This document discusses using machine learning and combining multiple datasets including historical track issue data, meteorological data, and satellite imagery to monitor and predict drought, flooding, and diseases like malaria. It provides examples of how this approach has been used to track water usage and irrigation needs in agriculture more efficiently, identify mosquito breeding sites to help prevent West Nile virus and malaria, and monitor air pollution levels. Integrating these different data sources allows for more accurate monitoring, prediction, and management of environmental and public health issues.

2. Combine historical track issue data
and historical high resolution
meteorological data with machine
learning.
Combine Multiple Datasets

3. Water
The Motive
Do well with the Details by
embracing the Big Picture
Prof. David Lary
+1 (972) 489-2059
http://davidlary.info
david.lary@utdallas.edu
Center for Space Science

4. Drought Has Increased Globally Since 1900

5. Worst Drought in 1,000 Years Predicted for American West
A paddle wheeler and a small motorboat sail on Lake Mead, North America's largest man-made reservoir. The water is at its
lowest level since the Hoover Dam was built in the 1930s. The white "bathtub ring" of mineral deposits on the rocks marks past
water levels.
PUBLISHED FEBRUARY 12, 2015

6. Western U.S. Drought Prompts
Disaster Declarations In 11 States
By MICHELLE RINDELS 01/16/14 07:51 PM ET EST
LAS VEGAS (AP) — Federal officials have designated portions of 11 drought-
ridden western and central states as primary natural disaster areas,
highlighting the financial strain the lack of rain is likely to bring to farmers in
those regions.
The announcement by the U.S. Department of Agriculture on Wednesday
included counties in Colorado, New Mexico, Nevada, Kansas, Texas, Utah,
Arkansas, Hawaii, Idaho, Oklahoma and California.
Rancher Ralph Miller, 79, checks on one of many “stock tanks” of water that are
receding due to the severe drought. “I’d say it’s just about as bad as it can get.”
Barnhart, Texas

7. “Water is the new oil”
Jim Rogers, chief executive of Duke Energy
... and many others
Water crisis in California, Texas threatens
US food security
Western water scarcity issues becoming more severe
Western Farm Press, Jun. 5, 2012
University of Texas at Austin
California and Texas produced agricultural products worth $56 billion
in 2007, accounting for much of the nation's food production. They
also account for half of all groundwater depletion in the U.S.,
mainly as a result of irrigating crops.
The nation’s food supply may be vulnerable to rapid groundwater
depletion from irrigated agriculture, according to a new study by
researchers at The University of Texas at Austin and elsewhere.
http://westernfarmpress.com/irrigation/water-crisis-california-texas-threatens-us-food-security

8. Since 1980 the population
of Texas has more than
doubled, but the reservoir
capacity has remained
almost unchanged.
During 2011the reservoir
levels were the lowest
during Sep-Dec that they
have been since 1990.
In 2015 we are starting out
with lower levels than 2013.

9. Smarter irrigation control is invaluable!
If we can use existing
infrastructure it is even better!
.... from farm, to corporate campus, to golf course, to your back yard.

10. When great societal need meets
appropriate scalable solution
there is much societal and
economic benefit to be
gained

11. How?
California Children Example

12. DATE: 12-Feb-2010
DOC NO: 0115056
ISSUE: 02
DMC DATA PRODUCT MANUAL
STATUS: FINAL
Figure 4: Detector and channel layout of the SLIM-6-22 imager
Imager Bank 0
Channel 6
Green
Channel 5
Red
Channel 4
NIR
Imager Bank 1
Channel 1
NIR
Channel 2
Red
Channel 3
Green
Pixel 1
Pixel 14436
Pixel 14436
Pixel 1

14. uneven irrigation
blown valves lead to flooding
Sports fields
agricultural test plots
22 m resolution

15. On average, systems have water
losses of about 17 percent.
49

16. Can you tell which grass has had more water?
Zooming in

17. Neighborhood
Trees
‘green’
pond
golf
course
gated community
dry grass
Trees
Trees
Sports Field

19. 20 lb Airborne hyperspectral imaging system
385 channels between 400-1,700 nm
Hyperspectral data cube

