Derivation of 16-day time-series NDVI data for environmental studies using a data assimilation approach

Figures & data

Table 1. Potential ecological and environmental applications of time-series NDVI data*

Metric name	Environmental and ecological meaning
1. Maximum NDVI	The greenness values of the vegetation at the peak greenness time
2. Mean NDVI	The average photosynthesis activity
3. Value of onset of greenness	Level of photosynthesis at start of the growing season
4. Value of the end of greenness	Level of photosynthesis at end of the brown season
5. NDVI amplitude	An index of the range of photosynthesis
6. Integrated NDVI	Net primary production approximation of a given year
7. Total length of growth season	Consecutive months the NDVI values are greater than onset/end of greenness
8. Fraction of growing season	The number of weeks of greenup divided by total weeks of growing season
9. Rate of greenup	Acceleration of increasing photosynthesis
10. Rate of senescence	Acceleration of decreasing photosynthesis
11. NDVI skewness	Integrated NDVI in greenup divided by integrated NDVI in senescence
12. Time of onset of greenness	Starting of photosynthetic activity
13. Time of end of greenness	End of photosynthetic activity
Note: *Reproduced according to Reed et al. (1994).

Figure 1. A graphical model for one-step forecasting of 16-day time-series NDVI data.

Table 2. Landsat TM images applied to NDVI time-series derivation

	USGS ID	Cloud coverage (%)	Acquisition date	Image ID
1	LT50210292010081EDC00	0	22 March 2010	20100322
2	LT50210292010097EDC00	100	7 April 2010	20100407
3	LT50210292010113EDC00	1	23 April 2010	20100423
4	LT50210292010129EDC00	0	9 May 2010	20100509
5	LT50210292010145EDC00	4	25 May 2010	20100525
6	LT50210292010161EDC00	48	10 June 2010	20100610
7	LT50210292010177EDC00	67	26 June 2010	20100626
8	LT50210292010193EDC00	54	21 July 2010	20100712
9	LT50210292010209EDC00	68	28 July 2010	20100728
10	LT50210292010225EDC00	42	13 August 2010	20100813
11	LT50210292010241EDC00	0	29 August 2010	20100829
12	LT50210292010257EDC00	32	14 September 2010	20100914
13	LT50210292010273EDC00	27	30 September 2010	20100930
14	LT50210292010289EDC00	6	16 October 2010	20101016
15	LT50210292010306EdC00	9	1 November 2010	20101101

Figure 2. TM 20100525: observed versus predicted NDVI image.

Figure 3. TM 20100914: observed versus predicted NDVI image.

Figure 4. The observed 20100829 NDVI image versus predicted 20100829 NDVI that was generated using predicted TM 20100813 and 20100914.

Figure 5. R ² of linear models regressing the forecast on the observed for bands 4, 3, and 2, and NDVI using 300,000 random pixels in each image 20100423, 20100509, and 20100829.

Figure 6. RMSE of linear models regressing the forecasted on the observed using 300,000 random pixels in each image 20100423, 20100509, and 20100829.

Figure 7. Willmott's index of agreement (d from 0 to 1) for assessment of the forecasted values and the observed values using 300,000 random pixels in each image 20100423, 20100509, and 20100829.

Reed , B. C. , Brown , J. F. , VanderZee , D. , Loveland , T. R. , Merchant , J. W. and Ohlen , D. O. 1994 . Measuring Phenological Variability from Satellite Imagery.” . Journal of Vegetation Science , 5 ( 5 ) : 703 – 714 .

 Web of Science ®Google Scholar