Remote Sensing Techniques used in nutrient studies in plant and soil . Prospects as far as space technology is concerned in agriculture. The usefulness of Hyperspectral and multispectral remote sensing in agriculture sector. How can remote sensing protect the soil from degradation and increase food production with sustainable management practice of agricultural land?
Use of remote sensing techniques for nutrient studies in soil and plant _Knight_RAD2020-42.pptx
1. PROFESSOR JAYASHANKAR TELANGANASTATE
AGRICULTURAL UNIVERSITY
DOCTORAL SEMINAR ON
USE OF REMOTE SENSING TECHNIQUES
FOR NUTRIENT STUDIES IN SOILAND
PLANTS
PRESENTATION BY
Knight Nthebere
RAD/20-42
Department of Soil Science
and Agril. Chemistry.
2. • Introduction
• What is Remote sensing?
• Remote sensing process
• Hyperspectral remote sensing in soil and plant studies
• Factors affecting soil reflectance
• Advantages and disadvantages of remote sensing
techniques in nutrient studies
• Case studies
• Summary and Conclusion
• Future line of study
CONTENTS
3. Impeding crisis of food
production
Scope
• Worlds population growth up to
10 billion by 2050, boosting
agricultural demand (FAO, 2017).
• Increase in food production must
be accompanied by sustainable
management of agricultural lands
to stop or slow down the impacts
on quality and quantity of soil
resources, land degradation and
biodiversity.
• Challenge in monitoring crop
growth and status in various
locations and environmental
contexts, with various temporal
resolutions, and for different
purposes (Weiss et al., 2020).
• Remote sensing appears as an
essential tool to respond to those
challenges since it offers a non-
destructive means of providing
recurrent information from local
to global scale in a systematic
way.
• Thus, enabling characterization
of the spatio-temporal variability
within a given area.
INTRODUCTION
4. Remote sensing is:
“The art and science of obtaining information
about an object without being in direct
contact with the object” (Jensen, 2009).
India’s National Remote Sensing Agency
(NRSA) defined as :
“Remote sensing is the technique of
deriving information about objects
on the surface of the earth without
physically coming into contact with them”
WHAT IS REMOTE SENSING?
5. Electromagnetic energy reaching the
earth’s surface from the sun is reflected,
transmitted or absorbed.
The specific target have an individual
and characteristic manner of interacting
with incident radiation that is described
by the spectral response of that target.
Egs- soils, of different types, water
with varying degrees of impurities, or
vegetation of various species.
PRINCIPLES BEHIND RS
6. (A)Energy source or
illumination.
(B)Radiation and the
Atmosphere
(C)Interaction with the
target.
(D)Recording of Energy
by the Sensor.
(E)Transmission,
Reception &
Processing.
(F)Interpretation and
Analysis.
(G)Application
Source: Canadian Centre of Remote Sensing
Figure 1. Remote sensing process
REMOTE SENSING PROCESS
8. There are different remote sensing indices such as
1. Normalised Difference Vegetation Index (NDVI)
2. Vegetation Condition Index (VCI)
3. Temperature Condition Index (TCI)
4. Vegetation Health Index (VHI
• NDVI is computed by the formula as:
• NDVI is the measure of greenness or
vigour of vegetation.
• The basic concept of NDVI is that the
healthy vegetation reflects NIR radiation and
absorbs RED radiation. This becomes reverse
in case of unhealthy or stressed vegetation.
Figure 3. Absorbance and reflectance of
radiation in healthy and unhealthy plant
REMOTE SENSING INDICES
9. ADVANTAGES OF REMOTE SENSING
Extent of coverage
Permanent and reliable record
Speed and consistency of interpretation of data
Reliable information
The data is available to multi-disciplinary use
The process of data acquisition and analysis is faster
Satellite data is too expensive
Remotely sensed data is complicated to use and not readily reliable
Satellite based remote sensing does not have sufficient resolution
DISADVANTAGES OF REMOTE SENSING
10. • Soil reflectance has hardly any peak
and trough, limiting use of
multispectral RS for soil studies.
