A study of characteristic between vegetation coverage and spectral reflectance of paddy field based on time series observation data

A STUDY OF CHARACTERISTICS BETWEEN VEGETATION COVERAGE AND SPECTRAL REFLECTANCE OF PADDY FIELD BASED ON TIME SERIES OBSERVATION DATA July, 2014 Supervisor: Professor Rikimaru Atsushi

A STUDY OF CHARACTERISTICS BETWEEN VEGETATION COVERAGE AND SPECTRAL REFLECTANCE OF PADDY FIELD BASED ON TIME SERIES OBSERVATION DATA

July, 2014

Supervisor: Professor Rikimaru Atsushi

Research Associate Sakata Kenta

Nagaoka University of Technology, Graduate School of Engineering,

Environmental systems Engineering

Phan Thi Anh Thu

LIST OF FIGURE v

LIST OF TABLES viii

ABSTRACT x

Chapter 1: INTRODUCTION 1

1.1 Introduction 1

1.2 Study area 2

1.3 Purpose 3

1.4 Research flow chart 3

1.5 Contents 4

Chapter 2: THE SPECTRAL REFLECTANCE CURVE AND RICE DEVELOPMENT STAGE 5

2.1 The reflectance and spectral reflectance curve 5

2.1.1 The reflectance 5

2.1.2 Spectral reflectance curve 6

2.2 Rice growth phase 8

2.2.1 Vegetative phase 10

2.2.2 Reproductive phase 10

2.2.3 Ripening phase 10

Chapter 3: EXPERIMENTAL STUDY AND GROUND DATA 12

3.1­ Experimental study 12

3.1.1 Equipment 14

3.1.2 Experiment specification 14

3.2­Experimental data processing 16

3.2.1 Calculate vegetation coverage from photos taken in paddy field 16

3.2.2 Calculate reflectance from spectral data 20

Chapter 4: THE CHARACTERISTICS BETWEEN VEGETATION COVERAGE AND SPECTRAL REFLECTANCE 33

4.1 The relationship between vegetation coverage and physical parameters of rice 33

4.2 The characteristics between vegetation coverage and spectral reflectance 38

4.2.1 The characteristics between spectral reflectance and rice growth (DAT) 38

4.2.2 The characteristics between spectral reflectance and vegetation coverage 43

4.2.3 The effects of cultivation condition 49

4.3 The relationship between reflectance of paddy fields and rice plant 50

Chapter 5: APPLYING TO LANDSAT IMAGES 53

5.1 Methodology of reflectance converting from Landsat ETM+ data 53

5.1.1 Digital number (DN) to Radiance 53

5.1.2 Radiance to ToA Reflectance 54

5.2 Calculate the reflectance of Landsat ETM + images acquired during experimental study period 54

5.3 Applying to satellite images 58

5.3.1 Spectral reflectance calculated from Landsat 7 ETM+ images 58

5.3.2 Field spectral measurements and satellite spectral data 59

5.3.3 Vegetation coverage estimated from satellite reflectance 62

Chapter 6: RESULTS AND CONCLUSION 64

6.1 Results 64

6.2 Conclusion 66

LIST OF FIGURE

Fig 1.1 Study area 2

Fig 1.2 Research flow chart 4

Fig 2.1 Illustration of the relation between radiance received at a sensor (Lsensor ), the radiance reflected from a surface (Lsurface ) and the incoming radiances at the top of the atmosphere (Ltoa) and at the surface (Lincoming) 5

Fig 2.2 Spectral reflectance properties of soil, vegetation and water, and spectral bands of LANDSAT 7( http://www.seos­project.eu/modules/remotesensing/remotesensing­c01­ p05.html) 7

