True Color Earth Data Set Includes Seasonal Dynamics

Eos, Vol. 87, No. 5, 31 January 2006 b a s e d m a t h e m a t i c a l analog to river basins.) T h e r e are plans by the author for validation of s u c h an effort on the Ganges-Brahmaputra-Meghna ( G B M ) basin using the flood forecasting system of the Bangladeshi authorities. S u c h blueprints c o u l d provide frugal m e a n s for c o n d u c t i n g an approximate, yet global, assessment of the numerous IRBs without resorting to c o n v e n t i o n a l distrib­ uted hydrologic m o d e l s that are data inten­ sive and usually require longer setup times. T h e blueprints should b e a m e n a b l e for rapid implementation over IRBs and as such, should b e a b l e to highlight the floodp r o n e nations that s e e m most likely to ben­ efit cost-effectively from anticipated GPM rainfall data. This a p p r o a c h c o u l d subse­ quently motivate flood-prone nations to invest in a range of more detailed studies to design and test an e n h a n c e d GPM-based prototype forecasting system by 2 0 1 0 . References Asian Disaster Preparedness Center (2002), Applica­ tion of climate forecasts in the agriculture sector: Climate forecasting applications in Bangladesh project, Rep. 3, Bangkok. Hossain,F(2005),Towards formulation of a spaceborne system for early-warning of floods: Can costeffectiveness outweigh prediction uncertainty?, Nat. Hazards, in press, doi: 10.1007/s11069-0054645-0 Hossain, F, and E.N.Anagnostou (2004),Assessment of current passive-microwave- and infrared-based satellite rainfall remote sensing for flood predic­ tion,./ Geophys.Res., 109, D07102,doi:10.1029/ 2003JD003986.

T a b l e 1. A n o n e x h a u s t i v e list o f l o w e r m o s t riparian n a t i o n s s i t u a t e d in f l o o d - p r o n e i n t e r n a t i o n a l river b a s i n s a

Name

of Downstream Country

International

Cameroon

S p a c e exploration has c h a n g e d our visual perception of planet Earth. In the 1950s, sat­ ellites revolutionized weather forecasting. Astronaut photography in the early 1970s showed us the Earth in color, the so-called 'Blue Marble' (Figure 1, left). S i n c e 1972, sat­ ellite sensors have been acquiring atmosphere, land, ice, and o c e a n data with increasing spectral and spatial resolution. Satellite remote sensing systems such as the NASA Earth Observing System ( E O S ) help us to understand and monitor Earth's physical, c h e m i c a l , and biological processes [Running etal, 1999]. The false-color Earth image shown in the center of Figure 1, n a m e d Blue Marble, was created in 2000 with data from the Advanced Very High Resolution Radiometer (AVHRR), the Geostationary Operational Environmental Satellite (GOES 8), and the Sea-viewing Wide Field-of-view Sensor (SeaWiFS). New sensors such as the Moderate-Resolution Imaging Spectroradiometer (MODIS), aboard NASA Terra and Aqua satellites, allow the derivation of a wide range of geophysical parameters from measured radiances of a single sensor. B Y R. S T O C K L I , E . V E R M O T E , N. S A L E O U S , R. S I M M O N , A N D D. H E R R I N G

Basin

Percent of Total Basin Area Occupied by the Country

Akpa/Benito/Ntem

41.8

Senegal

8.08

Ivory Coast

Cavally

54.11

Benin

Oueme

82.9

Senegal

Botswana

Okovango

50.6

Niger

26.6

Ganges-BrahmaputraMeghna

7

Bangau

46.03

Ca/Song Koi

35.1

Nigeria Bangladesh Brunei Laos Myanmar

Irrawaddy

91.2

Cambodia

Mekong

20.1

a

These nations would typically depend on rainfall information from the upstream regions (nations) of the IRB in order to realize the hydrologically possible flood forecasting range of the basin response time. Source: Aaron Wolf,Transboundary Freshwater Disputes Database, at Oregon State University, Corvalis (http://www.transboundarywaters.orst.edu). Nijssen,B.,and DPLettenmaier (2004), Effect of precipitation sampling error on simulated hydrological fluxes and states: Anticipating the Global Precipitation Measurement satellites, J. Geophys. Res., 109, D02103, doi:10.1029/2003JD003497. Paudyal, G. N. (2002), Forecasting and warning of waterrelated disaster in a complex hydraulic setting: The case of Bangladesh,//ydro/.5d.y.,4? (S),S5-S18. Webster, PJ, and C.Hoyos (2004), Prediction of mon­ soon rainfall and river discharges on 15-30 day time scales, Bull. Am. Meteorol. Soc, 55(11), 1745-1765. 7