• Hyperspectral RS is preferred to study
the complex features like soil having
varying compositions of number of
parameters
• Hyperspectral RS helps in
discriminating minute variation in
soil composition and quantification of
different parameters
Hoffer, 2003
Figure 4a & b. Reflectance of
different parameters
HYPERSPECTRAL REMOTE SENSING FOR SOIL STUDIES
11. FACTORS AFFECTING SOIL REFLECTANCE
Reflectance generally increases with increasing wavelength throughout the
visible, near- and middle infrared portions of the spectrum
Finer texture results in higher reflectance generally because of
denser packing of particles
Particle size (Texture)
Figure 5. Effect of particle size (texture) on soil reflectance
12. SOIL MOISTURE
With increase in moisture content reflectance decreases
Increased water content means increased absorption of radiation
Soil moisture strongly related to soil texture due to interstitial spaces
Higher moisture content in sandy and clayey soil results in decreased
reflectance throughout the VNIR region, especially in the water absorption
bands at 1.4, 1.9 and 2.7 µm and hydroxyl absorption bands at 1.4 and 2.2
µm Hoffer, 1978
Figure 6 a & b. Effect of soil moisture on soil reflectance
13. ORGANIC MATTER
OM has spectral activity throughout the entire VNIR-SWIR region (mainly in visible)
Higher the OM, lower the reflectance
At organic-matter contents greater than 2%, the decrease due to organic matter may
mask other absorption features in soil spectra
The spectra of soils with organic-matter contents greater than 5 % often have a
concave shape between 0.5 and 1.3 m
Beck et al. (1976) suggested that the region 0.90 to 1.22 m is suited for mapping
organic matter in soils
Figure 7. Effect of soil organic matter on soil reflectance
14. IRON OXIDES, SOIL COLOR
• Presence of iron oxides results in increase in reflectance in the red
portion of the spectrum (600 – 700 nm).
• Electronic transitions of Fe3+ and Fe2+ cause absorption bands near 0.7 and
0.87 µm as well as 1.0 µm respectively.
• Absorption due to Fe in the mid IR can be strong enough to obliterate the
water-absorption band at 1.4 m.
Figure 8. Effect of iron oxides, soil color on soil reflectance
15. MINERAL COMPOSITION
Hydroxyl bands near 1.4 and 2.2. m
are characteristic of layer silicates.
Calcite display absorption bands between
1.8 and 2.5 m due to carbonate.
Montmorillonitic soils had the lowest
average spectral reflectances between
0.52 and 1.0 m.
Kaolinitic soils generally display a wide
absorption band near 0.9 m due to
the common presence of free iron oxides
• Orthoclase feldspar, a dominant mineral in granite, shows almost no significant
absorption features in the Vis-NIR spectral range.
• Hydroxide ions (OH-1) produce the strong absorption band near 1.4 µm, along with the
weak 1.9 µm band in kaolinite
• Bound water molecules cause the stronger 1.9 µm band in montmorillonite
• Hydroxyl band at 2.2 m is difficult to identify in most spectra, but is apparent when
clay contents are greater than 20%. Stoner and Baumgardner (1980)
Figure 8. Effect of mineral composition on
soil reflectance
16. Surface roughness
Finer textured soil generally have higher reflectance compared to
coarse textured soil assuming they contain no moisture, organic matter and
iron oxides.
Dry, fine textured clayey soil have more reflectance in visible and
near-infrared region compared to sand and silt.
However
, the presence of soil moisture, organic matter and other
soil minerals masks the effect of reflectance of fine clay.
Fine grained clayey soils with high moisture, organic matter have lower
reflectance than coarser dry sands.
17. Reflectance of soil generally increases with the increase in salt
concentration of the surface soils.
Salt affected soils exhibit higher reflectance in the visible and near-
infrared spectrum compared to non-saline soils. (Rao et al., 1995)
Salt affected soils appear as bright white patches in color-infrared aerial
photography or multispectral false color composites.
