J Indian Soc Remote Sens (March 2012) 40(1):19 DOI 10.

1007/s12524-011-0122-2

RESEARCH ARTICLE

Formulation of Time Series Vegetation Index from Indian Geostationary Satellite and Comparison with Global Product
Rahul Nigam & Bimal Kumar Bhattacharya & Keshav R. Gunjal & N. Padmanabhan & N. K. Patel

Received: 24 October 2010 / Accepted: 6 May 2011 / Published online: 25 June 2011 # Indian Society of Remote Sensing 2011

Abstract To study impact of climate change on vegetation time series vegetation index has a vital role to know the behaviour of vegetation dynamics over a time period. INSAT 3A CCD (Charged Couple Device) is the only geostationary sensor to acquire regular coverage of Asia continent at 1 km1 km spatial resolution with high temporal frequency (halfan-hour). A formulation of surface reflectances in red, near infrared (NIR), short wave infrared (SWIR) and NDVI from INSAT 3A CCD has been defined and integrated in the operational chain. The atmospheric correction of at-sensor reflectances using SMAC (Simple Model for Atmospheric Correction) model improved the NDVI by 540% and also increased its dynamic range. The temporal dynamics of 16-day NDVI composite at 0500 GMT for a growing year (June 2008March 2009) showed matching profiles with reference to global products (MODIS TERRA) over known land targets. The root mean square deviation (RMSD) between the two was 0.14 with
R. Nigam (*) : B. K. Bhattacharya : K. R. Gunjal : N. K. Patel Agriculture, Terrestrial Biosphere and Hydrology Group (EPSA) Space Applications Centre (ISRO), Ahmedabad 380 015, India e-mail: rahulnigam@sac.isro.gov.in N. Padmanabhan Data Products Software Group, (SIPA) Space Applications Centre (ISRO), Ahmedabad 380 015, India

correlation coefficient (r) 0.84 from 200 paired datasets. This inter-sensor cross-correlation would help in NDVI calibration to add continuity in long term NDVI database for climate change studies. Keywords NDVI . Geostationary . INSAT . Asia

Introduction The influence of land-use change and landscape dynamics play a vital role in land atmosphere interaction and hence on climate change (Pielke et al. 2002). Vegetation index plays a major role in defining land use change dynamics and is being used in different dynamic vegetation model (Jarlan et al. 2008) embedded in climate models. The vegetation indices are derived by combining spectral data in different wave bands. The spectral indices are typically a sum, difference, ratio or other linear combinations of reflectance factor or radiance observation from two or more wavelength intervals (Wiegand et al. 1991). The time series vegetation index data provide a tool to study past conditions, monitor current conditions (van Leeuwen et al. 2006), and prepare future trends. Comparison of present vegetation records with past long-term averages have been widely used to study ecosystem monitoring, and help to evaluate the impact of rising global temperature and CO2 levels (Nemani et al. 2003). These vegetation indices have been found useful for vegetation density (Wiegand et al. 1979;

