Tip:
Highlight text to annotate it
X
>>
MURTUGUDDE: I'll describe what we mean by Regional Earth System Prediction. I think
you have some sense about it. So, it's basically about producing environmental information
at very high resolution down to a few meters so it can be used to make decisions. And the
main thing about Earth System is to then link and environmental connectivity to other issues
like human health, water, and so on. And this is a kind of a cartoon for the Chesapeake
Bay which has a lot of aquatic vegetation, which is well-known for providing very good
habitat for crabs and oysters which have been slowly degrading over decades, and all sorts
of regulations are not always able to bring it back. So, life goes from vegetation to
sediment exchanges, bacteria and viruses, microbes to shrimp and crab to--all the way
to fish like sardines and the bay anchovies, perch; you have all sorts of things. It's
human impact and it's a--it's a resting and migrating place for millions of birds every
year. So, can we build a tool that goes from microbes to man so that under climate change,
you can see what kind of impact it will have on the integrated assessment and health of
the bay, and so on? So, to deal with climate change issues, naturally, you will have partnerships
across Academia and National Labs, NGO's and so on; so that the team that works with me
includes people across the campus from Computer Science, School of Public Health, and so on,
people from Coastal Labs, from NASA, from other universities, from NOAA, from EPA and
the Department of Natural Resources, and so on. The entire thing is in the larger context
of initiative, that's actually across many universities but run by what's called Climate
Information: Responding to User Needs, where the idea is to invite users like insurance
industries, defense industries, vintners, public health officials, and so on. And let
them tell us what is it that they need in terms of climate information. So, we have
had two workshops; one focused on insurance and one was a broader, 400 people participating,
included people from the CIA Pentagon, and so on. And the idea is that, we are generating
incredible amount of data in, you know--from the satellites, from all sites of--all sorts
of institutions from N. Sanson. And we have no idea how to depict it properly for a wide
range of users. So, one of the main things we see in terms of the help we could get from
Google is how is it that we can appropriately portray all this incredible amount of information
for use. So, I will use a few examples before I get to the tool that we used to produce
this high resolution information. In terms of motivation, everybody has heard of the
IPCC which won the Nobel Prize with Al Gore, Intergovernmental Panel on Climate Change,
which does global and governance issues. Basically, what is the carbon dioxide concentration we
can deal with? What is the mean temperature rise we can live with? And those global models
basically go from what are called External Forcing; so volcanic forcing, changes in orbital
forcing, so changes in electricity, of liquidy, and so on to human impact. So, they are dealing
with very complicated system, and it's a huge system, computationally very expensive so
they can--they have gone from order 500 kilometer resolution globally in the first assessment
that came out in the late '80s, early '90s to the one that came out last year, down to
about 100 kilometers, the next few may go down to 10 kilometers. But still, despite
all the complexity there, adding in terms of human interactions, carbon cycle, all the
chemistry of the earth, and so on, it--we will still need information on the order of
a few meters to able to make decisions of the kind that I will explain over the top.
So, this is the framework that IPCC has been using, whether have Global Earth System Models
that try to simulate global change impacts and have responses from the society in terms
of adaptation and mitigation, and then they look for what kind of governance decisions
can we make as a global society for sustainability. And there is lots of data that's used for
monitoring, and then you try to blend the data with the Earth System Models under data
simulation and so on, and this is the loop you run through. What we are proposing here
is that, in fact, we need a Regional Earth System Model that's a much, much higher resolution,
that does dynamic downscaling so that you can do actual day-to-day management of resources,
and so on. So that's what I'll basically try to explain. The example I'll use is the Chesapeake
Bay. It's a prototype we are building, it's a system that then can be moved. We already
moved it to Gulf of Mexico and part of India, and so on. So the system can be moved anywhere
but you have to have an objective that is basically to develop a fully integrated model
for any region to do all sorts of things that are necessary. So you go literally from days
to decades of producing this information; not just physical environment, but also ecosystems,
pathogens. There are lots of things like harmful algal blooms, waves, inundation, and so on.