True Color Earth Data Set Includes Seasonal Dynamics PAGES 4 9 , 5 5

River

While false-color visualizations are artifi­ cially colorized from single- or multispectral data, true-color images are b a s e d on data which closely reflect the full spectral range of our visual perception: things in true-color images appear the way we would s e e them. In 2002, the authors of this article created the true-color Earth image on the right side of Figure l.This image consists of separate layers created from the underlying MODIS land, o c e a n , s e a i c e , and atmosphere s c i e n c e products. Both ( 2 0 0 0 and 2 0 0 2 ) Blue Marble images have b e e n widely used in museums, print media, and television documentaries, by mapping agencies, and in NASA's public c o m m u n i c a t i o n s about its missions and research initiatives. The wide public use of the Blue Marble imagery motivated the authors to continue the project. The Blue Marble: Next Genera­ tion (BMNG) is a true-color and normalized difference vegetation index (NDVI) data set that displays land surface state at 500-meter spatial resolution and monthly temporal res­ olution. The BMNG was created using Terra MODIS s c i e n c e data c o l l e c t e d in 2004; cloud disturbances were removed with a discrete Fourier technique. Whereas cloud-free Earth imagery is c o m ­ mercially available, the BMNG aims to pro­

Zielinski,S. (2005),Earth observation programs may still be at risk,£bs Trans.AGU,86(43),414. Author

Information

Faisal Hossain and Nitin Katiyar, Department of Civil and Environmental Engineering, Tennessee Technological University, Cookeville; E-mail: [email protected]

vide freely available imagery as a product c o m p l e m e n t a r y to the standard MODIS sci­ e n c e datasets. Although the spatial resolu­ tion of the BMNG true c o l o r data is c o m p a ­ rable to other data sets, seasonal variations have not b e e n shown before in s e a m l e s s true-color composites. Visualizations of snowfall, droughts, wet seasons, spring green­ ing, and so forth, c a n b e applied in formal and informal education.Visual perception of Earth system dynamics c a n foster interest to further explore the underlying s c i e n c e . Fur­ thermore, the BMNG c a n help to i n c r e a s e public understanding (and therefore a c c e p ­ t a n c e ) of satellite missions and awareness of c a u s e s and effects of c h a n g e s in Earth's cli­ mate system.

How to Create

Cloud-Free

Global

Imagery

S e a m l e s s cloud-free spatial and temporal compositing of the Earth's surface is not a trivial task. It is d e p e n d e n t on sophisticated atmospheric c o r r e c t i o n s (e.g., water vapor, ozone, and aerosol absorption and scatter­ ing [Vermote et at, 1 9 9 7 ] ) and cloud screen­ ing. Even then, cloudy pixels and remote sensing artifacts such as heavy dust and smoke, calibration errors, and illumination conditions [Los et at, 2000] c a n disturb sat­ ellite data. Temporal compositing c a n b e used to remove such irregularities. For the BMNG data set, a temporal adjust­ ment b a s e d on s e c o n d - and third-order dis­ crete Fourier series was used.This m e t h o d is

Eos, Vol. 87, No. 5, 31 January 2006 well suited to create consistent NDVI time series from AVHRR data [Los et al., 2000; Stockli and Vidale, 2004] but has not b e e n applied to true-color imagery until n o w The method assumes that the snow-free land surface c h a n g e s on a seasonal timescale with a yearly periodicity and ignores shortterm disturbances (e.g., clouds, a e r o s o l s ) . Eight-day c o m p o s i t e s of MODIS land sur­ f a c e r e f l e c t a n c e s (MOD09A1) from 2 0 0 4 [Justice et al, 2 0 0 2 ] were used. First, indi­ vidual MODIS b a n d s were quality-checked to identify high aerosol c o n c e n t r a t i o n s , clouds, a n d snow, followed by a snow-free Fourier adjustment. S o m e areas do not show substantial s e a s o n a l c h a n g e ( s u c h as lakes, o c e a n s , and permanently snow cov­ ered a r e a s ) , or do not have sufficient tem­ poral c o v e r a g e due to cloud c o v e r or per­ sistent a e r o s o l s (e.g. tropical rainforests). For t h o s e areas, weighted temporal averag­ ing was applied. As a last step, r e f l e c t a n c e values for snow-covered areas were added to the snow-free data set, and then the data were transformed into true-color imagery and NDVI maps. A t e c h n i c a l and scientific overview of the processing m e t h o d o l o g y is given at http://www.iac.ethz.ch/staff/stockli/ bmng

Resulting

Time

Fig. 1. Earth views: (left) Apollo 17 astronaut photograph (1972); (center) Blue Marble false-color composite (2000), see http://rsd.gsfc.nasa.gov/rsd/bluemarble/; (right) MODIS Blue Marble true-color composite (2002), see http://earthobservatory.nasa.gov/Newsroom/BlueMarble/ BlueMarble_2002.html Original color image appears at the back of this volume.