J Indian Soc Remote Sens (March 2012) 40(1):19

Spanner et al. 1990), green leaf density, photosynthesis active biomass (Tucker 1979; Wiegand et al. 1991) absorbed photosynthetically active radiation and used in climate modeling. The Normalized Difference Vegetation Index (NDVI) is widely used for vegetation growth monitoring (van Leeuwen et al. 2006). The NDVI is successful as a vegetation measure and it is sufficiently stable to permit meaningful comparisons of seasonal and inter-annual changes in vegetation growth and activity. The strength of the NDVI is in its ratioing concept, which reduces many forms of multiplicative noise (illumination differences, cloud shadows, atmospheric attenuation, certain topographic variations) present in multiple bands (Huete et al. 2002). NDVI was originally used as a measure of green biomass (Tucker et al. 1986). It has a strong theoretical basis as a measure of the fractional absorbed photosynthetic active radiation (fAPAR) by the canopy (Sellers 1985; Fensholt et al. 2004). The NDVI relates reflectance (or radiance) in the red range and in the NIR range to vegetation variables such as leaf area index (LAI) (Wang et al. 2005; Hasegawa et al. 2010), crop cover (Wardlow and Egbert 2008), and vegetation phenology (Zhang et al. 2003; Sakamoto et al. 2005). NDVI has been used to monitor and estimate primary productivity (Field et al. 1995; Matsushita and Tamura 2002) as well as its assimilation in regional numerical weather models give better weather forecasting (Dutta et al. 2009). It is sensitive to low chlorophyll contents, to low fraction of vegetation cover and, as a result, to low level of absorbed photosynthetic active solar radiation. For land surfaces dominated by vegetation, the NDVI values normally range from 0.1 to 0.8 during the growth season, the higher values being associated with greater density of vegetation and greenness of the plant canopy. Atmospheric effects, such as Rayleigh scattering from molecules, Mie scattering by aerosols, gaseous absorption by atmospheric constituents and sub-pixelsized clouds, all tend to increase the value of red with respect to NIR and reduce the values of the computed vegetation indices. The quantitative estimation of most of the biophysical parameters (LAI and fAPAR) using satellite based optical remote sensing requires time series NDVI data as model inputs. Daily and time composite NDVI series from moderate (1 km) to coarse (>1 km) resolution sensor data can provide full resolution of vegetation growth cycle than a single date high resolution optical data with low repeativity.

The normalized difference vegetation index (NDVI) product is now-a-days regularly available from observations in red and near infrared bands in large view global polar orbiting sensors such as: SPOT-VGT, MODIS TERRA and AQUA, NOAA AVHRR at spatial resolutions varying from 250 m to 8 km. These are available maximum twice per day on daily or time composite basis. The NDVI generated at multiple times in a day from geostationary satellite sensors provide opportunity to get more cloud free NDVI as compared to once or twice overpasses in a day by polar orbiting large view sensor (Fensholt et al. 2006). The effects due to orbital drift as in NOAA AVHRR will be least because of constant viewing geometry of geostationary sensors with respect to earth targets. Moreover, the diurnal behaviour of narrow band surface reflectances is ideal to study BRDF (Bi-directional Reflectance Distribution Function) characteristics of similar homogeneous land targets having similar sensor viewing conditions. This again helps further correct NDVI through surface BRDF modeling. No other existing geostationary satellite missions in the world except INSAT 3A CCD (1 km) of India and Feng Yung of China and MSG SEVIRI (3 km) have payloads that take multiple observations per day in narrow spectral optical bands (red, NIR and SWIR) at 1 km X 1 km spatial resolution. In INSAT 3A CCD payload was specifically designed to monitor vegetation and snow cover conditions over Asian region regularly. The preliminary documentation to estimate NDVI from CCD has been done by Bhattacharya and Pandya (2007). The present study was undertaken to do further investigation on atmospheric corrected NDVI with the following objectives: (i) Formulation of INSAT 3A CCD NDVI to make time series vegetation index from Indian geostationary satellite (ii) Comparison of formulated CCD NDVI with global available MODIS NDVI product for different land cover types

Methodology The radiances reaching satellite sensors from earth surface are generally influenced mainly by three factors viz. (i) sun-sensor viewing geometry, (ii) atmospheric noises, adjacency and (iii) BRDF effects.