Fig 2.3 Rice growth phases 8

Fig 3.1 The change of paddy field during rice development season 12

Fig 3.1 Obtaining the ground field of view 15

Fig 3.2 Experiment specification 15

Fig 3.3 Classified image processing 19

Fig 3.4 The photos taken in paddy field were affected by weather conditions 19

Fig 3.5 Vegetation coverage change during development time 20

Fig 3.6 Spectral data acquisition 21

Fig 3.7 Calculate reflectance from spectral data procedure 22

Fig 3.8 SL2 Mercury Argon Calibration Lamp 23

Fig 3.9 The linear regressions between wavelength and pixel number 25

Fig 3.10 Radiometric reference sources and color filters 26

Fig 3.11 The intensity calibration coefficient 27

Fig 3.12 The intensity of slave channel after calibrating intensity 27

Fig 3.13 The tube was covered by a white paper on the top 28

Fig 3.14 These tubes are used to test the transmittance of white paper 29

Fig 3.16 The reflectane curve of paddy field A on July 30 30

Fig 3.17 The reflectane curves of paddy field A 31

Fig 3.18 The reflectane curves of paddy field B 32

Fig 4.1 The increase in tiller number 34

Fig 4.2 The increase in height of rice plant 35

Fig 4.3 The correlation of original and data calculated by using function VC 37

Fig 4.4 Photo of paddy field 40

Fig 4.5 The reflectance of paddy field from Jun to Sep, 2013 40

Fig 4.6 The change of the ratio NIR/R over time 41

Fig 4.7 The change of NDVI during rice development season 41

Fig 4.8 The change of the subtract of NIR and R band during rice development season 42 Fig 4.9 The relationship between reflectance and vegetation coverage in visible band 44

Fig 4.10 The relationship between reflectance and vegetation coverage in NIR band 45

Fig 4.11 The relationship between reflectance and NIR/R 45

Fig 4.12 The relationship between reflectance and NDVI 46

Fig 4.13 The relationship between reflectance and the subtract between Near­infrared and Red band 46

Fig 4.14 Saturation point indicated from spectral reflectance of wide and narrow area observed in field A 51

Fig 4.15 Saturation point indicated from spectral reflectance of wide and narrow area observed in field A 52

Fig 5.1 ToA reflectace of visible and near­ infrared band of Lansat ETM+ image acquired on June 12th, 2013 58

Fig 5.2 Field spectral measurements and satellite spectral data of field A 60

Fig 5.3 Field spectral measurements and satellite spectral data of field B 61

Table 2.2 Important date 11

Table 2.3 Length of growing stages (days after transplanting­DAT) 11

Table 3.1 Data collected date 13

Table 3.2 The equipment 14

Table 3.3 Physical parameters of rice of paddy field A 16

Table 3.4 Physical parameters of rice of paddy field B 17

Table 3.5 Wavelength and pixel number 24

Table 4.1 The value of coefficient a, b and c 36

Table 4.2 Landsat 7 ETM+ bands 38

Table 4.3 Statistical summary for the polynomial relationship between plant growth (DAT) and reflectance spectral: Red, Green, Blue, NIR, NIR/Red, NDVI 42

Table 4.4 The correlation coefficients calculated from the whole data series 47

Table 4.5 The correlation coefficient when vegetation coverage is less than 90% 48

Table 4.6 Statistical summary for the linear relationship between reflectance spectral: Red, Green, Blue, NIR, NIR/Red, NDVI and vegetation coverage 48

Table 5 1 List of Landsat ETM+ images used 55

Table 5 2 Gain and bias of spectral band of Landsat 7 ETM+ 55

Table 5 3 ETM+ Solar Spectral Irradiances 55

Table 5 4 Earth­Sun Distance in Astronomical Units (updated July 29, 2013) 56

Table 5 5 Earth­Sun distance in astronomical units (d) 57

Table 5 6 Solar zenith angle ( ) 57

Table 5 7 The spectral reflectance of paddy field A calculated from satellite images 58

Table 5 8 The spectral reflectance of paddy field B calculated from satellite images 59

Table 5 9 Vegetation coverage estimated in field A 62

Table 5 10 Vegetation coverage estimated in field B 63

coverage and spectral reflectance of paddy field during rice growing season When does vegetation coverage saturate? How to identify saturation point of vegetation coverage? These question will be solved In addition, to calculate vegetation coverage of paddy fields by using satellite images reflectance is an advanced purpose of this study The field measurement data including spectral data, photos of paddy fields and physical parameters of rice plant were recorded in June to September, 2013 These photos of paddy fields taken by a digital camera

in nadir direction are used to calculate vegetation coverage Spectral data of paddy fields recorded by spectrometer are used to calculate the spectral reflectance