SOIL SALINITY
Figure 9. Effect of soil salinity on soil reflectance
18. HYPERSPECTRAL RS FOR SOIL AND CROP STUDIES
Soil related studies
Characterization, mapping
Parameter estimation
Soil fertility and quality monitoring
Crop related studies
Crop discrimination and
characterization
Abiotic stress monitoring
- Moisture
- Salinity
- Nutrient stress (N, P, K)
Biotic stress monitoring
- Pests and Diseases
Approaches
Spectral Indices
Spectral
matching/unmixing
Spectral feature
fitting/continuum
removal
Empirical /regression
modelling
Spectral shift – Red edge
19. Hyperspectral remote sensing data derived spectral indices in
characterizing salt-affected soils: a case study of Indo-Gangetic
plains of India
Mathura district of Uttar Pradesh, India
Area characterized by semi-arid climate
Hyperion-EO-1 data of summer month (May 2007)
when fields had no crops
Preprocessing of data (bad band removal, stripe
removal and conversion to reflectance)
Reconnaissance survey: identification of salt
affected areas
Collection of surface and subsurface soil samples
along with GPS locations
Table 1. Physico-chemical characteristics of salt-affected soils
Suresh Kumar et al., 2015
CHARACTERIZATION OF SALT AFFECTED SOILS
20. Table 2. Correlation analysis of spectral
bands with soil salinity parameters
Suresh Kumar et al., 2015
Table 3. Spectral indices based on
sensitive bands
Correlation analysis was performed between the mean spectral
reflectance value and soil salinity parameters to identify soil salinity
sensitive bands
Sensitive spectral bands were used to compute spectral indices
CONT…
21. Table 4. Correlation analysis of
spectral indices with soil salinity
parameters
Table 5. Regression equations
developed between spectral indices
and soil salinity parameters
Stepwise discriminant analysis (SDA) done to study the effectiveness of
spectral indices in discriminating varying degrees of salinity
High F value indicates more separability
Salinity index (SI) found to be the most suitable
spectral index in characterizing degree of salinity
Suresh Kumar et al., 2015
CONT…
22. Spatial distribution of soil salinity parameters using
Salinity Index (a) and Brightness index(b)
Suresh Kumar et al., 2015
Figure 10. EC distribution map Figure 11. ESP distribution map
CONT…
23. Saha et al.,2013
Table 6. Spatial Prediction of SOC using Geo-statistics Technique using Hyperspectral RS as
Auxiliary Variable
SPATIAL PREDICTION OF SOIL ORGANIC MATTER
24. SOC MAP GENERATED BY CO-KRIGINGWITH HYPERION BAND 57
Saha et al.,2013
25. SOC MAP GENERATED BY CO-KRIGING WITH HYPERION COLOURATION INDEX
Saha et al.,2013
27. Quantitative mapping of soil organic matter using field
spectrometer and hyperspectral remote sensing
Hengshan county in northern Shanxi province, China
Wind erosion coexisting with water erosion
Hyperion-EO-1 data of July 2003
Preprocessed and converted to reflectance
Image color indices generated using R, G and B bands
Correlation analysis between SOM and image indices
Multivariate regression
Table 7. Correlation of SOM with image indices
Zhuo et al, 2008
Table 8. Multivariate regression model developed
QUANTITATIVE MAPPING OF SOIL ORGANIC MATTER
28. Figure 13. Continuous map of soil
organic matter
Zhuo et al, 2008
Accuracy of quantitative mapping was evaluated by relationship between measured and
predicted values
Good relationship between the measured SOM content and the predicted ones convey
that we can predict the SOM concentration at 76% confidence
Figure 12. Scatterplot showing
accuracy of quantitative mapping
CONT…
29. Study area is located in Bharatpur district of Rajasthan, India
Ground truth observations were taken from various plots in a research farm managed
by NRCRM and farmers’ fields with varying levels of disease severity.
The disease severity scores were also recorded simultaneously
Hyperion-EO-1 (Feb, 2005) data preprocessed and converted to reflectance
Multi-date IRS LISS-IV data was used to delineate mustard crop using hierarchical
decision tree based classifier and a mustard crop mask was generated.
Dutta et al., 2006
Figure 14. Hyperspectral RS to identify disease in Mustard crop
HYPERSPECTRAL RS TO IDENTIFY DISEASE IN CROP
30. CONT…
The absorption in red region at 681 nm
decreased with the growth of diseases
intensity.