1 _

1

1

1

I I I

H I

a ) Mixed forests

1

•

1 1

•

II

IHHi

Evergreen broadleaf forest

b)

n

•

1 i

o

« °

„

i + 1

a o z

• 0

Q n

1

"

• %•

0

a

• 0.4

a |

-

•

* f

Series

Figure 2 shows time series of original (MOD09A1, s y m b o l s ) and Fourier-adjusted (BMNG, d a s h e d / s o l i d c u r v e s ) MODIS r e f l e c t a n c e s . T h e Alpine mixed forest (Fig­ ure 2 a ) h a s a long growing s e a s o n . T h e NDVI, which yields high values for lush veg­ etation b e c a u s e of strong near-infrared r e f l e c t a n c e off green plants, is slightly lower in winter, w h e n the d e c i d u o u s part of vege­ tation has lost its leaves. T h e MODIS c l o u d flag (indicating where the MODIS cloudmasking algorithm d e t e c t s clouds; Figure 2, at the top of e a c h graph) captures clouddisturbed pixels well, e x c e p t for o n e at the beginning of O c t o b e r 2004, which is then c o r r e c t e d by our Fourier adjustment algo­ rithm. Significantly heavier cloud c o n t a m i n a ­ tion is s e e n in r e f l e c t a n c e s of a tropical rainforest in Sumatra, Indonesia (Figure 2 b ) . However, the situation there c a n b e over­ c o m e s i n c e the cloud mask in c o m b i n a t i o n with the weighting s c h e m e used allows the r e c o n s t r u c t i o n of the constantly high NDVI time series and a low a l b e d o (reflective p o w e r ) of 2 - 5 % in the visible b a n d s . NDVI slightly d e c r e a s e d after July 2004 during the dry s e a s o n . A few tropical pixels are cloudy during the w h o l e year, which results in c l o u d - c o n t a m i n a t e d BMNG time series. High-altitude vegetation, such as shrubs in the Tibetan Plateau (Figure 2c), only grows dur­ ing a short period, being limited by temperature. Clouds obscure observations during the mon­ soon season, but the short peak in NDVI has been reconstructed by the use of discrete Fou­ rier series. Snow cover changes surface albedo from 10-50% within one eight-day period, as shown in a time series over the western United

1. Wk*4s F

M

A

M

«

J J Month 2004

w A

S

O

N

D

Fig. 2. Yearly time series of uncorrected (symbols) and Fourier-adjusted (lines) land surface reflectances and NDVI: (a) Swiss alpine mixed forest; (b) tropical rainforest in Sumatra, Indonesia; (c) shrubland in the Tibetan Plateau; (d) shrubland in Utah, U.S.

States (Figure 2d).The snow-free Fourier adjust­ ment provides summer reflectance time series, and snow is patched in winter based on a snow flag (where MODIS detects snow; Figure 2, at the top of each graph). However, the beginning and end of snow cover are not clear. The MODIS snow flag shows snow from January to March and at the e n d of December, whereas observed reflectances suggest snow from J a n u a r y to April and during November and D e c e m b e r . Discriminating snow cover from the snowfree reflectances is o n e of the main prob­ lems in the BMNG algorithm.The uncertainty of ±1 month in the timing of snow c o v e r affects the visual a p p e a r a n c e of areas with seasonal snow cover. Global

Maps

of Seasonal

Variations

Such time series were automatically pro­ c e s s e d for the whole planet at 500-meter

spatial resolution.The resulting monthly maps are different from previous cloud-free visualizations of the Earth in that they show seamless seasonal variations on the land sur­ face, with o n e set of maps in true c o l o r and another set as NDVI.The novel atmospheric correction and cloud-masking algorithms used in MODIS land products are important for producing these c o m p o s i t e s . However, while the Fourier adjustment of snow-free reflectance data uses sound assumptions b a s e d on seasonal phenology, compositing snow reflectances on a monthly interval presents a major difficulty in this study. A solution to this problem may b e the avail­ ability of a more sophisticated snow classifi­ cation s c h e m e , the compositing of multiple years of snow fall (snow climatology), or the rejection of short-term (e.g., less than 3 - 4 months) snow cover data. Figure 3 displays the two seasonal opposites, January and July 2004, in 0.5° (available

