GISdevelopment.net ---> AARS ---> ACRS 1992 ---> Poster Session R

Intercept Photosynthetically Active Radiation Estimated By Spectral

Jin Zhonghui, Zhang Hongming, Wang Jiasheng, Huang Xinshou
Beijing Agricultural University, Beijing, China


Abstract :
This paper introduce the theoretical formulas and the experimental formulas between the leaf-area index and intercepted photosynthetically active radiation PAR by wheat. Intercepted photosynthetically active radiation are calculated by measuring the leaf-area index of wheat. Then, to find out separately various spectral vegetation indices and the linear regression formulas between various spectral vegetation indexes and intercepted photosynthetically active radiation by wheat. It came to the conclusion that it is the most reliable method to estimate intercepted photosynthetically active radiation by using the spectral vegetation index PVI and the linear regression formula PAR = 78.6 PVI + 55.2.

1. Theoretical Fundamentals and Methods
It is often necessary to estimate intercept photosynthetically active radiation PAR by crop at inspecting crop growth and forecasting crop yield models. Having used the guantum sensor, Hipps et al measured intercept PAR by wheat canopy at different growth stages and measured the leaf-area index LAI of wheat at the same time. They got the function relation between PAR and LAI of wheat canopy before and after heading.

Before heading PAR = 93.5 1.0 – exp (-0.9 LAI) (1)
After heading PAR = 93.5 1.0 – 0.2 exp (-0.95 LAI) (2)

Monsi et al represents the function relation between interception of light rays P and LAI of plant canopy, as

P = 1 – e-k’ LAI (3)

Where k’, the angular shape coefficient, depends exclusively on the leaf angle distribution. In this paper, if the leaves are horizontal k’ is given as

K’ = 1 (4)

For uniform canopies with spherical leaf angle distribution k’ is given by

K’ = 0.5 / cos h (5)

Where c is the sun zenith angle. Because the scattering is low in the PAR region, the intercept radiation and the radiant absorption by leaves have a similar value. The P in (3) may be taken as the absorbed photosynthic radiation by leaves. (3) can be rewritten as

PAR = 1 – e-k’ LAI (6)

Mang researches has fund the correlation between spectral vegetation indices and biological variable of crop (for example leaf-area index, yield and so on). The most often using spectral vegetation indices are given as

Ratio vegetation index RVI = dhdg (7)

Normalized difference vegetation index ND = dh - dg / dh + dg (8)

Perpendicular vegetation index PVI = 0.939Jn – 0.344Jr + 0.09 (9) Where dhanddg area separately the spectral reflectance of the plant canopy at near-infrared and red wave bands. In this paper, dhanddg are taken as the spectral reflectance of TM4 (0.76 – 0.90mm) and TM3 (0.63 – 0.69mm) wave bands.

From (1) to (6) are utilized to find out intercept PAR by wheat through measuring the leaf-area index of what canopy, then we find out various spectral vegetation indices by measuring spectral reflectance of wheat canopy, finally we find our separately the linear regression formula between various spectral vegetation indices and PAR. It will be seen from above cumulative results that which spectral vegetation index will be used to estimate intercept PAR by wheat canopy. It will be the most reliable.

2. Experimental methods
Eight experimental plots of winter sheet are at test field of Beijing Agriculture University. Each plot was planted in north-south and sowed in 29.9.1989. The spectral reflectance measurements were made repeatedly eight item in the growth stages of wheat. The measurement are randomly carried out on nine sample points of each plot. The mean value of the spectral reflectance are taken as calculative values. The spectral reflectance measurements were made with a 15° filed – of – view eight channels radiometer made in BARNES company in America. The radiometer was hand-held at a height of 1.0 m or 1.5 m above the height of wheat and the sun was at one elevation of 35. We used the grey plate made in AN HUI optic machine institute as standard in measuring spectral reflectance. The spectral reflectance of two wave bands (TM3 and TM4) only are used in this paper calculation, though we have got the spectral reflectance of eight wave bands. We took twenty wheat plants from each test plot to measure area of each leaf. Then wheat plants from each test plot to measure are of each leaf. Then the mean leaf-area of each plant times the gross plants of each mu. Finally, the gross leaf-area of each mu are divided by the area of each mu gives the leaf-area index LAI.