In fact, some results are got from the measurement data The relationship between vegetation coverage and these physical parameters of paddy field can be expressed by using a two variables equation Vegetation coverage increases rapidly in both fields and saturates in early July, about 65 days after transplantation (DAT) The saturation point of vegetation coverage can also be indicated from spectral reflectance and photos of paddy field When vegetation coverage saturates, the difference (NIR­R) is approximately equal 25% Moreover, the reflectance responds differently to solar radiation in the visible and infrared region The characteristics of reflectance of paddy fields in the visible channels reach minimum value at

80 DAT in both field In case of NIR band, the reflectance increases rapidly during the growing season of rice plant In addition, to confirm the correlation between reflectance and vegetation coverage, correlation coefficient R (Pearson) is calculated This coefficient indicates the strongly positive linear relationship between vegetation indices and vegetation coverage This positive linear relationship is also noticed in NIR band Moreover, the satellite reflectance and field reflectance are similar in NIR band and the difference (NIR­R) From the results, when vegetation coverage of paddy fields begins to saturate the increase of spectral reflectance is affected by other factors instead of vegetation coverage as well as physical parameters of rice plant Hence, only if vegetation coverage is less than 90% can vegetation coverage be estimated from with satellite reflectance On the other hand, there are many factors effecting

on vegetation coverage and spectral reflectance The difference of varieties should be mentioned as the first reason The second reason is the difference of initial conditions such as transplanting density and the stem number per one plant Cultivation conditions is also mentioned as one of the reason Moreover, the combination of energy reflected from rice plant, water, soil, shadow of rice plant is the reflected energy of paddy field Therefore, their area also influence the reflectance of paddy field

Chapter 1: INTRODUCTION

1.1 Introduction

Rice is one of the main food source of many countries around the world including Japan For food security and water resource management, the distribution and area of paddy fields are collected as the important information They are essential matters and should be studied in agricultural countries On the other hand, there are many methods that have been applied in this field and remote sensing is one of them This technique can obtain the rice production information from space due to the images recorded by satellites The satellite sensors acquire temperature data as well as spectral reflectance data and store this information as a digital number Uchida (2010) argues that cropping pattern of rice could be monitored by analyzing phonological features obtained from high­temporal resolution satellite data To carrying out the relationship between rice spectral and rice yield, I Wayan Nuarsa (2011) states that rice yield can be estimated from total NDVI because there is the strongest exponential relationship between rice yield and the total NDVI Moreover, rice plants respond differently to solar radiation in the visible and infrared Additional, monitoring the growth of the crop during the growing season can improve the yield estimation V.K.Choubey and Rani Choubey (1999) claims that it is essential to understand the correspondence between crop radiance and field parameters such as species, and growth stage to utilize remote observations of crops

For high quality rice production, the rice growth during growing season should be figured out so that the rice growth as well as cultivation condition can be control Vegetation coverage, expressed by plant height and number of stem, and leaf color can indicate rice growth The parameters can be measured in paddy field but this method needs a lot of time and labor Moreover its value is affected by the measurement points Therefore, a laborsaving method using remote sensing is studied In this study, rice growth information will be tracked

by using spectrometer data recorded in paddy field and reflectance data of the Landsat images Because the reflectance of paddy field depends on rice canopy structure and leaf chlorophyll concentrate, rice growth can be monitored via observing the temporal changes of reflectance

Chapter 1: Introduction

On the other hand, rice growth can be indicated by some well know parameters such as plant height, number of stem and leaf color.If the rice plant is projected on the surface the simple parameter, vegetation coverage, is gotten This is a two­dimensional parameter, which changes easily during rice development period and represents the percentage of area covered by rice per one unit area of paddy field Vegetation coverage is necessary to study because of two reasons At first, it indicates rice growth Additional, there is the relationship between vegetation coverage and total nitrogen content of rice plant [2] Nitrogen content is important factor It effects on cultivation activity such as when to fertilize, how much fertilizer to use Therefore, rice growth and rice yield are also influenced When rice plant grows up, rice canopy also changes Vegetation coverage, along with rice canopy change, also changes It also effects on paddy field reflectance during development season Actually, vegetation coverage can be measured directly in paddy field but this method need a lot of labor and time

to collect the data Spectral reflectance of paddy field is effected by vegetation coverage In other words, vegetation coverage can be indicated if the relationship between them is known Because the reflectance can be gotten from satellite images easily, vegetation coverage estimated from spectral reflectance is a laborsaving method should be studied Hence, the characteristics of spectral reflectance of rice plant and vegetation coverage is the first issue which need to be achieved