I will show examples of that. One of the main things we are saying is that, you have to
device a decision-making tool where people can interactively play games, try out several
what-if scenarios. And I'll show an example of how we are doing that. So, in terms of
giving a decision-making tool to users, then you have to start working with the users from
the beginning itself. The old paradigm has been to produce lots of climate information
and just put it out there on the web or--and expect that people will come and use it. And
it has worked to some extent, but it's not very effective in terms of getting a feedback
as to what's really useful. So, what we have done is engaged a number of different user
groups like the Baltimore drinking water supply, Regulatory Board, a habitat suitability model
for our watermen that deal with striped bass on a daily basis. People who collect data
for harmful algae in the Chesapeake Bay, this is a big problem all the way from Gulf of
Maine to Gulf of Mexico to the West Coast of California to Oregon Coast Seattle, Alaska,
and so on. So there are lots of--and it's a global issue. And the point there is that,
with global warming, it seems to be--and with population increase, especially coastal population
increase has lots of eutrophication issues, nutrient loading to which seemed to be increasing
harmful algal blooms in every place. The Chesapeake Community Bay Modeling is--are riverkeepers
that are working with us. And there is a group that manages a 40-acre forest that's trying
to use the forecast and tell us whether this kind of thing works or not. One of the most
demanding group is Epidemiology and Public Health Group; especially because with climate
change, there have been increases in zoonotic and vector-borne diseases, water and food-borne
diseases, communicable respiratory diseases like the Swine Flu that's going around, and
invasive fungal diseases. And we've known for a very long time, since the time of Hippocrates
that it's not the demon and the Satan that create health problems, but it's environmental
effects. And the old paradigm of doing this has been mostly to look at the connection
between the time of the year, when temperature gets warm enough to cause infections. This
is an example of Razor Clams and Puget Sound where a harmful algal bloom called Alexandrium
affects the clams. So, there is over 30,000 people that do clamming--recreational clamming
of--in Puget Sound. It's a huge income for the state. And if you're not careful, then
you often get infected by these harmful algal blooms. And we would use those kind of correlations
and say, that's how we can predict impact on human health. And with climate change,
you would basically increase the window over which such an infection of Alexandrium would
happen, and we know for the physical climate, what kind of scales we can predict. So, this
is kind of an empirical approach that's being taken. But the problem is that the climate
change does not get to human health so directly just through temperature. There are lots of
things in the middle like Regional Weather Changes. So, you increase the number of heat
waves, you change weather in terms of extreme events, you get more tornados, more hurricanes
and so on, more rainfall coming more as huge events rather than drizzle, humidity is generally
increasing with warming because of increased evaporation, and so on. So, we collect this
sort of data everyday with a number of different satellites and incredible amount of data is
being generated, but there is no effective way of using this data. So, I think one of
the things we could really use help is how would we use all this information and make
people use it in an effective way. Plus, with these changes, it's not like Alexandrium is
affecting the Razor Clam as soon as temperature hits 13 degrees. There are lots of microbial
dynamics involved, transmission dynamics involved. For example, the way Swine Flu is spreading,
it's not what happened before. It's spreading even when there is no flu season, and we don't
understand exactly why this is happening. And with climate change, it's possible that
these things will continue to happen. So, the Biological and Genetic Research is increasing
our understanding of those things. There's lots of information being collected at molecular
and genetic level, antigen information, and so on. All kinds of water bodies are being
monitored for all kinds of pathogens, for example, buildings are monitored for bacteria,
and so on. And all this information is very different; and the users involved, they're
different. But we don't know how to combine this information. Agriculture, for example,
when you have E. coli outbreaks of spinach or tomato or jalapeno peppers, there is again
changes in habitats and transmission dynamics and transportation that are spreading these
things far and wide but we don't know how to monitor the connectivities of where things
are going. So--and there is responses in terms of withdrawing beef products and so on, and
that in turn affects how things spread. So, in a university environment, you can do research
on how these things affect each other before you see an impact on human health. But yet,
we don't know exactly how to deal with data that's very desperate. There are some physical
information that's very precise, very accurate; and there is some human interaction information
that's not very accurate. But it's still physical information, and we have to combine those
and somehow be able to navigate through the system and see how things spread. And so,
literally we are going from genetic information to community level ecosystem information and
human interaction, population, genetic information. And we have collected lots of data in all
spheres, but we don't know how to--how to look at the interactions between them. So,
this seems like a very Google-rich target in terms of just dealing with all the data
and seeing how it is that it can be made more effectively usable. The best example I can
give you is that, things that are often toxic to human beings are not really genetically
selecting to be toxic to human beings. For example, Vibrio Cholerae or harmful algal
bloom, if you go to the ocean and you get infected, you don't necessarily go back to
the beach and provide a feedback to them, to the bugs so that they can genetically fine-tune
their behavior to make it themselves more toxic or more virulent to human beings. What
they are doing is basically trying to pick a microhabitat for their own survival, for
their own genetic competition. So they're often toxic to human beings purely by accident
like Vibrio Cholerae, for example, bacteria likes a mucus that's on the hard shell of
crabs and copepods, and so on. And the same kind of mucus exists in the guts of human
beings. So, if you ingest water with Vibrio Cholerae, that then freaks out in your stomach
but it could found a similar environment. So, you have to understand how Vibrio Cholerae
picks this microhabitat, how it then exploits other similar habitats. So we have to basically
understand the transmission dynamics and microbial dynamics that go on. And it literally happened--even
Vibrio Cholerae itself has a bacterial; if not, toxic. It gets something called Phage
Transduction where viral--a virus attacks Vibrio Cholerae, the bacterium tries to inject
genetic material into it, and sometimes the bacteria dies and it's called Lysis. So there's
kind of something called Red Queen Hypothesis where there is a--there is like is a genetic
evolutionary arms race between the virus and the bacteria and the bacteria ends up dying.