20. J S Famiglietti, and M Rodell Science 2013;340:1300-1301
an accuracy of 1.5 cm equivalent water height.
Because GRACE measures changes in
total water storage, it integrates the impacts
of natural climate fluctuations, global change,
and human water use, including groundwater
extraction, which in many parts of the world
is unmeasured and unmanaged. GRACE-
derived rates of groundwater losses in the
world’s major aquifer systems (4–6) under-
score the critical need to improve monitor-
ing and regulation of groundwater systems
before they run dry.
Regional flooding and drought are driven
by the surplus or deficit of water in a river
basin or an aquifer, yet few hydrologic
observing networks yield sufficient data for
comprehensive monitoring of changes in
the total amount of water stored in a region.
GRACE observations have helped to fill
this gap. They have been used to character-
ize regional flood potential (8) and to assess
water storage deficits in the U.S. Drought
Monitor (9) and are included in annual State
of the Climate reports (10). As an integrated
measure of all surface and groundwater stor-
age changes, GRACE data implicitly contain
a record of seasonal to interannual water stor-
key tools for predicting future water avail-
ability, difficult to validate. Low-resolution
GRACE data, when combined with higher-
resolution model simulations, provide an
independent constraint on simulated water
balances, while also adding spatial detail to
GRACE’s low-resolution perspective (11).
They are widely used to evaluate land surface
models used by weather and climate forecast-
ing centers around the world (12).
Evapotranspiration is a key factor in
interbasin water allocations, yet because it
disperses into the atmosphere in the vapor
phase, it confounds standard measurement
techniques. The ability of GRACE to weigh
changes in water stored in an entire river
basin allows evapotranspiration to be esti-
mated in a water balance framework (13).
Transboundary water availability issues
require sharing hydrologic data across politi-
cal boundaries. However, national hydrolog-
ical records are often withheld for political,
socioeconomic, and defense purposes, com-
plicating regional water management discus-
sions. Several studies have used GRACE data
to circumvent international data denial prac-
tices, including in those involving lakes (14),
higher spatial (<50,000 km ) and tempor
(weekly or biweekly) resolution, for exam
ple through novel orbital configurations, s
that smaller river basins and aquifers can b
observed directly.The availability of GRAC
data at these finer scales, at which most plan
ning decisions are made, would likely ensu
their broader use in water management.
The GRACE-FO mission is on sched
ule for a 2017 launch, but a next-generatio
improved GRACE mission is still unde
design and as yet unconfirmed. Given i
demonstrated contributions to date and th
potential for much more, a future without
GRACE mission in orbit would be an unfo
tunate and unnecessarily risky backward ste
for regional water management.
References
1. P. J. Durack et al., Science 336, 455 (2012).
2. K. E. Trenberth, Clim. Res. 47, 123 (2011).
3. I. M. Held, B. J. Soden, J. Clim. 19, 5686 (2006).
4. V. M. Tiwari, J. Wahr, S. Swenson, Geophys. Res. Lett. 36,
L18401 (2009).
5. B. R. Scanlon et al., Proc. Natl. Acad. Sci. U.S.A. 109,
9320 (2012).
6. K. A. Voss et al., Water Resour. Res. 49, 904 (2013).
7. B. D. Tapley et al., Science 305, 503 (2004).
8. J. T. Reager, J. S. Famiglietti, Geophys. Res. Lett. 36,
L23402 (2009).
9. R. Houborg et al., Water Resour. Res. 48, W07525 (2012
10. J. Blunden, D. S. Arndt, Eds., Bull. Am. Meteorol. Soc. 9
S1 (2012).
11. B. F. Zaitchik et al., J. Hydrometeorol. 9, 535 (2008).
12. S. C. Swenson, P. C. D. Milly, Water Resour. Res. 42,
W03201 (2006).
13. G. Ramillien et al., Water Resour. Res. 42, W10403 (2006
14. S. Swenson, J. Wahr, J. Hydrol. 370, 163 (2009).
15. J. S. Famiglietti, Abstract GC31D-01, fall meeting, AGU
San Francisco, 3 to 7 December 2012.
Mixed picture. Between 2003 and 2012, GRACE data show water losses in agricultural regions such as Cali-
fornia’s Central Valley (1) ( 1.5 ± 0.1 cm/year) and the Southern High Plains Aquifer (2) ( 2.5 ± 0.2 cm/
year), caused by overreliance on groundwater to supply irrigation water. Regions where groundwater is being
depleted as a result of prolonged drought include Houston (3) ( 2.3 ± 0.6 cm/year), Alabama (4) ( 2.1 ±
0.8 cm/year), and the Mid-Atlantic states (5) ( 1.8 ± 0.6 cm/year). Water storage is increasing in the flood-
prone Upper Missouri River basin (6) (2.5 ± 0.2 cm/year). See fig. S1 for monthly time series for all hot spots.
Data from (15) and from GRACE data release CSR RL05.