1.2 Study area

Fig 1.1 Study area

The study area is in Niigata Prefecture, located on the island of Honshū on the coast of the Sea of Japan The major industry in that place is agriculture and rice is the principal product,

and among the prefectures of Japan Niigata is second only to Hokkaidō in rice output In particular, there are two paddy fields in Koshijinakazawa, Nagaoka City chosen Each paddy field has 90 meters in length and 30 meters in width Two rice varieties are transplanted in those field More detail, Gohyakumangoku, one of the types of sake rice, was transplanted in field A on May 3rd, 2013 and Koshihikari, a popular variety of rice cultivated in Japan, was transplating in field B on May 25th, 2013

1.4 Research flow chart

The rice growth is monitored during growing season by using spectrometer and digital camera The physical parameters of rice plant and temporal measurement of spectrum and photos of paddy field are also measured The spectral data is used to calculate the ground reflectance changing during rice development season On the other hand, the photos of paddy fields taken by digital camera is used to calculated vegetation coverage The characteristics between vegetation coverage and spectral reflectance is study by checking their values changing during development season Moreover, field surveying parameters show the growing condition of paddy fields They effect on characteristics between vegetation coverage and

Chapter 2: The spectral reflectance curve and rice development stage

Chapter 3: Experimental study and ground data

Chapter 4: The characteristics between vegetation coverage and spectral reflectance Chapter 5: Applying to Landsat images

Chapter 6: Summary

Chapter 2: THE SPECTRAL REFLECTANCE CURVE AND RICE

DEVELOPMENT STAGE

2.1 The reflectance and spectral reflectance curve

2.1.1 The reflectance

Spectral reflectance is the ratio of the total amount of radiation, as of light, reflected by

a surface to the total amount of radiation incident on the surface in a particular wavelength range It is usually expressed as percentage and can be calculated by using the following equation

=

Where : the intensity of the incident radiation

: the intensity of the radiation reflected back from a surface

Fig 2.1 Illustration of the relation between radiance received at a sensor (L sensor ), the radiance reflected from a surface (L surface ) and the incoming radiances at the top of the

Chapter 2: The reflectance curve and rice development stage

Skylight is solar radiation reaching the Earth's surface after having been scattered and absorbed from the direct solar beam by molecules in the atmosphere The important processes

in the atmosphere are elastic processes, by which light can be deviated from its path without being absorbed and with no change in wavelength These phenomena also happen to reflected light from the surface In remote sensing applications, the radiation emanating from the surface

is sensed The optical region represents for the most part reflected radiation In order to infer the properties of a land surface from the remotely sensed (reflected) signal, it is necessary to understand how vegetation, soil, and water interact with the incoming radiation to generate the reflected signal [3] The nature of the reflection, in terms of its intensity, its spectral properties, and its spatial or angular properties, contributes information about the surface being studied In order to estimate the corresponding reflectance at the surface atmospheric absorption and the scattering of electromagnetic radiation travelling between the sun, the surface, and the sensor need to be corrected

2.1.2 Spectral reflectance curve

Remote sensing is based on the measurement of reflected or emitted radiation from different bodies Objects having different surface features reflect or absorb the sun's radiation

in different ways The reflectance properties of an object depend on the particular material and its physical and chemical state (e.g moisture), the surface roughness as well as the geometric circumstances (e.g incidence angle of the sunlight) The most important surface features are color, structure and surface texture [6]

The solar spectrum which reflects from an object changes with the wavelength Every object have a specific curve which is not similar to other objects curve in its shape and value These differences make it possible to identify different earth surface features or materials by analyzing their spectral reflectance patterns or spectral signatures These signatures can be visualized in so called spectral reflectance curves as a function of wavelengths Configuration

of the spectral reflectance curves gives us insights into the spectral characteristics of an object Spectral reflectance curves guide us in selecting wavelengths region in which remote sensing data should be acquired for the given science goal The figure 2.2 shows typical spectral

reflectance curves for three basic types: dry bare soil, healthy green vegetation and clear lake water

Fig 2.2 Spectral reflectance properties of soil, vegetation and water, and spectral bands of LANDSAT 7( http://www.seos-project.eu/modules/remotesensing/remotesensing-c01-