It's like a terminal cancer. And there is another one called Cheshire Cat Hypothesis
where the bacteria can go from being deployed or having both male and female genes to just
get rid of one of the genes and become haploid and get rid of the viral genetic material
and escape the virus attack. So, these things are understood very well. But--and we have
collected this information; but then, to see how those things then begin to explode and
become a global epidemic needs other information. For example, there are environmental connectivities,
and we are very good at generating information about environmental connectivities by generating
either observed information from satellites, light temperature, winds, humidity, radiation,
and so on, or by using models and predict it for the future. But there is lots of physical
connectivities. This is something that started in Vietnam and explode it to--got transmitted
to the rest of the world. And this is a map of daily domestic flights in the US, and this
is a map of the commuter traffic in the US. And these give us a way to look for physical
connectivities. So if somebody comes into Boston, New York or Miami, Houston, whatever,
with a certain disease, we can use this physical connectivities and make scenarios of how quickly
things spread through this physical connectivities. And obviously, this is something that is again
very Google-rich target. So, if Google can track this and combine with their flu algorithm
they used last year based on search for example, to--they made flu predictions that were 10
days ahead of CDC. If we can connect that kind of search to environmental and physical
connectivities, then it's really a powerful tool that can extend that predictability to
many more days. Plus, we can collect data over people in a given city, for example,
Miami, Boston, New York, whatever, because certain sample of people who live there for
a certain number of years and have kids and so on. And you collect the data on where they
spend the time during the day; in a school, in a bus, in an office, and then you can use
simple Euclidean graph theories to connect them, to see if there is an outbreak in one
school of swine flu or whatever. What could be the best strategy to prevent rapid transmission
in terms of vaccinating them, quarantining them, closing down schools, et cetera? So,
you can create local connections, connectivities, environmental, physical and human interactions,
and global connectivities. And we know how to collect data more than we know how to depict
the information or how to navigate through this information. And I think this is a very,
very rich target for Google. And we are, you know---the technology is improving in terms
of going from molecular probes, genetic essays, and so on, where we can create using Nanotechnology,
not just observed temperatures and salinities, and so on or humidities and radiation, winds,
whatever. You can even begin to track DNA level information and see how bacterial phase
transactions are happening, how that is being picked up by other bigger bugs, moving up
the fish that we eat, or how it is getting into the atmosphere, and so on. And we can
even go all the way up to satellites and how to combine that information and use it very
rapidly. So literally, I would say this was something that was proposed by a French scientist,
Rosnay in the 1970s. It's very schematic but it's really worth thinking about in terms
of, we have a telescope for observing stars, we have a microscope for observing microbes.
How do we observe the interactions between nature and society which is basically what
determines everything that happens on a day-to-day basis? The concept is something called Macroscope;
and it's basically I think the way we can combine this incredibly different kinds of
information, and I think maybe Google is the best partner for people who do this kind of
research. And incredible amount of discreet data is being collected, for example, this
is Meteorological Stations by NOAA that beam data everyday. This is air-quality data by
EPA. And CDC collects a lot of survey data asking people questions about their respiratory
asthmatic conditions or, you know, older people heat waves connections to humidity increases,
and so on. So we are basically working with CDC and NOAA and EPA, collecting all these
information with computer science people. We are finding effective ways of kriging the
data, so that you can interpolate it and look for pathons. When you have discreet data,
it's very hard to look for pathons. So, if we can use this information that are, you
know--the survey data is an accounting level because of personal privacy requirements,
you cannot locate address by address where you got this information but it can aggregate
to a county and say, "This county has a swine flu outbreak or this county has a high level
of respiratory morbidity among all the people and so on." So we can combine these information
and use the tools that I'm going to show now in terms of producing environmental information,
and use statistical downscaling and dynamic downscaling techniques to actually even offer
information to hospitals so that they can call the people that they take care of and
say, "In the next few days or few weeks or this season, there's going to be a high chance
that you are going to be suffering more, so what precautions should you take, what kind
of allergies you might face?" You know, and so on. So, there's a very clear seasonality
and morbidity and so on. So, I will now show more specific examples of the environmental
information we're generating in the Chesapeake Bay. When you do this in any particular location--we
were in a meeting last week as I was telling [INDISTINCT] in Helsinki to try to do this
for Iran which has its own issues. For the Chesapeake Bay, the sea level is rising at
twice the global rate because there's something called post-glacial subsidence. So, after
the ice sheets on the continent melted about 10,000 years ago, continent--in front of the
ice sheet was raised and now it's sinking. Plus, the global sea level itself is rising,
so with this combination, sea level in the bays increasing at twice the rate which increases
the interaction with the septic systems and so on. Population has been increasing and
it's projected to increase before station has been going on very rapidly which increases
the chance of increasing all kinds of pollutants coming into the bay. Plus, if you add up all
the coastline of the rivers and tributaries, that's longer than the entire U.S. West Coast.