J Indian Soc Remote Sens (March 2012) 40(1):19

To overcome the effects of these stated factors corrections are needed to make at sensor spectral reflectances (Ri) into surface spectral reflectances. These corrections broadly defined as: Level 1: Generation of angular normalized atmospherically corrected surface spectral reflectances Level 2: Level 1+adjacency effect correction Level 3: Level 2+BRDF correction Level 1 correction has three different components A. Rayleigh scattering B. Gaseous absorption (ozone, water vapour, CO2) C. Aerosol scattering and absorption Post-Launch Vicarious Calibration of Band Radiances The electronic performance of CCD sensor elements generally is degraded due to space weathering. The cross-calibration has been carried out with high resolution (56 m) IRS-P6 AWiFS sensor with equal atmospheric perturbations for red and NIR bands. Three dates spread over December (2008), February (2009) and March (2009) for both INSAT 3A CCD and AWiFS having same overpass time (0500 GMT) were chosen for cross-calibration. Top of Atmosphere (TOA) band radiances from six different land covers such as: agriculture, forest, snow, bare soil, water body and cloud were used for recalibration. Cloud Filtering The optical properties of clouds showed that its reflectances in red, NIR or cloud albedo in broad visible band become high and even more than 90%. But SWIR band reflectances are less in presence of water clouds due to higher absorption. Three criteria were fixed for cirrus (high level), alto (medium level) and cumulus (low level) clouds based on several CCD scenes. The first two criteria are only based on TOA reflectance thresholds in red and NIR bands due to presence of more of ice cloud. In third criteria, SWIR TOA reflectance threshold was introduced in addition to red and NIR reflectances due to increasing presence of water clouds. Further processing was carried out only in cloud free pixels.

Atmospheric Correction The atmospheric noises such as molecular (Rayleigh) and aerosol (Mie) scattering along with gaseous absorption corrected by using simple model for atmospheric correction (SMAC) with wide view sensor coefficients (Rahman and Dedieu 1994). That has been successfully used for large view satellite sensors such as NOAA AVHRR, METEOSAT etc. The generalized functional form of SMAC model is rqs ; q v ; f tg q s; q v fra q s ; qv ; f et=ms tdqs with T q et=m tdq where =s or v * c s v f tg a TOA spectral reflectance at satellite sensor level spectral surface reflectance sun zenith angle view zenith angle relative azimuth between sun and sensor total gaseous transmission atmospheric reflectance which depends on optical properties of air molecule, aerosol, and sun-sensor viewing geometry. atmospheric optical depth (et=ms and et=mv being the direct atmospheric transmission) in sun (s) and view (v) directions atmospheric diffuse transmittances spherical albedo of the atmosphere taken in account multiple scattering between surface and the atmosphere 2 rc et=mv re tdq v g 1 re S 1

td() S 1 re S

This atmospheric correction scheme uses 1st order correction for additive and multiplicative atmospheric noises with the assumption that surface is lambertian. This scheme is simple to implement and tested against 5S atmospheric radiative transfer code (Tanre et al. 1990) and is thus increasingly used for generating surface reflectances from TOA radiances globally. Along with sun-sensor angular

J Indian Soc Remote Sens (March 2012) 40(1):19

geometry, this requires atmospheric inputs such as columnar ozone, precipitable water and aerosol optical depth (AOD) at 0.55 m. The database on daytime mean of five years (20022006) ozone, precipitable water and aerosol optical depth at 0.55 m from MODIS TERRA (0530 GMT) and AQUA (0800 GMT) was prepared through interpolation of MODIS eight-day atmospheric products (1X 1) equivalent to CCD resolution. These were further used as inputs to SMAC model to compute surface reflectances. NDVI Product Formatting The computation of NDVI was carried out in cloud free pixels using surface reflectances in CCD red and NIR bands. The NDVI was scaled to binary (8-bit) format with offset=110 and gain=0.01. The end product contains NDVI in h5 format that contains cloud free NDVI, surface band reflectances as well as files for angular geometry for the geographical bound (44.5105.3E, 9.8S45.5N) known as Asia mercator region.
70000 60000 Frequency 50000 40000 30000 20000 10000 0 -0.2 0.1 0.4 0.7

Satellite Data and Statistical Evaluation To validate INSAT 3A CCD NDVI with available global MODIS TERRA NDVI the data from May 2008 to March 2009 has been used. The pixel of MODIS TERRA within 1 km grid cell were averaged to create one pair of MODIS TERRA and INSAT 3A CCD NDVI. The NDVI values were randomly taken out from different known natural targets like agriculture, forest, desert and snow to cover all type of land cover. The root mean square error (RMSD) and mean absolute deviation (MAD) has been computed by following formulae
RMSD v uP u Pi Oi 2 t
i

