweden?)  DA allows the model to “lock up” on the real timing and distribution of weather events for a real day.

Why do we care about this for Mars?  We’re not really interested in weather forecasting for Mars.  But we are interested in being able to find the global atmospheric state that corresponds to sparse and irregular spacecraft observations of Mars on specific days.  Sparse and irregular data are difficult to handle and can cause errors in analysis if you’re not very careful.  If this care has already gone in to globally extrapolating the spacecraft data to a nice, regular model grid, the data can be much more widely and easily used.  This is the idea behind the creation of “reanalysis” data for Mars (reanalysis is another hold over from Earth weather forecasting terminology).

There is a lot more to be said about DA for Mars – we’ll have more blog posts on it in the future.  For now, the DA system can be accessed via Ashima Research’s www.marsclimatecenter.com website.

EDIT: We’re continuing to add more videos of and from the flier as we make them

Newer, Still Rather Cheap 5.8 GHz 720 CMOS Camera

Fighting winds in the park in Corona, Jan 25 (no onboard footage) – click here
Flight in the park, 17 Jan – click here
Nighttime flight with camera a 45degree view angle – click here
Evening flight over the building – click here
Nighttime flight with camera looking nadir – click here

Old, Super-Cheap 2.4 GHz Camera Test Camera

Long flight video – click here
Picture-in-picture from live camera during flight – click here
High-level flight over the building – click here
Nighttime flight – click here
Night roof flight 1 – click here
Night roof flight 2 – click here
Sunset onboard flight – click here
Park Flight 1 – click here
Park Flight 2 – click here
Park Flight 3 – click here
Parking Lot – click here

Recently, interest has been stirred again in orbital fuel depots. This is a brilliant alternative to the NASA’s new and doomed big rocket project (the Space Launch System). Doomed because its Apollo style approach means it will be impossibly expensive to build and will consequently also kill any projects in the near term that could ever use the rocket.  Fuel depots make vastly more sense because you can loft fuel on existing rockets. Placing depots in space also means that companies know there’s a market for certain types of launch. Fuel launch should be the cheapest of all. The fuel itself costs almost nothing, and thus the vast effort that goes in to making human rated rockets safe and comfortable for passengers can be completely avoided (only range safety needs be considered). Rockets can also be “right sized” for the market.  Secondary effects of having fuel depots could be significant in terms of businesses that might to refuel commercial and scientific satellites from fuel purchased from these depots.

(NOTE ADDED IN PROOF: Here’s what the former head of NASA’s Johnson Spaceflight Center has to say on the topic – somewhat similar!)

Why fixate on fuel? Because fuel mass ends up being a major factor of total vehicle mass. And there’s such a disparity in the “quality” of the mass between the fuel and the spacecraft hardware (one is very bulky and vanishingly cheap to make, while the other is vastly more expensive to build but can make up less than half the required mission mass).  There’s also the profound issue that if you can refuel you can reuse. We have a space station, after all.  We all ready have a harbour in space. It now makes essentially zero sense to start all missions on the surface.

We know finally have the opportunity to start building true spacecraft. If the vehicle can be reused and does not have to fly through / enter the atmosphere, vastly more interesting vehicles can be built.  The current “long range” vehicle being considered by NASA is the Multi Purpose Crew Vehicle – another sad and unimaginative throwback to Apollo. Capsules are necessary if you launch from Earth on a huge booster, fly your mission, and come straight back to Earth. They’re not very useful for long range missions – their crew mass and shielding is inadequate and you’re now bring your whole Earth reentry system with you as an incredibly expensive tourist on any mission you fly. If you’re not constrained by this, you can build bigger and more complex vehicles, and you can reuse them again and again.

Alternative ideas at NASA – ones not hopelessly stuck in the 1960′s – have involved building true reusable spacecraft that take advantage of fuel depots – and indeed can expand as new propulsion systems come along.  What’s also revolutionary about these kinds of systems is that we could be working on them now, we don’t have to wait for a mythical big booster, and we could be learning and evolving. Two different concepts whose pictures I’ve added are a Multi-Mission Space Exploration Vehicle and NAUTILUS-X (which has an acronym so hopelessly contrived I will not repeat it here).