3. Experiment date and calculative Results
The spectral reflectance of two wave bands (TM3 and TM4) and LAI values at different growth stages of wheat are listed in table 1. PAR values calculated from LAI values are listed in table 2, where A stands for PAR values calculated from (1) or (2), B.C.D stand for PAR values calculated from (5) and (6) when c are separately taken as 20°, 35°, 50°, E stands for values calculated from (6) when k’ is taken as 1. The spectral vegetation indices at different growth are listed in table 3. The linear regression formulas between PAR and x (spectral vegetation index) and their correlation coefficients and residual standard errors are listed in table 4.

Table 1. The spectral reflectance and LAI of the winter wheat at different growth stages.
MeasuringDate PlotsValue 1 2 3 4 5 6 7 8
9 0
3
3 0
TM3
TM4
LAI
3.87
29.0
2.01
3.43
33.57
2.08
4.59
26.74
1.83
3.56
34.45
1.89
3.62
33.99
2.57
4.08
30.15
1.78
4.66
27.81
1.52
4.36
29.69
1.81
9 0
4
8
TM3
TM4
LAI
5.76
34.99
2.95
5.18
37.15
2.91
6.26
37.49
1.99
5.35
36.13
3.08
4.86
40.59
4.00
5.50
37.60
2.65
6.50
35.60
2.25
5.05
37.91
2.63
9 0
4
2 3
TM3
TM4
LAI
1.39
32.76
4.72
1.35
35.68
5.39
1.45
27.96
4.41
1.44
35.14
5.02
1.80
43.41
6.65
1.72
30.97
3.41
1.57
33.65
2.86
1.50
33.65
6.41
9 0
5
4
TM3
TM4
LAI
1.54
39.46
5.79
1.62
45.93
5.97
1.62
40.82
6.01
1.49
42.60
6.04
1.62
46.34
6.84
1.83
44.92
5.72
1.51
40.23
4.78
1.58
40.69
6.59
9 0
5
1 3
TM3
TM4
LAI
1.85
42.06
6.87
1.91
44.58
8.60
1.89
39.89
6.24
2.01
44.40
6.95
1.97
45.98
8.90
2.03
44.78
6.45
1.85
41.83
6.15
2.03
43.96
8.94
9 0
6
5
TM3
TM4
LAI
4.67
38.71
4.25
4.26
35.67
5.00
4.82
34.58
3.31
5.59
35.80
2.55
6.03
34.38
2.55
5.12
36.58
2.72
5.00
36.77
2.33
5.46
35.19
2.84
9 0
6
1 2
TM3
TM4
LAI
9.67
28.64
1.26
8.78
26.69
1.39
9.31
25.14
0.16
12.73
27.39
0.22
12.73
27.39
0.22
10.43
26.16
0.08
10.74
23.21
0.16
11.74
25.44
0.18

Table 2 : Calculating PAR value by LAI under five model
PAR
Date
Plots
%
1 2 3 4 5 6 7 8
9 0

3

3 0
A
B
C
D
E
78.18
65.67
70.67
79.05
86.60
79.12
66.93
71.90
80.16
87.50
75.49
62.22
67.26
75.91
83.95
76.43
63.41
68.44
76.46
84.89
84.24
74.21
79.16
86.45
92.34
74.66
61.20
66.25
74.65
83.13
69.69
55.45
60.45
69.34
78.12
75.16
61.82
66.86
75.53
83.63
9 0