P MAD
i

ABSPi Oi N

Where Pi NDVICCD at ith case Oi =NDVIMODIS at ith case N=number of paired datasets
70000 60000 Frequency 50000 40000 30000 20000 10000 0 -0.2 0.1 0.4 0.7

(a)

(b)

Uncorrected NDVI 100000 80000 Frequency 60000 40000 20000 0

-50 -42 -34 -26 -19 -11 -3 5 13 21 29 37 45

Corrected NDVI

(c)

% Change in NDVI

Fig. 1 Comparison of atmospherically corrected NDVI with uncorrected NDVI for a particular day

J Indian Soc Remote Sens (March 2012) 40(1):19

Result and Discussion Effect of Atmospheric Correction on CCD NDVI Atmospherically corrected NDVI has been generated by applying Rayleigh and Mie corrections using average atmospheric conditions through SMAC model. The corrected NDVI was then compared with atmospherically uncorrected NDVI over Indian subcontinent. It was found that for a particular clear day, the NDVI range increased from 0.2 to 0.6 in uncorrected one to 0.2 to 0.7 atmospherically corrected one. The frequency distribution of uncorrected, corrected NDVI and percent difference between them are shown in Fig. 1a, b and c, respectively. Through respective difference of corrected and uncorrected NDVI stated above ranged from 25 to 45, but majority of pixels under different land cover types showed positive difference between 5 to 40%.

Analysis of Spatio-Temporal Dynamics of CCD NDVI The 16-day NDVI composite has been computed from daily NDVI to analyze the spatio-temporal dynamics of NDVI for a growing season over different land cover categories from June 2008 to March 2009. In agriculture, NDVI showed quite high dynamics as compared to desert and forest, respectively. From the Fig. 2, it is quite evident that in Indo-gangetic plain NDVI showed high dynamics of intensive agricultural activities from June 2008 to March 2009. In Indo-gangetic plain, overall spatial NDVI was low during June but it increased from July due to increase of vegetation cover with progression of monsoon rainfall over Indian landmass. The NDVI shows decreasing trend during October with the maturity of kharif crops but again shows increasing trend in November and peak in February due to progress of rabi crops.

Fig. 2 Spatio-temporal dynamics of CCD NDVI during a growing season

J Indian Soc Remote Sens (March 2012) 40(1):19

Validation For the validation purpose, 16 day NDVI composite at 0500 GMT satellite data were used from 26th May 2008 to 30th March 2009 to cover dynamic range of land cover type. The 16-day composite were prepared by extracting maximum NDVI value for each pixel from daily 16-day NDVI data. This will minimize the cloud interference on NDVI and to capture the phenological shift of vegetation. The MODIS TERRA cloud free NDVI were available as 16 day composite at 250 m spatial resolution. These were then linearly aggregated to target CCD grid resolution. The CCD NDVI and aggregated MODIS TERRA NDVI were
Fig. 3 Comparison of temporal profiles of CCD and TERRA. Comparison for different land cover types

extracted over different land targets such as agriculture (a), forest (b) and desert (c). The temporal evolution of NDVI and rate of change of slope of the curve matches well throughout growing year with MODIS TERRA NDVI (Fig. 3). Temporal profile over agricultural target in Punjab (agriculture) typically showed two peaks corresponding to growth of Kharif rice and winter wheat. In case of desert, NDVI values from both CCD and TERRA showed little change between 0.05 to 0.2 except small peak during south west monsoon season. The NDVI profiles from CCD and MODIS TERRA showed similar pattern over forest target. The spikes in the monsoon months could be due to difference in detection technology in
MODIS TERRA INSAT 3A CCD

(a)
1 0.8 NDVI 0.6 0.4 0.2 0

Agriculture : Punjab (30.50 N, 76.40 E)

10 11 12 13 14 15

Number of 16 days interval starting from 1 June 2008

(b)
1 0.8 NDVI 0.6 0.4 0.2 0

Rajasthan : Desert (26.60 N, 70.90 E)

