You are on page 1of 6

IDENTIFICATION OF RELATIONSHIP ON SOIL NUTRIENT (N, P, K)

OF HARUM MANIS MANGO (MA 128)

Ernieza Suhana Mokhtar 1 Nazalyyussma Yusop 2 Roseliana Mohammad Saufi 1, Siti Aminah Anshah1,
Masayu Norman1, Norfariza Zainal1
Roseliana Mohammad Saufi (*)
1
Faculty of Architecture, Planning and Surveying, UiTM PERLIS
2
Faculty of Plantation and Agrotechnology, UiTM PERLIS
Email: liana_pinkrosas@yahoo.com

KEY WORDS: GIS, Geostatistical analyst, Ordinary Kriging, Nitrogen, Phosphorus, Potassium

ABSTRACT: The integration of ArcGIS and Geostatistical analyst tool is used in order to predict optimal
condition for efficient and more reliable production of the crops which is Harum Manis (MA 128). Soil fertility
is very important which applied for great productivity of plant. It must be applied in reasonable amount and as
well as with suitable balance. Due to laboratory cost analyst are very high in determining all the macronutrients
which is Nitrogen (N), Phosphorus (P) and Potassium (K) at once, resulting the contribution of algorithm which
helps laboratory analyzer in observing the analyst of this macronutrient in order to reduce the lab cost and time.
The aim of this study is to develop regression model for soil nutrient of Harum Manis crop. Regarding the aim,
the objectives contributed are i) to develop the correlation between predicted and measured of the Harum Manis
crop (Nitrogen, Phosphorus and Potassium); ii) to verify the soil nutrients using regression model; iii) to
compare interpolation map based on prediction value from regression and actual value. The study area for this
research was selected in the Harum Manis crop in UITM Perlis. This study only concentrates on relationship
between N, P and K soil elements. For overview, the methodology implemented for this research involves i)
collecting data which is soil samples data, ii) data processing which is generate linear algorithm and interpolation
map and iii) data analysis which include the correlation soil nutrients, verification of develop regression and
comparison the interpolation map based on actual values and predicted values from regression. The result shows
that the coefficient of determination (R2) and RMSE of relationship between N and P is R2 = 0.419 and RMSE =
± 0.0876, P and K is R2 = 0.058 and RMSE = ± 0.1509, and N and K is R2 = 0.102 and RMSE = ± 0.0252. The
linear regression is used to verify the predicted soil nutrients and the verification between predicted and
measured values of N, P and K is determined using RMSE where the results obtained for first, second and third
linear algorithm are RMSE = ±0.010, ±0.008 and ±0.007 respectively. The range of the prediction map based on
predicted values from regression are between 0.075 – 0.371 % for N, 0.104 – 0.168 ppm for P and 0.183 – 0.296
ppm for K. While, the range for prediction map using actual values are between 0.037 – 0.390 % for N, 0.069 –
0.304 ppm for P and 0.090 – 0.469 ppm for K.

1. INTRODUCTION

Harum Manis product of mango can only be obtain from Perlis and this state also is one of the biggest
manufactures of the mango product especially Harum Manis type. The study of this kind of crops is carry out
where it can give benefit not only to the planters but also Agriculture Department and as well as public.
Perlis is known well of their Harum manis mango and widely planted by most of the farmers in their farming
practice and this type of fruit contribute great earnings when it comes to their season which is two times per year.
Therefore, it is suitable for make research in this stated focus on UITM Perlis farm of Harum Manis. The need
for timely and reliable agricultural information has become more important in decision making. A complete
knowledge of the basic soil resource is of primary importance for efficient land use planning. Soil fertility is very
important which applied for great productivity of plant. It must be applied in reasonable amount and as well as
with suitable balance with also taken some consideration of climate zone which give moisture, light, and
sufficient heat. This means, the soil of Harum Manis should be applied correct quantity of soil nutrient, which
are Nitrogen, Phosphorus and Potassium in order to growth. Those major nutrients are needed and important to
plants. Excessive and deficiency of these nutrients might reduce the productivity of plant itself. The farm
management for apply the N, P and K need to be organize well in promoting the productivity of plant.

The problem is, the state demands maximum possible output of Harum Manis in each unit of cultivated
land area. The soil fertility changes due to cropping, manure, and fertilizers application. Thus, the soil fertility
evaluation requires a unified method and it should be generalized. One of such method is via soil test.
Supposedly would be to sample every farm of Harum Manis crop. The more soil samples taken in determining
the soil nutrients needed by Harum manis, the better the result for applying correct quantities of soil nutrients for
growth purpose. However, we may not be able to sample each farm due to too costly and time consuming,
especially with multiple small farm holdings. Soil test results from one farm have to be linked with broader
scope of the population of all farms of Harum Manis.