The MMSEV is based on submersible ideas. The NAUTILUS-X is based on modular spacecraft design and the incorporation of artificial gravity. These are immature designs, but this radically futuristic stuff is what we could be starting today. Add a resuable lunar lander, and you have a lunar mission system. No Saturn V required. Vehicle is assembled at the space station with minimal fuel, all pieces launched on existing commodity launch vehicles (Atlas, Delta, Falcon, Ariane, Proton, etc). Crew flies to the station on a commercial flight operated by Boeing or SpaceX or SNC, etc. The MMSEV or NAUTLIUS-X or whatever the final design is called is moved to the depot, fueled, and off we go.

The great thing about building these systems in LEO is we already have a platform to test the components of a true spacecraft. Test modules can be brought up to the space station. Attached is an example artificial gravity system being tested on the end of ISS’s Node 2.

The crucial thing about all of this is that by bringing commodity purchase into the core of the program maximizes the opportunity for business to come up with systems optimized for cost and safety, not performance and safety. This is the only hope for viable and affordable spaceflight and economic expansion into space.  It also allows us to do revolutionary spaceflight TODAY, rather than squandering 5-10 years on a doomed attempt to rebuild Apollo (SLS and MPCV). We cancelled Apollo for valid reasons – it was a road to nowhere and ridiculously expensive. It was designed as a command economy solution to beat the Soviets to the moon. It was never designed to be a sustainable and economical movement in to space. However, we now have the opportunity to do it right and start actually investing in spaceflight (like the Eisenhower interstate) rather than simply spending money on impressive but ultimately dead-end monuments like Apollo (the Giza pyramids are impressive, but they don’t enable anything).

Kill SLS and MPCV and use those $bns/year to start doing REAL spaceflight and go to the Moon and beyond now!

ADDED: some more recent ideas on using components derived from the International Space Station (ISS) for a main spacecraft. Also ideas on a reusable lunar lander. Neither the SLS nor MPCV (Orion) are needed for any of this. It’s clear that NASA and contractors are thinking through what could be done. If only congress would stop strangling them with “pork rockets to nowhere” and “Apollo on steroids”…

(or maybe “while SLS and MPCV (Orion) are being killed / collapsing under their own weight” because we don’t need them)

On a side note, sometimes the problem of climate change is inverted in public discussion from how I typically conceive it (and opposite to the direction of the underlying uncertainty). The question shouldn’t be “Why is the Earth warming?” (since it’s only relatively recently that we’ve gained confidence that it is). Rather, it has always been much more compelling and reasonable to ask: “Given that we know we’re dumping CO2 into the atmosphere to concentrations not seen in the last million years – and we know that CO2 is a greenhouse gas – what is the likely impact on climate going to be?” I’ve attached two figures that show the recent trends in CO2 and the record over the last 500,000 years. There is no credible argument that the CO2 rise is due to anything but fossil fuel burning and human land-use changes. The expected impact of these kinds of CO2 increases are also consistent with (but there is a broad band of uncertainty here) the types of temperature changes observed. At this point, plausibility is very strongly on the side of our causing the observed temperature increases with CO2 emissions (with a little CH4 and water vapour feedback for good measure).

From a political point of view, it’s probably more appropriate now to be debating what kind of response is optimal – is it cheaper to try and reduce CO2 and CH4 emissions, or is it better to “book keep” for cost impacts associated with a warming climate and just let it evolve? Likely a mix of those is most credible. In any case, that seems to be the message that’s becoming clearer as each year goes by – the questions of if we’re warming and why are resolved (a definitive “yes” and “human greenhouse gas emission”). By how much (and more importantly, with what impact) and what we should do about it or in response to it remain open.

Image from the ESA Venus Express Venus Monitoring Camera.