Increase in reflectance % with disease severity
was observed
At 1660 nm, the diseased crops showed sharp
increase in reflectance from 0.122 to 0.177 %
for disease severity level from 0 to 46 %.
Separability analysis revealed that at 1660 nm
the reflectance of healthy and diseased crops
are significantly separable
Highest difference in reflectance % was found
at 2143 nm, among the various disease
severity levels (0 to 46%)
For identification of diseased crop the five
Disease-Water-Stress-Indices (DWSI) were
used (Apan et al ., 2004)
650-1100 nm
Dutta et al., 2006
1450-1850 nm
2000-2400 nm
31. DWSI-3 (1660/680 nm) was found to best correlated (68%) with the disease score
Dutta et al., 2006
Multidate LISS-IV data classified using DT and mustard crop mask generated.
DWSI indices derived from hyperspectral data were superimposed on the crop mask
DWSI varied from 1.4 - 1.7 for disease score from 15 to 46 %.
Thus diseased mustard crop have been discriminated from healthy crop
Figure 15. Scatter plot
Figure 16. Disease map
CONT…
32. Application of Hyperion data to land degradation mapping in
the Hengshan region of China
Wu et al., 2010
Hengshan county in northern China
Nearly two-thirds of the cultivated
land in this county is in ecological
degradation
Field reflectance spectra collected
were used for generation of spectral
library of different degraded lands.
Figure 17. Methodology used
Table 9. Spectral indices and equations
HYPERION DATA TO LAND DEGRADATION MAPPING
33. Contd…
Figure 19. Reflectance spectra of soils
at different wavelengths
desertification levels
Desertification soil has the highest DSI value (bright)
Bare land, wild grass ground and cultivated land have
relatively high values (middle)
Vegetation and water body have low values (dark).
Wu et al., 2010 Figure 18. DSI image using
Hyperion data
35. LAND DEGRADATION ESTIMATION SYSTEMS FOR STUDY AREA
High SAVI value depicts an abundant vegetation covering and a strong capability of withstanding
wind (water) erosion
High DSI value reflects little vegetation cover with serious soil and water loss
High SOMI and SFOI value indicates an abundant content of total potassium and total
phosphorus and illustrates the soil fertility and vegetation health
Low NDWI value denotes severe soil aridity (indicates soil and water conservation ability of land)
Table 10. Land degradation
Mapping
Wu et al., 2010
Fig.21
Table 9. Land degradation estimation systems for study area
36. Retrieval of soil physicochemical properties towards assessing salt affected
soils using Hyperspectral Data
Ahmedabad district of Gujarat, India
Area characterized by hot semi-arid climate
Hyperion image of April, 2013
Preprocessing and conversion to reflectance
MNF transformation for dimensionality reduction
Generation of spectral library of different salinity classes
Tarik Mitran et al., 2015
feature
Derivation of various absorption
parameters(depth, width, asymmetry, area)
Table 11. Correlation matrix between EC and absorption feature parameters
ASSESSMENT OF SALT AFFECTED SOILS USING HYPERSPECTRAL DATA
37. Area and depth of the absorption features of salt-affected soils decreases with
the increasing EC, ESP and CEC values
Tarik Mitran et al., 2015
Regression equations between EC, and reflectance at selected wavelength
(1020, 1450, 1920 nm)
CONT…
38. Tarik Mitran et al., 2015
Figure 22. Scatter plots of measured vs. predicted Mg2+ and EC derived from
PLSR analysis
CONT…
Table 12. Regression models or equations
39. • Spectral library generated from
standard spectra of different salinity
classes
• Spatial distribution of salt affected
soils was mapped using spectral library
through Spectral Angle Mapper
method (SAM) with considerable
accuracy (89%)
Tarik Mitran et al., 2015
CONT…
Figure 23. Spatial distribution map of salt
affected soils using Spectral angle mapper
Table 13. Overall accuracy of SAM method