Eos, Vol. 87, No. 5, 31 January 2006 at 0.004667°) resolution.The algorithm used efficiently removes cloud contamination in tropical areas. Boreal forest albedo in north­ ern latitudes is low during wintertime, since the trees project out of the snow layer, con­ trasted by the very high albedo of short vege­ tation in these areas. The brownish surface during the dry season (January) in northern India and China changes into dark green during the monsoon season (July). These, among many other interesting fea­ tures, c a n b e discovered while exploring the data set, and the images' application in for­ mal and informal education is especially encouraged. NASA is publicly sharing the BMNG data set at n o cost at the following Web site: http://bluemarble.nasa.gov Acknowledgments Special thanks go to the MODIS S c i e n c e Team and the MODIS S c i e n c e Data Support Team for collecting, processing, and providing the MOD09A1 and MOD12Q1 data. System administration by J o h n Hord, Scott Sinno, and Bill Ridgway is much appreciated.This project was funded by NASA contract NAS5-01070, task 4a, as subcontract 2101-03-002, issued by S c i e n c e Systems and Applications, Inc.

References Justice, C. O., J. R. G.Townshend, E. F Vermote, E. Masuoka, R. E.Wolfe, N. Saleous, D. PRoy, and J.T.Morisette (2002), An overview of MODIS Land data processing and product status, demote Sens. Environ., 83,3-15. Los,S.O.,et al. (2000), A global 9-yr biophysical land surface data set from NOAA AVHRR data,./ Hydrometeoroi, 7,183-199. Running, S. W, D. D. Baldocchi, D. PTurner, S.T. Gower, PS. Bakwin, and K. A. Hibbard (1999), A global Atmospheric correction of visible to middle-infra­ terrestrial monitoring network integrating tower red EOS-MODIS data over land surfaces: Back­ fluxes, flask sampling, ecosystem modeling and ground, operational algorithm and validation, Eos satellite data, Remote Sens. Environ., 70, J. Geophys.Res., 102,17,131-17,141. 108-127. Stockli, R.,and PL.Vidale (2004),European plant Author Information phenology and climate as seen in a 20-year AVHRR land-surface parameter data set, Int. Reto Stockli, Institute for Atmospheric and J. Remote Sens., 25(17), 3303-3330. Vermote, E. F, N. El Saleous, C. O. Justice.Y J. Kaufman, Climate Science, Federal Institute of Technology, J.L.Privette,L.Remer,J.C.Roger,and D.Tanre (1997), Zurich, Switzerland; E-mail: [email protected].

nasa.gov; Eric Vermote and Nazmi Saleous, Depart­ ment of Geography, University of Maryland, College Park; Robert Simmon, NASA Earth Observatory, Goddard Space Flight Center, Greenbelt, Md.; David Herring, NASA Headquarters, Washington, D.C.

Eos,Vol. 87, No. 5, 31 January 2006

Fig. 1. Spatial dependency (qualitative) between the occurrence of large-scale floods and the geographical location of International River Basins (IRBs) (compare bottom panels) shown through two connecting themes: rainfall and lack of economic resources (top panels), (top left) A sixyear climatologic rainfall map produced with data from the Tropical Rainfall Measuring Mission (TRMM; http://trmm.gsfc.nasa.gov). (bottom left) Global distribution of floods in 2003 (Source: Brakenridge, G.R., Anderson, E., Caquard, S, 2003, Flood Inundation Map DFO 2003-282, Dartmouth Flood Observatory, Hanover, N.H., digital media, http://www.dartmouth.edu/~floods/2003282.html). (top right) World Bank estimate of GDP per capita of nations in 1999 (lighter shaded countries are poorer), (bottom right): Map of IRBs with the darker shaded IRBs projected to be in a state of water stress due to lack of cooperative agreements for sharing water resources (© Transboundary Freshwater Dispute Database, Oregon State University, 2002).

Page 49

Fig. 1. Earth views: (left) Apollo 17 astronaut photograph (1972); (center) Blue Marble false-color composite (2000), see http://rsd.gsfc.nasa.gov/rsd/bluemarble/; (right) MODIS Blue Marble true-color composite (2002), see http://earthobservatory.nasa.gov/Newsroom/BlueMarble/ BlueMarble_2002.html

Eos, Vol. 87, No. 5, 31 January 2006

Page 55

Fig. 3. Blue Marble: Next Generation—true-color, and (bottom) July 2004.

cloud-free

composites

during (top) January

2004