Venus is the Earth’s nearest neighbour, and has by far the thickest atmosphere of the rocky planets in our solar system. Venus has a CO2 atmosphere with a surface pressure almost 92 times higher than Earth, and 15,000 times higher than Mars, another CO2 dominated atmosphere. Instead of water clouds near the surface Venus is covered in a haze of Sulphuric acid droplets, mixed with water and various sulphur compounds. The extreme surface pressure and thick cloud decks raises the surface temperature to 500 degrees Celsius (900 degrees Fahrenheit) and makes it impossible to see the Sun at the surface!

As part of our NASA funded projects to simulate the atmosphere of Venus, we have been working on a model that reproduces these extreme temperatures using the best information we have on carbon dioxide, water, and the Sulphuric acid clouds. Our model isn’t the first to do this, but we’ve tried very hard to make the model self-consistent, validated against observations, and flexible. We don’t correct the temperature if it’s wrong, we go back and fix our ideas about the atmosphere. This made our life a little harder, but in the end makes the model far more useful for a variety of applications.

In our recent paper introducing this model (Lee and Richardson, 2011) we showed how we used the model in three very different applications, all using the same model and same information on gases and clouds.

In one example, we showed that we could use the model at high (spectral) resolution to calculate the radiation coming from the atmosphere that would be observed by a satellite such as the Venus Express VIRTIS instrument. We didn’t try to match a specific observation, because that requires a careful examination of the differences between what our model suggested and what the satellite actually observed (this process is called ‘signal inversion’ or ‘temperature retrieval’), but we showed that we could see the same features and that we are sensitive to small changes just as the real atmosphere is.

In our second example, we went back in time to measurements by the NASA Pioneer Venus satellite (and more modern data from Venus Express). When these satellites looked at Venus they found a very bright planet that reflected as much as 75% of the sunlight back to space. This makes Venus much brighter than the Earth, which reflects about 30% of the sunlight back to space. These values are captured in something called a Bond Albedo, which measurements of Venus stretching back over 4 decades suggest is between 70% and 80%. Out model gives us a value of 74%! We get an even more detailed picture than the satellites of course, because we can ‘see’ below the clouds with our model. We looked at the measurements from five Soviet era Venera landers (8-12) and the NASA Pioneer Venus lander (if you’ve read the paper, you will have noticed the errors in the labels on the axes of the figure that shows this. The lower X axis should be cosine(zenith angle), and the upper X axis should be Latitude, that’s a twofer!)

In the first two examples, we fixed the temperature and composition (gases and clouds) in the atmosphere and modeled the radiation throughout and above the atmosphere. In the final example, we let the model decide on the temperature profile, allowing the model to heat or cool the atmosphere until it reaches an equilibrium (the radiative equilibrium). Then we added the constraint that the atmosphere should be stabilized by vertical mixing or convection (radiative-convective equilibrium). In the former case our model suggests that the atmosphere wants to reach a surface temperature closer to 850K(~580C,1080F) than 750K(480C,900F). The difference in temperature between the radiative equilibrium and the radiative-convective equilibrium tells us how much the convection affects the lower atmosphere. With the effect of convection included, we reach a temperature profile that is very close to the few measurements that exist for Venus from the surface to 100km. These results show the model predicting the heating by the Sun and by a green-house effect caused by the Carbon Dioxide and clouds, cooling from the atmosphere at night, and smoothing of the atmosphere by convection. All things considered, this is the most comprehensive test for the model and we have one of the few models that can pass this test!

We’re now hard at work on the next stage of our project. We have a radiation model and we have a dynamical model (a Venus GCM), so we need to combine these to make a much better simulation of the Venus atmosphere. This work presents new challenges, from simple things like making the radiation model fast enough to work inside the GCM, to more complex issues like how we include clouds and convection in the GCM. We already have the two models combined, with a speed-up between 200 and 200,000 (really!) depending on some interesting and some not-so-interesting choices, and we’re working on the finer details of the combined model.

We’ll have more information on our progress at the American Geophysical Union Fall Meeting 2011 in San Francisco, with a poster presentation on Monday afternoon, and an update on the blog soon.