The spectral reflectance curve of bare soil is considerably less variable The reflectance curve is affected by moisture content, soil texture, surface roughness, presence of iron oxide and organic matter These factors are less dominant than the absorbance features observed in vegetation reflectance spectra

Chapter 2: The reflectance curve and rice development stage

The water curve is characterized by a high absorption at near infrared wavelengths range and beyond Because of this absorption property, water bodies as well as features containing water can easily be detected, located and delineated with remote sensing data

2.2 Rice growth phase

Vegetative phase Reproductive

phase

Ripening phase

Fig 2.3 Rice growth phases

Growth and development of the rice plant involve continuous change This means important growth events occur in the rice plant at all times Therefore, the overall daily health

of the rice plant is important If the plant is unhealthy during any state of growth, the overall growth, development and grain yield of the plant are limited It is important to understand the growth and development of the plant [7] The ability to identify growth stages is important for proper management of the rice crop Because management practices are tied to the growth and development of the rice plant, an understanding of the growth of rice is essential for management of a healthy crop The growth duration of the rice plant is 3­6 months, depending

on the variety and the environment under which it is grown It is divided into three main phases including many stages: vegetative (germination to panicle initiation); reproductive (panicle initiation to flowering); and ripening (flowering to mature grain)

Table 2.1 Rice growth stages

I Vegetative (germination to panicle

initiation)

Stage 0 from germination to emergence

Stage 1: Seeding

Stage 2: Tillering

Stage 3:Stem elongation

II Reproductive (panicle initiation to

flowering)

Stage 4: Panicle initiation to booting

Stage 5: Heading or panicle exsertion

Stage 6: Flowering

III Ripening (flowering to mature

grain)

Stage 7: Milk grain stage

Stage 8: Dough grain stage

Stage 9: Mature grain stage

Chapter 2: The reflectance curve and rice development stage

2.2.1 Vegetative phase

The vegetative growth stage is characterized by active tillering, a gradual increase in plant height and leaf emergence at regular intervals Tillers that do not bear panicles are called ineffective tillers The number of ineffective tillers is a closely examined trait in plant breeding since it is undesirable in irrigated varieties, but sometimes an advantage in rain fed lowland varieties where productive tillers or panicles may be lost due to unfavorable conditions The length of this stage primarily determines the growth duration of varieties Some very early maturing varieties have a shortened vegetative growth stage, while others have both shortened vegetative and reproductive growth stages Panicle initiation may occur before the maximum tiller number is reached in very short ­season and some short season varieties

2.2.2 Reproductive phase

The reproductive stage is characterized by culm elongation, a decline in tiller number, booting, emergence of the flag leaf, heading and flowering of the spikelets The reproductive stage usually lasts approximately 30 days in most varieties This stage is sometimes referred

to as the internode elongation or jointing stage and varies slightly by variety and weather conditions

2.2.3 Ripening phase

The grain filling and ripening or maturation stage follows ovary fertilization and is characterized by grain growth During this period, the grain increases in size and weight as the starch and sugars are translocated from the culms and leaf sheaths where they have accumulated, the grain changes color from green to gold or straw color at maturity and the leaves of the rice plant begin to senesce Light intensity is very important during this interval since 60 percent or more of the carbohydrates used in grain filling are photosynthesized during this time interval This period is also affected by temperature

The final component, individual grain weight, is determined during the ripening stage Although grain weight is relatively stable for a given variety, it can be influenced by the environment High temperatures tend to reduce the grain filling period and may reduce grain

weight Low temperatures tend to lengthen the time required for grain fill and ripening The ripening process may cease after a significant frost occurs

In this study, two rice varieties are transplanted in two paddy fields Their growth phase are expressed specifically in the following tables

Table 2.2 Important date

Field Transplanting date Heading date Harvesting date

A May 03rd, 2013 July 21st, 2013 Aug 29th, 2013

B May 25th, 2013 Aug 10th, 2013 Sep 21st, 2013

Table 2.3 Length of growing stages (days after transplanting-DAT)