4

8
A
B
C
D
E
86.92
79.18
83.47
89.91
92.97
86.68
78.73
83.06
89.60
92.71
77.90
65.30
70.31
78.72
83.32
87.65
80.57
84.3
90.88
95.40
90.94
88.09
91.29
95.54
98.16
84.89
75.58
80.15
87.27
92.93
81.15
69.98
74.67
82.62
89.46
84.73
75.32
79.91
87.07
92.79
9 0

4

2 3
A
B
C
D
E
92.16
91.88
94.38
97.45
99.10
92.76
94.31
96.27
98.48
99.54
91.73
90.42
93.22
96.76
98.78
92.48
94.07
95.32
97.98
99.34
93.26
97.09
98.27
99.43
99.87
89.15
83.70
87.52
92.95
96.69
86.37
78.16
82.54
89.18
94.27
93.20
96.70
97.99
99.31
99.83
90

5

4
A
B
C
D
E
92.99
95.40
97.08
98.89
99.69
93.06
95.83
97.38
99.03
99.74
93.08
95.91
97.44
99.06
99.75
93.09
95.97
97.49
99.08
99.76
93.30
97.37
98.46
99.51
99.89
92.95
95.23
96.95
98.83
99.67
92.23
92.13
94.59
97.57
99.16
99.73
96.99
98.20
99.40
99.86
9 0

5

1 3
A
B
C
D
E
93.30
97.41
98.48
99.52
99.89
93.45
98.96
99.47
99.87
99.98
93.15
96.38
97.78
99.21
99.90
93.32
97.52
98.56
99.55
99.90
93.46
99.12
99.56
99.90
99.99
93.21
96.76
98.04
99.33
99.84
93.13
96.20
97.65
99.16
99.78
93.47
99.14
99.57
99.90
99.99
9 0

6

5
A
B
C
D
E
93.09
89.25
92.52
96.33
98.57
93.29
93.00
95.27
97.95
99.32
92.55
82.81
86.73
92.38
96.34
91.61
74.24
78.90
86.23
92.19
92.78
85.42
89.03
94.01
97.32
9.188
76.47
80.98
87.94
93.41
91.20
71.04
75.87
83.67
90.27
92.04
77.92
82.92
89.01
94.15
9 0

6

1 2
A
B
C
D
E
87.85
48.84
53.65
62.46
71.63
88.54
52.26
57.18
66.07
75.09
77.43
8.15
9.30
11.70
14.78
78.32
11.04
12.56
15.72
19.74
75.49
2.10
2.41
3.06
3.92
76.16
4.16
4.76
6.03
7.68
77.43
8.15
9.30
11.70
14.78
77.74
9.13
10.40
13.06
19.72

Table 3. The spectral vegetation indices of winter wheat at different growth stages
Date
Plots
Indices
1 2 3 4 5 6 7 8
9 0
3
3 0
RV1
ND
PV1
7.49
0.764
0.3490
9.78
0.814
0.3934
5.82
0.706
0.3253
9.67
0.812
0.4012
9.39
0.807
0.3967
7.39
0.761
0.3590
5.96
0.712
0.3351
6.81
0.744
0.3537
9 0
4
8
RV1
ND
PV1
6.07
0.717
0.3987
7.17
0.755
0.4210
5.98
0.713
0.4204
6.75
0.742
0.4108
8.67
0.793
0.4550
6.83
0.744
0.4241
5.47
0.690
0.4019
7.50
0.764
0.4286
9 0
4
2 3
RV1
ND
PV1
23.56
0.918
0.3928
26.42
0.927
0.4204`
19.28
0.901
0.3475
2.40
0.921
0.4150
24.06
0.920
0.4905
18.00
0.894
0.3749
21.43
0.910
0.4005
22.47
0.914
0.4008
9 0
5
4
RV1
ND
PV1
25.62
0.924
0.4552
28.35
0.931
0.5157
25.19
0.923
0.4677
28.59
0.932
0.4849
28.60
0.932
0.5204
24.55
0.921
0.5055
26.64
0.927
0.4625
25.75
0.925
0.4666
9 0
5
1 3
RV1
ND
PV1
21.65
0.911
0.4785
23.34
0.917
0.5020
21.10
0.909
0.4580
22.09
0.913
0.4996
23.34
0.917
0.5149
22.06
0.913
0.5035
22.61
0.915
0.4764
21.65
0.911
0.4958
9 0
6
5
RV1
ND
PV1
7.92
0.775
0.4371
8.37
0.786
0.4102
7.17
0.755
0.3981
6.40
0.729
0.4069
5.70
0.701
0.3920
7.14
0.754
0.4158
7.35
0.760
0.4180
6.44
0.731
0.4016
9 0
6
1 2
RV1
ND
PV1
2.96
0.495
0.3256
3.03
0.503
0.3097
1.24
0.107
0.2940
2.15
0.365
0.3034
2.42
0.415
0.2826
2.50
0.428
0.2997
2.25
0.384
0.2724
2.16
0.367
0.2885