21. Summary
• Vegetation Index is dependent on amount of
irrigation
• Regular (weekly) remote sensing inspection
could allow us to:
• Appropriate irrigation zones
• Help identify regions of over watering
• Help identify any burst pipes/valves
• Optimize irrigation patterns
• Automate sprinkler system controls
• Progressively more benefit as a specific history
of the plots/site is built up

22. Stage 2

23. FUTURE Water Management
Why Agriculture? ~80% water use US (USDA 2013)
Challenges: Climate change, Drought, Population, non-ag water uses.
Water Use Efficiency: ~50% US (USDA 2004)
Water Mgmt.
“Smart-GRID*”
Delivery
Models
Basin
Geodata
Water/Crop
Status & Forecast
Water Need
Status & Forecast
Water
Agric +Others
Status & Forecast

24. Current Water
Mgmt.
Delivery
Models
Basin
Geodata
Water/Crop
Status & Forecast
Water
Status & Forecast
Water
Agric +Others
Status & Forecast
Water use based on:
Experience Limited estimations
No related info
CURRENT Water Management

25. CWMIS Case Example: Water Use vs. Delivery
TOP: crop water use vs. water delivery (ac-ft).
BOTTOM: water use difference (ac-ft)
Typically save at least 10%
Can be done on a field by field, campus by campus, home by home,
or golf course by golf course basis or for an entire basin.
Alfonso Torres

27. Culex tarsalis
West Nile Virus
The same data infrastructure can also
be used to help combat West Nile Virus
by identifying breeding sites.

28. Malaria Prediction
Using Satellite Data

29. P. vivax is carried by the female Anopheles mosquito

30. Plasmodium vivax is a protozoal parasite and a human
pathogen. The most frequent and widely distributed
cause of recurring (Benign tertian) malaria, P. vivax is
one of the six species of malaria parasites that
commonly infect humans.[1] It is less virulent than
Plasmodium falciparum, the deadliest of the six, but
vivax malaria can lead to severe disease and death.[2]
[3] P. vivax is carried by the female Anopheles mosquito,
since it is only the female of the species that bite.
Plasmodium vivax
Plasmodium falciparum http://www.worldmalariareport.org/

31. Seasonal climatic suitability for malaria transmission (CSMT)
Climatic conditions are considered to be suitable for transmission when the monthly precipitation
accumulation is at least 80 mm, the monthly mean temperature is between 18°C and 32°C and the
monthly relative humidity is at least 60%. These thresholds are based on a consensus of the
literature. In practice, the optimal and limiting conditions for transmission are dependent on the
particular species of the parasite and vector.
Commentary: Web-based climate information resources for malaria control in Africa
Emily K Grover-Kopec, M Benno Blumenthal, Pietro Ceccato, Tufa Dinku, Judy A Omumbo and Stephen J Connor*
Malaria Journal 2006, 5:38 doi:10.1186/1475-2875-5-38