So, there's a lot of water front property here and people like *** Cheney and Rumsfeld
have moved into this and now it's called, "The Master's Vineyard of the South." So,
lots of very fancy neighborhood is coming up. So with that, the habitat degradation
is almost continuous. There's lots of impact on the water quality and so on. And in any
given region, you have to know what is predictable when you want to produce environmental information.
And most of the water bodies around the world and around the U.S., you have things that
you know exists in the water whether it's a pathogen, whether it's a harmful algal bloom
or whether it's a kind of a toxin that affects fish and so on. In the Chesapeake Bay, we
have one main thing called Vibrio Vulnificus which is a relative of Vibrio Cholerae and
Vibrio Parahaemolyticus. Vibrio Cholerae is what causes cholera, a big problem in Bangladesh,
India, Africa, South America and so on. Vibrio Cholerae exists in this water. In fact, in
the 1800s, before sanitized water supply became a routine thing, there used to be big outbreaks
and more than a 1000 people died in Baltimore in the 1850s. But now, you don't have cholera
anymore because you don't ingest as much Vibrio Cholerae per liter. But Vibrio Vulnificus
affects fisherman through cuts and nicks and it causes up to 50 hospitalizations per year,
more than 10 deaths per year and it's pretty serious. So, we work with Waterman to collect
data and this is a predictable signal. So, there are lots of other predictable signals
like something called Microcystis which is a sign of bacteria which produces a hepatotoxin
and it kills livestock and wild life every year. But the main thing there is that none--most
of the states don't have a regulation as to how much sign of bacteria you can have in
the water supply. For even a sub-acute dose causes severe liver damage and often times
it's too late to know that you've been affected by a microbac--microcystis. There are other
things like Karlodinium, there is Prorocentrum, there is Pseudo Nitzschia and so on which
are all harmful algal blooms which get affected by the warming and by increased population,
producing lots of nutrients that run into the bay. So, every time there's a nice lawn,
people put fertilizers and so on. So, all those things run in to the water bodies. And
this is a problem everywhere, off of the West Coast of Florida; you have something called
Karenia Brevis which in fact produces a toxin that gets aerosolized so you don't have to
be in the water. You'd be walking on the beach and you be in to have very serious asthmatic
problems because the toxin is in the air. So, we are predicting some of those things.
Sea nettles is basically jellyfish. A lot of rich people sailing in the Chesapeake Bay,
they want to know if they can go in the water or not so there's a lot of demand for sea
nettles. Human pathogens, I already mentioned. Anoxia is where basically when you have too
many nutrients coming in, you have big blooms of Phytoplankton that suck up all the oxygen
when they die and sink so that produces the so called Dead Foams. Fish that can't swim
away--do swim away but crabs can not go running very far so they all end up dying. This is
a huge problem in the Gulf of Mexico where massive shrimp kills happened so that's a
big problem. With the Warming, we know that we are decrease--we are increasing the night
time minimum temperatures so the insect infestation is going up. And Obama administration, for
example, has an idea to digitized all PHR or personal health records and we are trying
to convince them that in fact, environmental information should be part of this PHR because
a lot of people are affected by the environmental conditions, their asthmatic conditions, arthritis,
etcetera depend on what the--whether is--going to do. So, we can produce personalized preemptive
and predictive health information and you know that this is another Google rich target.
There are lots of other things that I'll come back to in a minute. The basic tool we are
using is a dynamic down scaling using regional atmospheric model. So, as I mentioned before,
the IPCC models and climate forecast models basically use--produce information at about
100 kilometers, maybe they'll get to 10 kilometers at some point but we want to produce--this
can be run at half a kilometer and then you can combine with [INDISTINCT] observations
and do dynamic down--statistical down scaling down to several meters so you can produce
information that order off a block of a city so the hospitals can use this information.
And there is a terrestrial water shed model that tracks all the nutrients that are getting
into the water, E. Coli, new--sediments, chemical and biological oxygen demand and so on. Everything
runs into the water already that you're considering where you're doing tides, circulation, salinity,
temperature, chlorophyll water quality; you're tracking air quality and so on. So, you obviously
produce winds, temperature, humidity, radiation etcetera. But the number of people interested
vary across so it could be aviation industry, it could be public health, it could be companies
that are trying to do solar and wind energy. So, we--if we can reliably predict in the
next month, for, example, or even next 15 days, how many sunny days you're going to
have, how much wind you're going to have? They can properly plan how to produce energy
and distribute energy because it's not very easy to save energy in those situations or
how to trade between conventional energy and solar and wind energy. So, we can produce
designer forecasts of various fields at 8 days and beyond. And the challenge would be
that we will need help from somebody like Google to be able to say how uncertain our
information is and to be able to give it in the proper format for these different users.
They—-insurance company cares about certain level of accuracy but aviation industry wants
a much different accuracy. Solar and wind energy will need a very different accuracy.