Table 4 The linear regression formulas between PAR and the spectral vegetation indices, correlation coefficients and residual standard errors
Date Regression model PAR (%)( =ax +b) r Se %
X = RV1 A
B
C
D
E
0.537x + 80.2
2.19x + 44.9
2.09x + 49.1
1.89 x + 56.3
1.67x + 62.8
0.680
0.742
0.711
0.649
0.589
5.33
18.4
19.3
20.6
21.4\3
X = ND A
B
C
D
E
25.1x + 68.1
138x – 31.5
137x – 28.0
133x – 20.8
126x – 12.2
0.641
0.932
0.929
0.915
0.893
5.58
9.96
10.1
10.9
11.8
X = PV1 A
B
C
D
E
78.6x + 55.2
348x – 68.0
340x – 61.9
320x – 49.5
295x – 35.5
0.731
0.856
0.836
0.800
0.756
4.96
14.2
14.9
16.2
17.2

4. Analyses and Conclusions
Above calculation results pointed out that all correlation coefficients of 15 linear regression formulas are more than 0.354, it is a critical correlation coefficient value when the sample number is 50 and confidence is 1%, however the sample number is 56 I our experiment. It is clear that there is really a correlation between PAR and three spectral vegetation indices introduced in this paper. There are more smaller residual standard errors in three formulas relating to a condition. Their se are separately about 5%. The residual standard errors se of other twelve formulas relating B.C.D.E. condition are separately more than 10% and even upto 20%. When Se value of the regression formula is more than 10.11%, it is not difficult for us to show that PAR value will be over 100% at same x values (spectral vegetation index). It is obvious that this condition are contrary to physical facts, so twelve linear regression formulas relating to B.C.D.E must be given up. The above analysis showed that three linear regression formuls relating to A condition can be useful. It also showed the function relation (1) and (2) between PAR and LAI suggested by Hipps et al are very successful in out experiment. Which one is more better among the three linear regression formulas relating to A condition ? We have made it clear that using PVI is the best because of its relating to the largest correlation coefficient and the least residual standard error Se. F test are carried out to the three linear regression formulas too. The calculating results also showed that using PVI has the best effect, because it has the largest F value and greatly exceed in the critical F value (n-4, a-1) as given by table 5. In short, this research showed that by using PVI and the learn regression formula.

Table 5.F value of the linear regression formulas under the F test
Indices F value Critical F value (n=40, a=1%)
RV1 46.6 7.31
ND 37.7 7.31
PV1 62.1 7.31

Reference
  • Hipps.L.E.Asar, G & Kanemasu, E.T. (1983) Argric Meteorol 38: 253-259
  • M/Monsi & T. Sacki : Jpn.J.Bot 14: 22-52. 1953
  • J.L. Halfield : Remote Sensing of Environment. 1 984, 14:65 – 75
  • J.L. Halfield : Int. J. Remote Sensing 1985 Vol 6. No. 1 167 – 175
  • Zhang Hongming, Wang, Jiasheng : Crop Remote Sensing Monitoring and Estimating yield. Beijing Agricultural University Publishing House, 1989.