33. 0 500 1,000 Km
Vectorial Capacity
In Zones with Malaria Epidemic Potential
05 August - 12 August 2013
VCAP Values
0
0 - 2
2 - 4
4 - 6
6 - 8
8 - 10
10 - 15
15 - 20
> 20
Country Boundaries

34. Satellite imagery can be used to track mosquito habitats.
High-resolution (5 m) satellite images can identify
very small water bodies, wetlands and other
malaria-relevant land-cover types.
Of the 225 million annual reported
cases of the disease, 212 million of
these occur in Africa. Of the
800,000 Malaria-related deaths
each year, 90% of these fatalities
occur in sub-Saharan Africa.
http://www.itweb.co.za/index.php?option=com_content&view=article&id=52695

35. T H R I V E
T I M E LY H E A LT H I N D I C AT O R S U S I N G R E M O T E S E N S I N G &
I N N O VAT I O N F O R T H E V I TA L I T Y O F T H E E N V I R O N M E N T
Why we care so much?
Approximately 50 million Americans have
allergic diseases, including asthma and
allergic rhinitis, both of which can be
exacerbated by PM2.5.
Every day in America 44,000 people have an
asthma attack, and because of asthma
36,000 kids miss school, 27,000 adults miss
work, 4,700 people visit the emergency
room, 1,200 people are admitted to the
hospital, and 9 people die.

37. Air pollution in Ulaanbaatar, Mongolia

38. Unprecedented levels of air pollution in Singapore and Malaysia in June led to respiratory illnesses, school closings, and
grounded aircraft. This year it was so bad that in some affected areas there was a 100 percent rise in the number of asthma
cases, and the government of Malaysia distributed gas masks.
MODIS Aqua July 21, 2013.
David Lary

40. PM2.5 Invisible Killer

42. TypesofbiologicalMaterialTypesofDustTypesofParticulatesGasMolecules
0.0001 μm 0.001 μm 0.01 μm 0.1 μm 1 μm 10 μm 100 μm 1000 μm
Pollen
Mold Spores
House Dust Mite Allergens
Bacteria
Cat Allergens
Viruses
Heavy Dust
Settling Dust
Suspended Atmospheric Dust
Cement Dust
Fly Ash
Oil Smoke
Smog
Tobacco Smoke
Soot
Gas Molecules
Decreased Lung Function < 10 μm
Skin & Eye Disease < 2.5 μm
Tumors < 1 μm
Cardiovascular Disease < 0.1 μm
Hair
Pin
Cell
0.0001 μm 0.001 μm 0.01 μm 0.1 μm 1 μm 10 μm 100 μm 1000 μm
PM10 particles
PM2.5 particles
PM0.1 ultra fine particles PM10-2.5 coarse fraction
0.1 mm 1 mm
Table!1.!PM!and!health!outcomes!(modified!from!Ruckerl*et*al.!(2006)).!
!!
Health*Outcomes!
Short9term*Studies* Long9term*Studies*
PM10! PM2.5! UFP! PM10! PM2.5! UFP!
Mortality* !! !! !! !! !! !!
!!!!All!causes! xxx!! xxx!! x! xx! xx! x!
!!!!Cardiovascular! xxx! xxx! x!! xx! xx! x!
!!!!Pulmonary! xxx! xxx! x! xx! xx! x!
Pulmonary!effects! !! !! !! !! !! !!
!!!!Lung!function,!e.g.,!PEF! xxx! xxx! xx! xxx! xxx! !!
!!!!Lung!function!growth! !! !! !! xxx! xxx! !!
Asthma!and!COPD!exacerbation! !! !! !! !! !! !!
!!!!Acute!respiratory!symptoms! !! xx! x! xxx! xxx! !!
!!!!Medication!use! !! !! x! !! !! !!
!!!!Hospital!admission! xx! xxx! x! !! !! !!
Lung!cancer! !! !! !! !! !! !!
!!!!Cohort! !! !! !! xx! xx! x!
!!!!Hospital!admission! !! !! !! xx! xx! x!
Cardiovascular!effects! !! !! !! !! !! !!
!!!!Hospital!admission! xxx! xxx! !! x! x! !!
ECG@related!endpoints! !! !! !! !! !! !!
!!!!Autonomic!nervous!system! xxx! xxx! xx! !! !! !!
!!!!Myocardial!substrate!and!vulnerability! !! xx! x! !! !! !!
Vascular!function! !! !! !! !! !! !!
!!!!Blood!pressure! xx! xxx! x! !! !! !!
!!!!Endothelial!function! x! xx! x! !! !! !!
Blood!markers! !! !! !! !! !! !!
!!!!Pro!inflammatory!mediators! xx! xx! xx! !! !! !!
!!!!Coagulation!blood!markers! xx! xx! xx! !! !! !!
!!!!Diabetes! x! xx! x! !! !! !!
!!!!Endothelial!function! x! x! xx! !! !! !!
Reproduction! !! !! !! !! !! !!
!!!!Premature!birth! x! x! !! !! !! !!
!!!!Birth!weight! xx! x! !! !! !! !!
!!!!IUR/SGA! x! x! !! !! !! !!
Fetal!growth! !! !! !! !! !! !!
!!!!Birth!defects! x! !! !! !! !! !!
!!!!Infant!mortality! xx! x! !! !! !! !!
!!!!Sperm!quality! x! x! !! !! !! !!
Neurotoxic!effects! !! !! !! !! !! !!
!!!!Central!nervous!system!! !! x! xx! !! !! !!
x, few studies; xx, many studies; xxx, large number of studies.