MODIS TERRA INSAT 3A CCD

10 11 12 13 14 15

Number of 16 days interval starting from 1 June 2008

(c)

Evergreen Forest : Western Ghat (18.50 N, 73.10 E) 1

MODIS TERRA INSAT 3A CCD

0.8

NDVI

0.6 0.4 0.2 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Number of 16 days interval starting from 1 June 2008

J Indian Soc Remote Sens (March 2012) 40(1):19

cloud contaminated pixels for two sensors. In MODIS both optical and thermal bands used for cloud screening, but only optical band data is used for cloud screening in CCD. The overall RMSD of 0.14 with a correlation of 0.84 (n=200) was obtained as shown in Fig. 4. MODIS show overall high NDVI value over INSAT 3A CCD. This may due to difference in spectral and spatial resolution of sensor itself along with difference in atmospheric correction schemes. Miura et al. (2008) also reported high MODIS NDVI as compared to ASTER derived NDVI with mean difference of 0.031 from same (TERRA) platform having identical suntarget-view geometry for both the cases. In General, MODIS bands are much narrower in spectral bandwidths than ASTER red and NIR bands. Likewise, the central wavelengths from two sensors differ. MODIS red band completely avoids the red edge region (~ 680 nm), the ASTER counterpart extends to cover that wavelength. The MODIS NIR band overlaps at the longest wavelength portion of the ASTER counterpart. These are the consequences of the MODIS band selection requirements to avoid Fraunhofer lines and atmospheric absorption lines. Similarly, MODIS TERRA NDVI showed high NDVI with mean bias (NDVIMODIS NDVICCD) of 0.07 as compared to CCD NDVI. In addition CCD NDVI computation does not explicitly consider complex modeling of surface BRDF and adjacency effects as incorporated in MODIS. Geostationary meteosat second generation (MSG) SEVIRI derived cloud free daily averaged NDVI (3 km) compared with resampled daily MODIS TERRA/ AQUA NDVI (250 m) by Fensholt et al. (2006). They
1
r = 0.83, n = 200 MAD = 0.10 RMSD = 0.14

showed fairly good agreement in the dynamic range with a tendency to little higher MSG NDVI in the beginning of the growing season (July-August) and lower towards the end (October-November). This has been attributed to seasonal variation in solar azimuth angle, which influence observations from geostationary platform as compared to those from polar orbiting platform having lesser swath. Therefore, present comparison of CCD NDVI with MODIS TERRA shows good agreement in connection with earlier findings. NDVI product from INSAT 3A CCD has been derived at 30 min interval and due to changing sun zenith and azimuth angles with reference to target, red and NIR reflectance get changed. It has been reflected in diurnal pattern of NDVI for different targets as shown in Fig. 5. NDVI experienced minimum change

0.3 0.2 0.1 0

Vegetation

830 930 1030 1130 1230 1330 1430 1530

Local time(hrs)

Himalayan Evergreen Forest

0.5

NDVI

0.4

0.3 830 930 1030 1130 1230 1330 1430 1530

INSAT 3A CCD NDVI

0.8

Local time(hrs)
0.3 0.2 0.1 0

0.6

Desert

0.4

0.2

0 0 0.2 0.4 0.6 0.8 1 MODIS TERRA NDVI

830 930 1030 1130 1230 1330 1430 1530

Local time(hrs)
Fig. 5 Diurnal (1 hr) variation of NDVI from INSAT- 3A CCD (9th May10)

Fig. 4 Pooled comparison of INSAT 3A CCD NDVI with MODIS TERRA NDVI

J Indian Soc Remote Sens (March 2012) 40(1):19 Director, SAC and Dr. J.S. Parihar, Deputy Director EPSA, SAC for their timely help during this study. The authors are grateful to Dr Sushma Panigrahy, Group Director, Agriculture, Terrestrial Biosphere and Hydrology Group for her valuable suggestions while carrying out the analysis.

during 11:30 to 14:30 LMT (Local Mean Time). This is because the daily NDVIs are generated primarily from data gathered around midday. Further studies from INSAT 3A CCD used this time period for composting of NDVI under different time domain. The similar diurnal NDVI pattern has been reported by Proud et al. (2010) for Meteosat second generation using similar (SMAC) atmospheric correction scheme.