40. HYPERSPECTRAL ANALYSIS OF CLAY MINERALS
Gwalior and Shivpuri districts of MP
Hyperion of April 2011
Soil sampling for XRD and spectral analysis
Clay mineral identification using XRD
Absorption features in spectra identified by
characterizing the shape and wavelength
position of strongest absorption features
Derivation of absorption feature parameters
Suresh et al., 2015
(position or center of absorption band,
depth, width, area and asymmetry of the
continuum removed spectra)
Table 14. Clay mineral identification
using XRD
41. Suresh et al., 2015
K=Kaolinite, M=Montomorillonite, I=Illite,
G=Goethite, V=Vermiculite
Figure. 24: Dominant clay mineral in the soil samples were predicted using
absorption feature parameters as input parameters employing random forest
approach
Table 15. Error matrix of the random
forest based classification
Width of the absorption – most
important variable for prediction
(explains 31% of variance)
Unknown samples were classified
with an accuracy of 0.795
CONT…
42. data was classified using SAM
Hyperion
algorithm
Algorithm was trained based on the field
observations in combination with XRD analysis
Kaolinite was found to be the dominant
mineral followed by Montmorillonite
Overall accuracy of the classification was
found to be 68.43%
In the study area the uplands are mainly with
non-expanding clay minerals, while lowlands
(bottom portion of the image) are dominated
by expanding clay minerals
Suresh et al., 2015
CONT…
Figure 25. SAM
Classification
43. study area
Felcra Kampung Kurnia
Gelong Gajah, Ayer Tawar,
Perak, Malaysia.
Three phases of the oil palm
plantation on phase 3 of the
project
Cultivated on 100 acre.
Methodology used
Image processing on RS.
soil sampling (N.P.K), pH &
testing for palm trees ground
data
Identification of healthy &
unhealthy palm
Detecting nutrient deficiencies of oil palm trees with remotely sensed Data
Figure 26 . Study area (Ayer Tawar, Perak)
Marzukhi et al., 2016
45. Table 16. Identification on the plot area .
Plot sample Coordinate location of plot area(m) Area (m2) Physical
condition of the
plot area
x y
1 311554.58 485331.00 261605.45 Healthy
2 312203.02 485689.58 461572.07 Healthy
3 311884.60 484837.14 385592.61 Unhealthy
4 312531.23 485258.92 160433.07 Unhealthy
5 311796.43 485933.77 600754.37 Healthy
Marzukhi et al., 2016
CONT…
46. Table 17. Result of analysis for nutrient content
Plot pH. value
Total Nitrogen
(N)%
Status Phosphorus
(P) mg/kg Status Potassium(K)
meq/100g
Status
1 0.25 Very
High 197 High 0.75 Very
High 5.8
2 0.20 Very
High 40 Very
Low
0.98 Very
High 5.4
3 0.17 Moderate 7 Very
Low
0.20 Low 5.0
4 0.14 Moderate 81 Very
Low
0.19 Low 4.8
5 0.21 Very
High 180 Low 0.65 Very
High 5.0
Marzukhi et al., 2016
It can be concluded that the healthy oil palm has the high NDVI value and the
unhealthy oil palm tree has the lowest NDVI value.
Cont…
47. CONT…
NDVI Min Max Median Mean
Plot 1 0.5000 0.67816 0.65032 0.649
Plot 2 0.45833 0.68000 0.64796 0.645
Plot 3 0.21818 0.34375 0.31285 0.312
Plot 4 0.28000 0.3813 0.30707 0.306
Plot 5 0.56204 0.67251 0.30705 0.645
Table 18 . Results of NDVI .
Figure 28 . NDVI process on the plot area.
Marzukhi et al., 2016
NDVI range from 0.30 to 0.65.
The healthy tree has low red light
reflectance and high near infrared
reflectance.
Therefore, the value of NDVI is
analyzed as high.
Based on that study, the unhealthy
tree has high red light reflectance
and lower near infrared reflectance
so the result of the NDVI value is
low.
It also indicated that the lowest of
NDVI value are on the Plot 3 and
Plot 4
48. CONT…
Figure 29 . Result of SAVI analysis on the plot area
SAVI Min Max Median Mean
Plot 1 0.89701 1.0143 0.97262 0.972
Plot 2 0.68512 1.0171 0.96911 0.964
Plot 3 0.90492 1.0148 0.97491 0.975
Plot 4 0.87197 1.0029 0.95843 0.958
Plot 5 0.8400 1.0058 0.96826 0.965
Table 19 . The result of SAVI.