So, this information will have to be depicted in very different ways. The other things we
do is stream forecasts. As I said, it can do sediment loading, nutrient loading, E.
Coli loading, chemical and biological oxygen demand and so on. We carry skills course basically,
validation is a very important part. And the other thing we can do that we are doing routinely
now is to down scale this information for the future so that policy decisions can be
made. So, this is our depiction of how the 2040s will look over the Chesapeake watershed.
So, this you combine with various land use scenarios. So, you have to change the crop
types, you have to use Smart Growth concepts and so on because global models do not do
this very efficiently. So, when you do dynamic down scaling for very high resolution, what
you can do is change how the land will look with erosion or with more population, more
bridges, more urban centers, more paved areas and so on. And that leads me to how the terrestrial
water shed model works and this is an incredibly important part of the whole system. We have
to have information--3D information, digital elevation maps with all the cropped types
that are in there, how much forest there is, how much urban land type there is, how the
cities are growing, what kind of crop changes are happening, how much fertilizer is being
put in to the water, how much water is being withdrawn, what kind of tillage practices
are used and so on. So, you can imagine taking Google Earth to actually tract all that information
and this is incredibly crucial to figure out what is it that is getting in to the water
and ending up in the water body that you are looking at. So, we've been combining that
information from very different sources and trying to put it in but it's not very easy
to get this data routinely for different regions that you want to look at. It's an incredibly
labor intensive work. And for the rivers to do them right, we need to know how wide they
are, how deep they are so it needs to get really 3D in terms of what information we
need to get. And, this is supposed to be an animation but we don't even know how to depict
this information properly but, again, something we can really use Google's help. This is our
projection for the coming season, for example, in terms of the stream low water run off in
Rappahannock which runs to the Chesapeake Bay and it's going to be below normal and
that's going to affect how much oxygen there is in the water, how many fish there will
be and so on. So, what kind of help can we get from Google in that? The watershed itself
has, as I said, other than the physical quantities like temperature, salinity, currents, tidal
water levels, non-tidal water levels and so on, has Sea Nettles so that is being forecasted
routinely. Vibrio Cholerae that I mentioned, one of the pathogens but there are lots of
other pathogens that we can track. Carodimium is a harmful algal bloom. There are lots of
other algal blooms but you can already see the very different kind of information being
generated. So, in terms of looking at 3D Ocean, we want to be able to show all these things
together and it will become a little bit more obvious in a minute when I show the decision
making system. So, combining what I mentioned before in terms of physical connectivity and
environmental connectivity, and human connectivity, we can generate information of seasonal--order
a season on how the weather will change and how the pathogens will change. And we want
to be able to combine with human decision making process and be able to provide early
warning systems. So, I think this is another system that's very rich for Google. We are
able to monitor a lot of chemistry and eco systems and the physical parameters. So, how
do we depict these things for users, that's a very crucial thing. So, I will skip this
part which is basically how to combine Smart Growth concepts in a particular neighborhood.
So, we took out certain buildings and used sustainable methods and try to estimate the
cost of cleaning the bay with sustainable cost included versus business as usual and
you can argue that—-actually, you can save money in the long run. So, let me focus a
little bit more on this one. Our main goal, as I said, is really to be able to provide
an interactive decision making tool for users that could be policy makers, police, ambulances,
runners or, you know, people who are planning for the future, city planners and so on. So,
I'll show an example. The next slide is--got a narration where you give people an interface
with the model running in the background and they can change a few things like population
density or pollution or land use change and agricultural mix and so on and then want to
be able to look at various things that they're interested in. So, how would we want to do
that? Here is just one example. But--we'll discuss...
>> We can develop the prediction software for evaluating how land use in the Chesapeake
Bay watershed affects the pollution in the Chesapeake Bay. In the left panel, the current
land use map is drawn and a legend is shown on the right: corn, cotton, sore crumb, soy
beans, peanuts and so on. You can see the bay itself, the Baltimore area, Regency area,
this is the Potomac River, Patapsco River, the bay bridge would be here and this the
Susquehanna River flowing into the bay. The map can be dragged around by the user in order
to see the whole watershed. >> MURTUGUDDE: So, then we can select different
types of crops or land use types like urban or forest or whatever, change them and look
at what comes into the bay and that how it affects jellyfish concentration.
>> In the right panel, you can select the particular output variable that you want to
see plotted over the bay. We are currently able to track nitrogen and phosphorus plots
in the bay, water oxygen content, sea nettle or jellyfish presence, and harmful algal bloom
density. You can also select a month and the year of the calendar. We are using predicted
future climate data such as temperature and rain fall in making some of the plots here.
The middle panel is the comparison panel. Once you compute an output map for the particular
land use pattern, you can store it in the middle panel. Then you can make changes to
the land use map and see how those changes affect the health of the bay by visually comparing
the output map for the regional and for the modified land use patterns. In this way, one
can answer the, "What if" question, what happens to the bay if land use in the future is changed?