References
Bhattacharya, B. K., & Pandya, M. R. (2007). INSAT geophysical parameter retrieval system algorithm theoretical basis definition (pp. 269284). Ahmedabad: Space Applications Centre, ISRO. Dutta, S. K., Das, S., Kar, S., Mohanty, U. C., & Joshi, P. C. (2009). Impact of vegetation on the simulation of seasonal monsoon rainfall over the Indian subcontinent using a regional model. Journal of Earth System Science, 118, 413440. Fensholt, R., Sandholt, I., & Rasmussen, M. S. (2004). Evaluation of MODIS LAI, fAPAR and the relation between fAPAR and NDVI in a semi-arid environment using in situ measurements. Remote Sensing of Environment, 91, 490507. Fensholt, R. I., Sandholt, I., Stisen, S., & Tucker, C. (2006). Analysing NDVI for the African continent using the geostationary meteosat second generation SEVRI sensor. Remote Sensing of Environment, 101, 212229. Field, C. B., Randerson, J. T., & Malmstrom, C. M. (1995). Global net primary production: combining ecology and remote sensing. Remote Sensing of Environment, 51, 74 88. Hasegawa, K., Matsuyama, H., Tsuzuki, H., & Sweda, T. (2010). Improving the estimation of leaf area index by using remotely sensed NDVI with BRDF signatures. Remote Sensing of Environment, 114, 514519. Huete, A., Didan, K., Miura, T., Rodriguez, E. P., Gao, X., & Ferreira, L. G. (2002). Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sensing of Environment, 83, 195213. Jarlan, L., Mangiarott, S., Mougin, E., Mazzega, P., Hiernaux, P., & Le Dantec, V. (2008). Assimilation of SPOT/ VEGETATION NDVI data into a sahelian vegetation P. dynamics model. Assimilation of SPOT/VEGETATION NDVI data into a sahelian vegetation dynamics model. Remote Sensing of Environment, 112, 13811394. Matsushita, B., & Tamura, M. (2002). Integrating remotely sensed data with an ecosystem model to estimate net primary productivity in East Asia. Remote Sensing of Environment, 81, 5866. Miura, T., Yoshioka, H., Fujiwara, K., & Yamamoto, H. (2008). Inter-comparison of ASTER and MODIS surface reflectance and vegetation index products for synergistic applications to natural recourses monitoring. Sensors, 8, 24802499. Nemani, R. R., Keeling, C. D., Hashimoto, H., Jolly, W. M., Piper, S. C., Tucker, C. J., et al. (2003). Climate-driven increases in global terrestrial net primary production from 1982 to 1999. Science, 300, 15601563. Pielke, R. A., Marland, G., Betts, R. A., Chase, T. N., Eastman, J. L., Niles, J. O., et al. (2002). The influence of land-use

Conclusions The surface reflectances in red and NIR were used to compute NDVI by applying simple model for atmospheric correction. The atmospherically corrected INSAT 3A CCD NDVI was further evaluated and validated with global product to evaluate its spatio-temporal profiles and its range over different natural targets. This study suggests that atmospheric corrected CCD temporal NDVI profile follow the same trend as globally available NDVI products in a growing year for different vegetation systems and land cover types. The CCD NDVI showed fair good correlation with global available MODIS TERRA NDVI product. The prospects appear good for future efforts to reprocess CCD data sets with a goal of continuity generating of NDVI product to study land surface changes with time. The NDVI generated at continental scale at high temporal resolution through INSAT 3A CCD has made a way to produce NDVI at global level by clubbing different geostationary satellite data. With the availability of time series operational INSAT 3A CCD NDVI products, an accurate quantification of radiation and rainfall interception by canopy and their related response can be further evaluated. Moreover, the vegetation fraction, a derivative of NDVI, can also be assimilated into land surface process module in weather forecasting schemes. These have a great relevance to ecological, hydological as well as meteorological applications. In addition, long term NDVI of the past, present and future can be generated through inter-sensor calibration for climate change studies.
Acknowledgments The authors are extremely grateful to Satellite Meteorology Division, IMD, New Delhi for providing the time series INSAT 3A CCD data from IMDPS in the purview of IMD-ISRO project. The authors would like to thank Dr. R.R. Navalgund, Director, Space Applications Centre (SAC), ISRO for his support given for this study. The authors would also like to thank Shri. A.S. Kiran kumar, Associate