The average of the SAVI value is 0.966 and it shows that the
value is not much different.
49. CONT…
Figure 30. The correlation relationship of NDVI and SAVI towards
nutrients analysis.
Marzukhi et al., 2016
50. CONT…
Correlations element R2 value
NDVI SAVI
Nitrogen (N) 0.7442 0.2171
Phosphorus (P) 0.3813 0.0009
Potassium (K) 0.8811 0.012
pH. value 0.4789 0.1969
Table 20 . Comparison of both correlation relationship
of vegetation indices towards soil properties.
Figure 31 . Oil palm tree spot on the plot 3
area due to N deficiency
Figure 32 . Leaves dryness indicate in
brown colour. Fruit bunches is
unusually small due to P deficiency
Marzukhi et al., 2016
NDVI values are
higher than SAVI in
comparison.
51. CONT…
Figure 33 . Symptom in lacking potassium on
leaves.
Marzukhi et al., 2016
52. DETERMINING SPATIOTEMPORAL DISTRIBUTION OF
MACRONUTRIENTS IN A CORNFIELD USING REMOTE SENSING
AND A DEEP LEARNING MODEL
PROPOSED METHOD
STUDY AREA
The field for cultivation of corn was located in
SacheonCity, Republic of Korea.
The cultivation plan spanned from May to
September 2019, which mostly coincided with
the summer season.
Generally, in this period of the year, the city
experiences temperatures as high as 39◦C
And there are infrequent raining which cannot
affect the fertilizer distribution in the
cornfield.
The field was divided into 5 equal sections,
S1-S5, from which soil samples were collected
for determining the NPKC nutrient
distributions.
Figure 34. Parrot Bluegrass
UAV and its components.
Jaihuni et al., 2021
53. The above study was intended to tackle the fertilizer consumption
inefficiencies by utilizing non-destructive remote sensing
technologies, soil macronutrient distribution analysis, and a deep
learning model.
Specifically, an Unmanned Air Vehicle (UAV) was used in a
cornfield to capture the plant’s reflectance information for retrieving
the Normalized Difference Vegetation Index (NDVI) during the
vegetative and reproductive growth stages.
Consequently, the field’s soil samples were examined for their
Nitrogen, Phosphorus, Potassium, and Carbon (NPKC) macronutrient
constituencies.
Finally, a Convolutional Neural Network-Regression model was
developed to predict infield NPKC spatiotemporal variations in soil
using the NDVI measurements.
The deep learning model effectively determined the surpluses or
shortages of the NPKC macronutrients within the cornfield
throughout the growth stages.
The model performed vigorously with R2 values of 0.93, 0.92, 0.98,
and 0.83 in predicting N, P, K, and C levelsin soil, respectively.
CONT…
54. CONT…
Figure 35. The UAV flight plan created
according to the field corn seedling
positions.
Figure 36 . Image acquisition, data preprocessing
and model trainingflowchart.
Jaihuni et al., 2021
55. Figure 37 . Basic structure of a
CNN-Regression model.
Figure 38. Comparison of the predicted and
actual macronutrient valuesof a) K b) P c) N d) C
CONT…
57. RADIATION- TARGET INTERACTIONS
1. Absorption (A)
2. Transmission (T)
3. Reflection (R)
Figure 38. Immature leaves contain less chlorophyll
than older leaves, they reflect more visible light and
less infrared radiations
Source: NRCAN, Canada
58. VARIABLE RATE OF N FERTILIZER RESPONSE IN WHEAT USING REMOTE SENSING
Figure 39. Spatial distribution of the
randomly determined N treatments
for the growing seasons 2008/09
Figure 40. Map of mean
grain yield for each N plot
for 2008/09 growing season
Basso et al., 2016
60. Figure 41. Soil salinity mapping using remote sensing
• The use of remote sensing data, followed by site observations is a powerful tool
in detecting salt affected areas.
• As nearly 80 % of saline areas could be delineated, it is a good indication for
the validity of the model. Hence, this model can be used in similar areas that
experience salinization problems. The simplicity of this model and acceptable
degree of accuracy make it a promising tool for use in salinity prediction.