Field Varieties Vegetative phase Reproductive phase Ripening phase

Chapter 3: Experimental study and ground data

Chapter 3: EXPERIMENTAL STUDY AND GROUND DATA

3.1- Experimental study

In this study, the rice growth information is monitored during rice development season Rice grows up and its reflectance also changes during development season Rice growing season is divided into three main phase (in page 9) Photos of rice fields recorded by digital camera throughout the development season is displayed in figure 3.1 According to these photos the vegetation coverage increases over time while soil and water area surrounding rice plant decreases Therefore, the reflected energy recorded from paddy fields is recognized as

a combination of rice, water and soil mixed together By recording the reflectance, the rice growth can be tracked This is the main idea of this study

Fig 3.1 The change of paddy field during rice development season

34 days

(June 6th)

55 days (June 27th)

83 days

(July 25th)

108 days (Aug 19th)

There are twenty five experiments performed in paddy field A and B The process was carried out in the same period of the day In each times, there are three types of data recorded They are spectral data, photo data and physical parameters of the rice plant The calculation

of vegetation coverage and reflectance of paddy fields during development season is performed in order to understand the characteristics of the vegetation coverage and spectral reflectance

Table 3.1 Data collected date

Chapter 3: Experimental study and ground data

3.1.1 Equipment

Spectrometer Ocean Optics SD2000 connected to a laptop is used to collect spectral data

of paddy fields All data is saved in that minicomputer Moreover, these photos of paddy fields are taken by a digital camera in nadir direction for each experiment The camera automatically takes images at many different times and the interval between two shots is one minute These photos are taken simultaneously with spectral data All of them are used to calculate vegetation coverage of paddy fields

Table 3.2 The equipment

intensity of reflected light from an area with 0.5 m radius in paddy field is collected Moreover

a digital camera is also used to take the photo of paddy field in nadir direction

Fig 3.1 Obtaining the ground field of view

(a) Observe the reflectance of wide area

(b) Observe the reflectance of narrow area

Fig 3.2 Experiment specification

Chapter 3: Experimental study and ground data

For each experiment, there are two types of target area that are observed for each sample area using spectrophotometer The first one is wide area including rice plant and background (shadow, soil, water ) The second one is narrow area The radiation intensity of skylight and reflected radiation from the object surface are acquired at the same time For each target objects these data is recorded 5 times Moreover, the photos of paddy fields are also taken several times during experiment in order to record rice canopy surface Thanks to this, vegetation coverage of paddy field is calculated

3.2-Experimental data processing

3.2.1 Calculate vegetation coverage from photos taken in paddy field

Vegetation coverage (VC), a concept which is mentioned, is a two­dimensional parameter It represents the percentage of area covered by rice per one unit area of paddy field This is a parameter changing easily during rice development period It need a lot of labor and time to measure vegetation coverage directly on the paddies Moreover, its value is affected by the physical parameters of rice and depends on the transplanting density The physical parameters of rice such as the plant height or number of stem per one plant can be measured simply During rice development time, these physical parameters of rice are also observed For each paddy fields, there are five rice plants chosen to measure The average value of them would be considered as representative values of paddy field

Table 3.3 Physical parameters of rice of paddy field A

Date

Plant height (cm) Number of stem A1 A2 A3 A4 A5 A1 A2 A3 A4 A5 06­Jun­13 29 29 32 28 30 12 5 11 3 10

Table 3.4 Physical parameters of rice of paddy field B

Date

Plant height (cm) Number of stem B1 B2 B3 B4 B5 B1 B2 B3 B4 B5 20­Jun­13 43 41 42 41 42 16 14 12 10 11 24­Jun­13 50 50 49 50 47 19 19 18 18 17 27­Jun­13 50 50 49 50 48 28 24 23 23 22 01­Jul­13 54 51 53 53 50 30 30 27 29 28 04­Jul­13 58 53 54 54 52 37 31 30 36 32 08­Jul­13 64 62 64 66 66 43 38 37 46 41 11­Jul­13 79 77 75 76 75 44 40 36 45 42