So, let's see what is the current distribution of pollutants in the bay. Here, you can see
nitrogen, phosphorus, oxygen, jellyfish, and harmful algal bloom. Let us turn this map
as a baseline reference map. Now, let's see what happens if in the next year, in the year
2010, a lot of corn is planted say in Richmond area for ethanol fuel. Land use can be changed
simply by selecting the crop from a list and draw it on the map. Now, let us compute the
pollutant levels from modified land use. You can see that the difference in nitrogen pollution
is relatively small and is confined to the small area in James River. Similarly, other
output variables are also relatively unchanged. So, we can say that planting of corn won't
significantly change the state of the bay. Let us undo the changes to the map and consider
the second scenario where the crop planted is soy beans instead of corn. Once we compute
the results, we see that the level of pollution is much higher now and the pollution spreads
all over the lower part of the bay including the Rappahannock and York Rivers. If you check
the harmful algal bloom density, we can see that compared to the original land use, the
HAB density is quite high in James River. Let us return to the original map once again
and see what happens if in 40 years--let's say in 2048 or '49, the major cities, Baltimore
and D.C., grow to twice their size. The similar the change but draws the city larger on the
map. You can see what changes that would bring to the bay. Phosphorus levels are significantly
higher, and a particular bad increase is observed in jellyfish density which would occupy most
of the upper bay. Finally, another ability of the software is to see what happens to
the pollutant levels if some types of land use in the bay are used or totally eliminated.
So, let's see, if you want to see what happens if we eliminate all the poultry farms on the
Eastern Shore. You can see, when we compute the pollutant levels, you can see that the
bay is much cleaner now and that the pollution levels go way down and oxygen content is higher
and harmful algal bloom density is much lower. In this way, one can visually check which
specific land use types are most harmful to the bay and what would happen if those land
uses would be restricted in the future. >> MURTUGUDDE: So, basically, that gives you
an idea. That has an incredible amount of--number of different ways it can go. We want to make
it full 3D so that, you know, you have full depiction of the atmosphere and the water
body and the land use and--so people can--like--even as an educational tool, kids can pollute some
part of the bay, see where the water goes, and swim around in the bay and look at where
the oxygen disappears, where dead fish end up floating to the surface and so on. Plus,
it's a very useful decision making tool in terms of people in Montgomery County, for
example, want to see if they clear out the 30 acre of land and put a shopping mall in
there, what could happen to the streams in the area. Plus, you can use it for agricultural
practices, where you think organic methods like low tillage, it's very good for soil
organic quality and that's true but sometimes those methods are very bad in terms of water
quality, we call it leeches, more nutrients into the water. So, you can then play all
sorts of water scenarios for the future and seems to be very useful to do that. This is
another powerful tool that, again, is very Google rich. This is a digital elevation map
that we obtained commercially so that we have street level pot holes and side walks and
everything depicted correctly. And the model, basically, predicts when there is storm or
an invasion surge happening and it goes from being 3D to 2D on the streets so that we can
navigate with the GIS interface through the streets to see which streets are actually
passable, which streets are too flooded to--for ambulances or police to get through, which
hospitals may need to be evacuated and so on. And there are restaurants in Alexandria
and Baltimore, for example, if we can do this accurately, even 2 days before or several--12
hours before, they can decide on whether they want to store perishables like meat and vegetables,
whether the restaurant will be open that night or not because oftentimes the decision has
to be made very quickly. Even though you put sandbags and so on, if the flood is too high,
they have to close and they lose lot of money on the perishables and so on. So, this can
be very powerful in terms of combining digital elevation map with high accuracy, zoomable
capabilities, and so on. Combined with predictions of inundation and storm surge and depiction
of street by street and being able to navigate through those things. So it seems like this
is another thing where we can really use Google's help to see how to properly depict these things.
Basically, that's kind of the story I want to convey, if I can get out of this. Okay.
The other thing we do is fish forecast. I won't go into the details but basically the
very detailed theories on how habitat temperatures relate to metabolic rates so you can literally
go from large scale--you can go from molecular level, kinetic reactions of metabolic activities
to population level densities and their relations between body mass and biomass. In other words,
small things tend to be much more abundant when you add up the biomass whereas large
things like elephants, even though they are large, their total biomass is not as much.
So if you know how much food is available to any given ecosystem, you can use that relation
to basically see how much efficiency of food transfer there is. And we have used that to
make global forecast of different kinds of tuna, blue fin and yellow tail and albacore
and so on. And in a given water body like the Chesapeake Bay, there are relations between
temperature and oxygen and how the predators and prey interact with each other, in this
case menhaden and anchovies, and we are able to make forecasts of recruitment habitats
suitability, in other words, how many larvae will survive given the environmental conditions
and how many adults will survive and how much disease pressure there will be from something
called micro bacterium, which is in the water and so and so on. So this kind of things can
be done. The other thing is to really interact with people who collect the data so you can
use data assimilation techniques with these kinds of predictive models to be able to say
where it would be most useful to collect the data instead of going in the water and throwing
instrument wherever you feel like. You can optimize your distribution and this works
even on land. We have--using that original diagram, I showed a physical connectivities.