• Combining these remotely sensed and ECe variables into one model yielded the
best fit with R2= 0.78.
Asfaw et al., 2018
SOIL SALINITY MAPPING
61. WEED MANAGEMENT USING RS
France
(a) Multispectral orthoimage;
(b) Crop (green) and weed (red) location deduced from spatial information;
(c) Weed (green) and crop (red) location deduced from spectral information;
(d) Weed (green) and crop (red) location deduced from the combination of spatial
and spectral information.
Figure 42. Example of the spatial and spectral combination
results using SVM classifier
Marine et al., 2018
Maize crop having weeds
like Chemopodium album,
Cirsium Arvence.
62. Carrot field having weeds like Chemopodium album, poa annua, Stellaria
• The label application for Glyfonova Plus ranges from 540 g/ha to 2880
g/ha depending on the types of weeds and weed pressure (Cheminova AS,
2005).
• A treatment scheme with the robot and DoD system, would consist of 2-3
treatments in combination with mechanical weed control in between the
rows. Building on the experience from the lab and field trials, we would
estimate a total application of 50 – 150 g/ha glyphosate. This represent a
ten- fold reduction in applied herbicide.
Utstumo, 2019
Norway
Figure 43. Visualization on Drop-on- Demand herbicide application
PRECISION FARMING
63. SOIL FERTILITY MAPS GENERATED FROM REMOTE SENSING
DATA
CPRS, Jalandha Ray et al., 2001
64. REMOTE SENSING TO DETECT NITROGEN DEFICIENCY IN CORN
Figure 44A. Corn where
nitrogen was applied due to
fertilizer applicator problem.
Corn on person’s side is 5 ft.
tall; to the left where nitrogen
was not applied is 3 ft. tall
Figure 44B. Remote sensing image
collected on 07/05/03 by aircraft
platform. Corn field outlined in
yellow. Low nitrogen area outlined
in black and confirmed by ground
scouting
Shanahan et al.,2001
65. PROBLEMS OF REMOTE SENSING IN INDIAN CONDITIONS
Small size of plots
Diversity of crops sown in a particular area
Variability of sowing and harvesting dates in different fields.
Intercropping and mixed cropping practices.
Extensive cloud cover during the rainy season
66. SUMMARY AND CONCLUSION
The deep learning model performed effectively in spotting fertilizer
surpluses or shortages within the field, providing RMSE levels of
0.09%, 0.16%, 0.09%, and 1.21% for N, P, K, and C, respectively,
Vegetation indices viz., NDVI resulted in better information compared to
SAVI in detecting the nutrients deficiencies of oil palm trees from
remotely sensed data,
Soil organic matter was mapped quantitatively using a spectrometer and
hyperspectral remote sensing with the prediction at 76 % confidence,
Up to 80% of saline areas can be delineated using remote sensing data.
Support Vector Machine (SVM) classifier was found to be effective for
classifying maize crop from the weeds in the field using both spatial and
spectral information.
67. As per remote sensing data, treatments applied with 70 kg N shown
good performance which was in good correlation with ground based
yield data while studying the variable Nitrogen rates fertilizer
response on durum wheat yield across the field.
From these studies, it is evident that remote sensing is an accurate,
time-saving technique with large coverage, non-destructive tool for
detecting nutrient deficiencies, SOM mapping, mapping of
problematic soils, weed management, crop yield, crop growth
development and identification of stressed crop etc.
CONT…
68. FUTURE LINE OF WORK
Future applications in Remote sensing should combine available
resources from space/ aerial/ UAV platforms with ground-based data.
The prerequisites of such resource integration are as follows:
(i) The spatial resolution of the satellite data should be high
enough to match ground-based data; for example, both spatial
data and ground data are in same order of accuracy.
(ii) Cloud- base calculations should support big dataset from
crowd- sourced remote sensing resources.
The current situation with the use of remote sensing shows promising
support for integration of multiple sources of remote sensing data.
We can expect to see new applications developing years to come.
69. The material for the presentation has been compiled from
various sources- books, tutorials, and several resources on
the internet.
knthebere@gmail.com