01­Jul­13 67 64 64 60 65 37 21 37 15 36 04­Jul­13 72 67 70 66 71 38 22 39 18 37 08­Jul­13 74 71 72 69 71 34 21 35 16 33 11­Jul­13 76 75 77 73 78 29 19 30 15 30 16­Jul­13 86 80 85 82 82 29 18 30 13 29 18­Jul­13 93 92 90 92 92 26 17 30 13 26 19­Jul­13 100 99 94 97 96 27 17 31 13 26 22­Jul­13 104 104 100 104 102 26 17 31 13 26 25­Jul­13 111 108 107 109 109 26 17 31 13 26 30­Jul­13 120 116 116 114 116 26 17 31 13 26 02­Aug­13 121 117 121 115 116 26 17 31 13 26 06­Aug­13 119 116 118 116 116 26 17 31 13 26 08­Aug­13 122 115 119 116 116 26 17 31 13 26 15­Aug­13 123 116 119 116 115 26 17 31 13 26 19­Aug­13 121 117 119 116 117 26 17 31 13 26 22­Aug­13 122 115 116 113 117 26 17 31 13 26 26­Aug­13 123 114 118 107 118 26 17 31 13 26 28­Aug­13 122 115 116 108 113 26 17 31 13 26

Chapter 3: Experimental study and ground data

16­Jul­13 89 91 88 87 87 40 37 32 38 37 19­Jul­13 92 93 91 89 91 41 36 37 34 34 22­Jul­13 93 94 92 90 93 38 35 32 31 34 25­Jul­13 97 94 95 93 98 36 35 30 30 32 30­Jul­13 101 98 96 98 101 36 34 29 29 33 02­Aug­13 101 101 96 99 100 34 34 28 30 32 06­Aug­13 103 102 100 104 105 34 34 28 29 34 08­Aug­13 106 104 102 106 108 34 34 28 29 34 15­Aug­13 120 118 116 120 117 34 34 28 30 34 19­Aug­13 126 122 118 124 124 34 34 28 30 34 22­Aug­13 126 122 117 122 124 34 34 28 30 34 26­Aug­13 126 123 117 123 123 34 34 28 30 34 28­Aug­13 126 123 118 122 124 34 34 28 30 34

In this study, these photos of paddy fields are taken by a digital camera in nadir direction The camera automatically records images at many different times and the interval between two shots is one minute These photos are shot simultaneously with spectral data All of them are used to calculate vegetation coverage in paddy fields Basically, the objects in the images are classified into two classes, vegetation and non­vegetation Vegetation class represents for rice plant and non­vegetation class represented for background objects including water, soil and shadow of rice plants In order to perform this process, an index is calculated for all images

by using PCI Geomatica That is greenness index calculated from the red, green and blue band

of the original images (eq 3.1) Corresponding to each images, the threshold value is set The pixels having greenness index which is greater than the threshold will be classified into vegetation class and the rest will be non ­vegetation class (fig 3.2) The vegetation coverage

is calculated as a ratio of the number of vegetation pixels to total pixels of the photo and expressed as percentage (eq 3.2)

(%) = ℎ

(a) Greenness image (b) Classified image

Fig 3.3 Classified image processing

Fig 3.4 The photos taken in paddy field were affected by weather conditions

Additional, the photos are taken every minute during experimental study There are many photos taken and they have a little difference because of weather condition effects such as: cloud, wind, sunlight The vegetation coverage is calculated based on these photos Therefore, the images taken in the same experiment are used to calculate the average value of vegetation coverage representing for whole paddy in order to eliminate errors caused by weather conditions mentioned above (fig 3.3) The following figure shows the vegetation coverage change of paddy field A and B during rice development time Figure 3.4 shows that

VC increases rapidly in both fields and reaches 90% in early July, about 65 days after

Background Vegetation

Chapter 3: Experimental study and ground data

transplantation (DAT) Then, this status is maintained and there is a little decrease before harvesting time In fact, field B saturates sooner than field A because of different rice varieties

Fig 3.5 Vegetation coverage change during development time

3.2.2 Calculate reflectance from spectral data

Spectral data of paddy fields are recorded for each experimental test They are used to calculate the spectral reflectance Regarding to the fundamental, the reflectance is calculated

as the ratio between the intensity of light reflected from the object surface and the intensity of the incident light as equation (2­1) However, in the process of data acquisition, there are some factors that lead to a little change in the calculation process To acquire the intensity of the skylight and reflected light from the object surface there are two spectral cables used One spectral cable ends is attached to the spectrometer and another one is attached to a black hollow plastic tube with one end Each tube is high 4.4 cm and its diameter is 3.8 cm Because the intensity of skylight is many times as much as the intensity of the light reflected from ground objects surface it is difficult to collect them at the same time in visible wavelength range Therefore, when the experiment is performed, in case of the cable receiving energy from sunlight the tube is covered by a white paper on the top to reduce the intensity of the skylight (fig 3.5) Thanks to that the intensity of reflected light from the object surface and skylight