43. Hourly Measurements from 55 countries and more than 8,000 measurement sites from 1997-present

44. Aqua DeepBlue
Rank Source Variable Type
1 Satellite Product Tropospheric NO2 Column Input
2 Satellite Product Solar Azimuth Input
3 Meteorological Analyses Air Density at Surface Input
4 Satellite Product Sensor Zenith Input
5 Satellite Product White-sky Albedo at 470 nm Input
6 Population Density Input
7 Satellite Product Deep Blue Surface Reflectance 470 nm Input
8 Meteorological Analyses Surface Air Temperature Input
9 Meteorological Analyses Surface Ventilation Velocity Input
10 Meteorological Analyses Surface Wind Speed Input
11 Satellite Product White-sky Albedo at 858 nm Input
12 Satellite Product White-sky Albedo at 2,130 nm Input
13 Satellite Product Solar Zenith Input
14 Meteorological Analyses Surface Layer Height Input
15 Satellite Product White-sky Albedo at 1,240 nm Input
16 Satellite Product Deep Blue Surface Reflectance 660 nm Input
17 Satellite Product Deep Blue Surface Reflectance 412 nm Input
18 Satellite Product White-sky Albedo at 1,640 nm Input
19 Satellite Product Sensor Azimuth Input
20 Satellite Product Scattering Angle Input
21 Meteorological Analyses Surface Velocity Scale Input
22 Satellite Product Cloud Mask Qa Input
23 Satellite Product White-sky Albedo at 555 nm Input
24 Satellite Product Deep Blue Aerosol Optical Depth 550 nm Input
25 Satellite Product Deep Blue Aerosol Optical Depth 660 nm Input
26 Satellite Product Deep Blue Aerosol Optical Depth 412 nm Input
27 Meteorological Analyses Total Precipitation Input
28 Satellite Product White-sky Albedo at 648 nm Input
29 Satellite Product Deep Blue Aerosol Optical Depth 470 nm Input
30 Satellite Product Deep Blue Angstrom Exponent Land Input
31 Meteorological Analyses Surface Specific Humidity Input
32 Satellite Product Cloud Fraction Land Input
In-situ Observation PM2.5 Target