J Indian Soc Remote Sens (March 2012) 40(1):19 change and landscape dynamics on the climate system: relevance to climate-change policy beyond the radiative effect of greenhouse gases. Philosophical Transactions of the Royal Society A, 360, 17051719. Proud, S. R., Rasmussen, M. O., Fensholt, R., Sandholt, I., Shisanya, C., & Mutero, W. (2010) Improving the SMAC atmospheric correction code by analysis of Meteosat Second Generation NDVI and surface reflectance data. Remote Sensing of Environment, 114, 16871698 Rahman, H., & Dedieu, G. (1994). SMAC: a simplified method for the atmospheric correction of satellite measurements in the solar spectrum. International Journal of Remote Sensing, 15, 123143. Sakamoto, T., Yokozawa, M., Toritani, H., Shibayama, M., Ishitsuka, N., & Ohno, H. (2005). A crop phenology detection method using time-series MODIS data. Remote Sensing of Environment, 96, 366374. Sellers, P. J. (1985). Canopy reflectance, photosynthesis and transpiration. International Journal of Remote Sensing, 6, 13551372. Spanner, M. A., Pierce, L. L., Running, S. W., & Peterson, D. L. (1990). The seasonality of AVHRR data of temperate coniferous forests: relationship with leaf area index. Remote Sensing of Environment, 33, 97112. Tanre, D., Deroo, C., Duhaut, P., Herman, M., & Morcrette, J. J. (1990). Description of a computer code to simulate the satellite signal in the solar spectrum-the 5S code. International Journal of Remote Sensing, 11, 659668.

9 Tucker, C. J. (1979). Red and photographic infrared linear combinations for monitoring vegetaion. Remote Sensing of Environment, 8, 127150. Tucker, C. J., Justice, C. O., & Prince, S. D. (1986). Monitoring the grasslands of the Sahel 19841985. International Journal of Remote Sensing, 7, 15711571. van Leeuwen, W. J. D., Orr, B. J., Marsh, S. E., & Herrmann, S. M. (2006). Multi-sensor NDVI data continuity: uncertainties and implications for vegetation monitoring applications. Remote Sensing of Environment, 100, 6781. Wang, Q., Adikua, S., Tenhunena, J., & Granier, A. (2005). On the relationship of NDVI with leaf area index in a deciduous forest site. Remote Sensing of Environment, 94, 244255. Wardlow, B. D., & Egbert, S. L. (2008). Large-area crop mapping using time-series MODIS 250 m NDVI data: an assessment for the U.S. Central Great Plains. Remote Sensing of Environment, 112, 10961116. Wiegand, C. L., Richardson, A. J., & Kanemasu, E. T. (1979). Leaf area index estimated for wheat from Landsat and their implications for evapotranspiration and crop modeling. Agronomy Journal, 71, 336342. Wiegand, C. L., Richardson, A. J., Escobar, D. E., & Gerberman, A. H. (1991). Vegetation indices in crop assessments. Remote Sensing of Environment, 35, 105119. Zhang, X., Friedl, M. A., Schaaf, C. B., Strahler, A. H., Hodges, J. C. F., Gao, F., et al. (2003). Monitoring vegetation phenology using MODIS. Remote Sensing of Environment, 84, 471475.