65 DAT VC=90%

can be recorded at the same time In order to calculate the reflectance correctly the transmittance coefficient of the white paper have to be known Therefore, the equation (2­1) can be rewritten as following:

=

Where : the intensity of the incident radiation

: the intensity of the radiation reflected back from a surface

: Transmittance coefficient of white paper

Fig 3.6 Spectral data acquisition

Because of the effects of spectrometer itself the accuracy of spectral data, which is recorded, is also influenced In this case, wavelength and intensity calibration have to perform

in order to reduce that effects The new wavelength is interpolated and the interval is 0.25 micrometers The intensity is also interpolated correspondingly The entire process is displayed in the figure 3.6 and will be applied to all spectral data

Spectrometer

Digital camera White paper

Chapter 3: Experimental study and ground data

Fig 3.7 Calculate reflectance from spectral data procedure

3.2.2.1 Wavelength calibration

To calibrate the wavelength of spectrometer, a calibration lamp with Mercury and Argon gas emission lines from 253.65­1013.98nm is used to perform wavelength calibration and demonstrate spectrometer resolution and operational range for UV/VIS/NIR/XNIR configurations Through each spectrometer is calibrated before it leaves Ocean Optics, the wavelength for all spectrometers will drift slightly as a function of time and environmental condition The relationship between pixel number and wavelength is a second­order polynomial

Fig 3.8 SL2 Mercury Argon Calibration Lamp

Where is the wavelength of pixel p

I is the wavelength of pixel 0

C1 is the first coefficient (nm/pixel)

C2 is the second coefficient (nm/pixel2)

At first, a spectrum of light source is taken Adjust the integration time until there are several peaks on the display screen that are not off scale Move the cursor to one of the peaks and carefully position it so that it is at the point of maximum intensity Record the pixel number that is displayed in the status bar (located beneath the graph) Repeat this step for all of the peaks in the spectrum A table with two column is created In the first column, place the exact

or true wavelength of the spectral lines used Most calibration line sources come with a wavelength calibration sheet In this case, this sheet is attached on the lamp In the second column of this worksheet, place the observed pixel number Now the wavelength calibration coefficients can be calculated The functions to perform linear regressions is found Select the true wavelength as the dependent variable (Y) Select the pixel number as the independent variables (X) After the regression is executed, these coefficients are found out This process

is repeated for two channel of all spectral data After wavelength calibration, the new wavelengths are interpolated with the wavelength interval is 0.25 micrometers

Chapter 3: Experimental study and ground data

Table 3.5 Wavelength and pixel number

wavelength (nm) pixel number (ID) wavelength (nm) pixel number (ID)

(a) Master channel

λ= ­2E­05ID2+ 0,3835ID + 337,84

R² = 10

20040060080010001200

20040060080010001200

Chapter 3: Experimental study and ground data

should be calibrated to compensate for wavelength dependent CCD sensitivity and the effect

of color filters This is the main purpose of this step

Fig 3.10 Radiometric reference sources and color filters

To perform this procedure, a radiometric reference sources (RRS) is used At first, a spectrum of this reference source is taken Then, this spectral data will be calibrated wavelength The new values of wavelength are interpolated with the interval is 0.25 micrometer The new values of intensity are also interpolated correspondingly The intensity calibration coefficients is calculated as a ratio between the intensity of master and slave channel Finally, this value will be used to calibrate the intensity of slave channel of all observed data The procedure can be display as the following equations

Where (Cal_skylight ) : the calibrated intensity of slave channel of observed data

(skylight ) : the intensity of slave channel of observed data

C : the intensity calibration coefficient

C = (Slave channel )

Where (Slave channel ) : Intensity of slave channel of reference source

(Master channel ) : Intensity of master channel of reference source

Fig 3.11 The intensity calibration coefficient

Fig 3.12 The intensity of slave channel after calibrating intensity

3.2.2.3 Calculate transmittance coefficient

Transmittance is the fraction of incident light (electromagnetic radiation) at a specified wavelength that passes through a sample It can be calculated by using the following equation

Before