We can optimize the web of sensors. So we can decide--if we are trying to track swine
flu, for example, and you know the physical connectivities and you want to monitor incoming
traffic of flights coming in then which airport would you pick to monitor people coming in?
You don't want to do it and you can't always do it at all the airports, for example, but
you can use these techniques to say "You should monitor Atlanta, Boston and Seattle" or whatever.
So you can use physical and environmental connectivities to optimize those kinds of
distribution of sensors and monitoring activities. So that's basically what we are trying to
do with in situ instruments, satellites and some automated sensors. The basic goal I want
to just leave you with is that this scale is covered very well by the national weather
service. It goes from several minutes to about a week. They don't do anything beyond that.
And IPCC is doing these scales where they are looking into what will happen in 2040,
2050 and so on. But not much is happening in this work beyond day eight to a couple
of seasons, to a decade. So we are arguing that this is where lots of decisions get made.
There are lots of emergency type decisions happen here. There are a lots of benefits
for all sorts of industries in this range but almost all society level decisions get
made in this range and it's a very rich target for improving environmental prediction information,
gathering data, optimizing the data gathering efforts and being able to combine all those
data and environmental information to, you know, portray it in a proper format that people
can navigate through it. So that's where I think Google can really, really help us. You
have to know lots of other things about what kind of regional cooperation is needed and
so on. But--those are details but hopefully something resonates with people who work on
these sorts of things at Google and we can talk further on what is the low-hanging fruit
can be, you know, work as a collaboration between an academic setting and Google. So,
if you have any questions. >> This is overwhelming, I have to say. [INDISTINCT]
How do you see the collaboration [INDISTINCT] that has resources and [INDISTINCT]?
>> MURTUGUDDE: You--I mean, one of the richest targets I see is the land use data. So if
Google can't find a way to help us, you know, not just--Google Earth is used so widely now
and Google Ocean also is beginning to be used heavily. But this--be able to monitor, you
know, at--literally at several meter level what kind of crops are being plant--planted,
what kind of slopes there are, how wide the rivers are, how deep they are, then I think
that basically there's a huge incredibly intimate interaction between agriculture, water and
the environment. And that determines pretty much everything including human health and
that all happens basically through these things. So if Google can find a way or if it can help
us provide that kind of information then we can translate it into usable information for
monitoring waters, pathogens, water quality, air quality, and so on. The other one is really
the--just being able to take information that's, you know, atmospheric not just physical but
chemical and even bacterial level information can be collected, same for the water, same
for land. And this information is--it goes across, you know, all physics, chemistry,
biology and human interactions. We don't have any idea how to project this information to
look at the interactions, to make decisions. So, I think if we start with a sample of one
location, combined meteorological information with air quality information, land use information,
and water information then what would we do to best project this information so that you
can--decisions are now being made kind of separately. If you look at the way swine flu
was being dealt with, it's not being combined with this connectivity information to see
how it's spreading because we already know that there's been an outbreak--there's been
an outbreak in Emory University, there's been an outbreak in University of Washington. The
next one showed up in University of Maryland. So how are these things popping up? I think
we can do that and I think Google is probably in the best position to offer these physical
connectivities of commuter transportation--commuter movement, people movement with flight movement
and all the biological information that goes with it to see how can we map these. Can we--I
mean, in the flu situation, I think Google tracked what keywords people were using in
a given location and then they said that the flu outbreak here and that was almost 10 days
before CDC could make a decision about it. So if this--if we combine this with the environmental
connectivity I think you can advance that time by, you know, another fold so you could
go--in the case of swine flu maybe in November or December what kind of swine flu map you'll
have can probably be done by now if we combine this information. I think Google is in the
best position to think about how to combine and project this information so you just can
navigate through it. We don't--we don't know exactly how to do it right now. I mean, we
are using simple cringing methods and so on to produce gridded fields out of this discreet
information of air quality and metrology and survey data from CDC. So of Google can help
us, say, how would we combine this and look for patterns then we can say how they will
evolve in the coming weeks and months. >> [INDISTINCT] It sounds like you're saying
that gathering information and what is available before that can be processed easily and combined.
The first data type is what you need. You need computational, resources as well, new
algorithms, human vehicle to provide this at...
>> MURTUGUDDE: Yes, I think we'll have to kind of work out the details. We will have
to come together as a team from, you know, some people from our side and your side, too.
There is definitely just the graphics of it. How would you project this information in
3D and, you know, make it see how things are spreading or how they are getting around,
you know. So we will have to bring our--our environmental people and computer science
people together with your people to see exactly how you would go about it. Right now, it's
like the users are from very different areas, from public health to nurses to, you know,
city managers, to people who are trying to decide if a person with swine flu could--should
get on the flight or not and so on. And there'll all dealing with very different information.