45. This is a BigData Problem of
Great Societal Relevance
• Collecting data in real time from national and
global networks requires bandwidth.
• With the next generation of wearable sensors and
the internet of things this data volume will rapidly
increase.
• A variety of applications enabled by BigData,
higher bandwidth and cloud processing.
• Future finer granularity and two way
communication will dramatically increase the size
of the data bringing air quality to the micro scale,
just like weather data.
Time Taken
10 Mbps 20 Mbps 50 Mbps 1 Gbps
40 TB training data
4 Gb update
185 days 93 days 37 days 1 day 21 hours
54m 27m 11m 32s

46. Long-Term Average 1997-present

47. Four Corners Power Plants
Sonoran Dessert
Los Angeles Area
CentralValley
Common Fire Area
Close Ups Showing Good Agreement With Observations
Alaska
(a)
(b) (c)
(d)
Great Salt Lake Desert

48. Automated traffic patterns,
driverless cars routing

49. Flight
on
Nov
18,
2014
clear
skies Flight
on
Dec
04,
2014
hazy/overcast
Japan USA WHO/EU
Annual
Avg.
: 15μg/m3
Annual
Avg.
: 12μg/m3 Annual
Avg.
: 25μg/m3
24
hour
Avg.
: 35μg/m3
24
hour
Avg.
: 35μg/m3
Annual
Avg.
:20μg/m3
PM2.5
Air
Quality
Standards
Day
within
EPA
Air
Quality
Standards Day
with
exceedance
of
EPA
Air
Quality
Standards

50. 50

51. Model
Airplane
Details
A
12s
5400
mAh
baLery
pack
per
motor
(current
setup)
provides
approximately
8
minutes
of
flight
Pme.
Flight
Pme
can
be
increased
by
using
higher
capacity
baLery
packs.

52. Flight
Photos

53. Accomplishments
To the best of our knowledge the first time the full sub-
pixel aerosol size distribution has been characterized at
high spatial resolution (sub meter) and high temporal
resolution (every second) using:
• A zero emission, low cost, electric remote control model
aircraft at multiple vertical levels in the lower most 100
m of the atmosphere.
• A car driving daily across a 10 km pixel over an extended
period.
Satellite Pixel
Full Aerosol Size Distribution

54. G E O L O C AT E D A L L E R G E N S E N S I N G P L AT F O R M
G A S P
Four objectives:
1. Develop and deploy an array of Internet of Things remote airborne
particle sensors within Chattanooga to be used to provide real-time
streamed data on hourly particulate levels, both pollen- sized (10-40
micron) and smaller (<2.5 micron) particles.
2. Deploy an in-situ pollen air sampler in Chattanooga to identify specific
pollen types.
3. Merge locally streamed data with already-collected, satellite-based
NASA data to complement and enhance the newly-collected particulate
data and generate Chattanooga-focused particulate maps.
4. Develop web-based visual tools to provide real-time pollen and smaller
particle alerts to end users such as asthma patients, health institutions,
and businesses and other institutions affected by elevated pollen levels.

55. Think Big: Holistic & Comprehensive Informatics
Bio
InformaPcs
Medical
InformaPcs
Environmental
InformaPcs
THRIVE
MulPple
Big
Data
+
EMR
+
Social
Media
+
Machine
Learning
+
Causality
A
Cross-­‐cuXng
PlaYorm
for
Comprehensive
InformaPcs
for
Data
Driven
Decisions
in
Pa<ent
Centered
Care
facilitated
by
High
Speed
Low-­‐Latency
networks,
mulPple
massive
datasets
from
large
distributed
sensor
networks,
EMR,
and
local
cloud
compu:ng.

56. Combine historical track issue data
and historical high resolution
meteorological data with machine
learning.
Combine Multiple Datasets

57. Satellite Images can be used
to automate the highlighting of
vegetation near the tracks
Highlight Vegetation

58. 1
Routine satellite acquisition of
multispectral and SAR imagery
2
Periodic high resolution ground
truth from aerial surveys
3
Image processing & Machine Learning
BNSF Decision Support
4
The synergy between routine satellite imagery,
periodic high resolution ground truth surveys and
automated machine learning and image processing is
a powerful combination for decision support.
Preparing for Routine Decision Support