So how can even combine the information so they all look at the same thing? And, you
know. >> [INDISTINCT] I'm finding a little confusing.
The Chesapeake Bay information has a very detailed model. You have lots of data, includes
lots of data available in your--or you're using that as a--are you using that as an
example of what could be done nationally or globally? Is that your [INDISTINCT]?
>> MURTUGUDDE: Yes. It's a--it's a prototype and why we did it for the Chesapeake Bay is
because these things have regional specificities. So you don't want to build a global--this
works only well at a very resolution and you can only accomplish this in a regional setting.
But it's also advantageous because each region has something that's corky to that region.
The kind of topography you have, the kind of climate you have, the kind people interaction
you have, the kind of water body you have, kind of disease outbreak you have, kind of
agriculture you have and so on, the kind of economy, population movements, et cetera.
So there are--the way it works in the great lakes, the way it works here, the way it works
here or even here and here, they're all very different. So we built the prototype here
and the basic model structure itself can be moved anywhere but it's the land used that
goes into it. For example, what grows here is very different than what grows here. What
will happen in the future here is very different than what will happen in the future here.
So the basic models are the same but the details are different. So we--the way we see it that
you will have multiple structures running, you will have, like, three in the--not three
in the center, let's say three in the South and you would have to combine all that incredibly
detailed and rudimental information and human interactions in each location to be able to
project so that federal government will be trying to make certain decision. Regional
and local governments will be making a certain decision. Military may be wanting to use it
in a certain way. Department of Homeland Security may be trying to use it in a certain way.
So a it evolves, you know, they want all kind of scenarios. If terrorists released something
in Texas, like [INDISTINCT] and Equine Encephalitis, how it will spread to the U.S.? How fast will
it spread? So then you will need to know these environmental connectivities, physical connectivities,
people movement and so on. So, in the coming years--so, you cannot do it with one model.
You have to have high resolution model here connecting to a high resolution model over
there and so on. So there is a natural variability going in and then there are human perturbations
going in and then Homeland Security is interested in just various scenarios. So this is the
best tool to make scenarios from one season to many decades.
>> [INDISTINCT] Are you getting plenty of people here at Google during your visit? [INDISTINCT]
people at Google.org, Google Maps, [INDISTINCT] I know that there are people who are interested
in--well, we might be able to get people decided if he could take the Chesapeake Bay model
and start generating a San Francisco Bay model. >> Yes.
>> Yes. >> It's not a [INDISTINCT] legal system but
it's comfortable in size. >> MURTUGUDDE: But it has incredible issues
with water, water resource. So you would focus more on water resource and hydrology. And
restoration of San Francisco Bay is a big interest for a certain part of the government.
So--and--yes. >> [INDISTINCT] it's a--it's a rich ecosystem
with all different [INDISTINCT] and Central Valley are operable here. So we might be able
to get people understand it that way because there is a lot of [INDISTINCT] and, in fact...
>> And fire is a big thing here. >> Yes. We don't worry so much about that
parallel here. >> MURTUGUDDE: So that--so this--the good
thing is it has all the elements and you can focus on one particular element. That's what
I meant by regional specificity and for me I had already different focus on that, [INDISTINCT]
just be [INDISTINCT]. >> It seems like that there are several groups
at Google who would be [INDISTINCT] interested in this. Google.org and the kind of work that's
been done is by tracking the habits of various [INDISTINCT]...
>> MURTUGUDDE: I think how to just--how to combine very different kinds of information
in a good way, one system that you can navigate through, seems like something Google would
love to do, that's my imagination anyway. >> [INDISTINCT].
>> MURTUGUDDE: That's what we were trying to do here. This data--this data goes like
to almost 1960s and the survey data from CBC goes back to the 1960s. So, we're trying to
look for trends and that relates to vegetation changes because allergens that affect people
often are related to vegetation types, in addition to temperature and humidity. So,
you already see the richness of the different kinds of information, so there's air quality
information involved, just in terms of increasing industrial activity and traffic and so on.
There's meteorological changes sent in, there are people changes just in terms--in terms
of increased populations. So, this is basically to look at trends over the last, at least
20 years to try to say, "What might come in the next 20 years?" So, you know.
>> [INDISTINCT] how long have you been using them?
>> MURTUGUDDE: Oh, this project? We only started about 18 months ago. Brand new in the sense
that the concept where there aren't people wanted to do it but we got funding from [INDISTINCT]
said "Let's just do it." So, it's the first time that somebody is trying to do this.
>> You're also using it to calibrate environment. >> MURTUGUDDE: Yes, so we are doing it basically
to say, existing data helps us do forecast of [INDISTINCT]. We bring a colony and so
on. To validate it, we will need to do this. So--and their argument is in five years, we
have no choice, so we better be doing this right. And if we want to get there in five
years, what kind of data do we need? So, we're trying to drive data needs to say, if you
want this kind of information, this is the kind of info--data we need to start collecting
to validate to put better skills on these things and so on.