On behalf of solving the problem, the method is by generate the algorithm which known one of the
element value and later capable to identify other element values. Once, the soil fertility algorithm is created, it is
possible to transform the information of fertilizers that needed by soil. Such algorithm will provide site-specific
recommendation without testing the soil for all Harum Manis farm. The average values of soil samples on the
ground obtained help in avoided the excessiveness and deficiencies of N, P and K and later on lead the Harum
Manis in well growth.

2. MATERIALS AND METHODS

2.1 Planning

The study is held in the UITM Perlis where the Mango crops especially Harum Manis is planted. For the
study area, the plantation area is approximately 1.619 hectare. On behalf of implementing this project, several
software that has been use. The software used is ArcGIS software and Microsoft excel. The extension from
ArcGIS software which is geostatistical analyst tools is used for processing the soil elements dataset in
interpolating the prediction map using Ordinary Kriging technique. Whereas, Microsoft Excel is used in order to
perform analysis of the research where to identify the relationship between N, P and K respectively. The
relationship of those soil elements is determined by regression, R² in the Microsoft Excel and linear regression
algorithm also generated by using this software.

2.2 Data acquisition

For this phase, at first, the data is determined primarily which will be used in the project. The data that
required for this research are the surveyed area for Harum Manis Crop in UITM Perlis, soil samples for Nitrogen
(N), Phosphorus (P), and Potassium (K) dataset with coordinates. The data will be work with the ArcGIS
software where the powerful statistical analyst tool, geostatistical analyst tools were planned to be used and other
tools as well. The data that required and descriptions are as follows:

Table 1: Data acquired


Data Description
1. Harum Manis (MA 128) surveyed plantation These data already exists and some are taken from plantation
plan section.
2. N, P, K soil element datasets

1. Coordinate of added soil samples. The coordinate of soil samples was obtained via GPS
observation using MyRTKnet technique. The observation is only
for samples soil that does not have the coordinate values.

2.3 Data Processing

The data processing is involved and begins with determine the linear algorithm of N, P and K, followed
by identifying the correlation and verification and later on performs interpolation in the ArcGIS software.
The linear algorithm is used in order to identify the relationship of N, P and K. The linear algorithm is generated
by making the regression in identifying the relationship of soil elements between N and P, P and K, and N and K
respectively. The ground values of N, P and K are used on behalf of determining the algorithm itself. The
generated linear algorithms are as follow:
Table 2: Linear Algorithm
No Equation
1 N predicted = a0 + a1P measured
2 P predicted = a0 + a1K measured

3 K predicted = a0 + a1N measured

In order to identify the relationship of N, P and K, the coefficient of determination (R2) is determined in
the Microsoft Excel sofware. The relationship of soil nutrients is based on 86 soil samples. The relationship is
identified between N and P, P and K and N and K. Moreover, the RMSE also calculated to identify the different
values that from estimated by model and true value. The RMSE is calculated using Equation 1. Root Mean
Square Error (RMSE):

RMSE = ±√ (Σ (X 1 – X2)2 / n)
Equation 1

Verification is made on each soil elements of N, P and K using other 34 soil samples. The verification is
carried between predicted N, P and K using linear algorithm created and the measured N, P and K which obtain
from ground sample. RMSE is calculated in the Microsoft excel for the verification. On behalf of generating
interpolation map, the soil samples of verification are overlaid on the surveyed area plan for MA 128 in the
ArcGIS software and later on the prediction map based on actual values and prediction map based on regression
is generated.

3. RESULT AND ANALYSIS

The result summarized in Table 3 shows the coefficients of determination (R2) and RMSE obtained based
on relationship between N and P, P and K and N and K.

Table 3: Relationship of soil elements


No Relationship between soil Equation R2 RMSE
nutrients
1 N&P Y = 0.921 x + 0.011 0.419 ±0.0876
2 P&K Y = 0.169 x + 0.089 0.058 ±0.1509
3 N&K Y = 0.321 x + 0.171 0.102 ±0.0252

Using linear algorithm, the relationship between soil nutrients have shown that the N and P correlates
moderately as compared to others. There is no relationship between P and K while low relationship between N
and K. The correlation of soil nutrients shows low in R2 because of each element which is N, P and K are
standalone. The result summarized in Table 4 shows the RMSE obtained based on verification made for soil
nutrient of N, P and K using the regression generated.

Table 4: RMSE of verification using regression models (relationship between predicted and
measured soil nutrient)
No Verification RMSE
1 Measured N with predicted N using ±0.010
Y = 0.921 x + 0.011
2 Measured P with predicted P using ±0.008
Y = 0.169 x + 0.089
3 Measured K with Predicted K using ±0.007
Y = 0.321 x + 0.171

Using linear regression algorithm, the relationship between predicted and measured soil nutrient shown
low RMSE once verify using regression models generated where RMSE = 0.010 using linear algorithm from
relationship of N and P, RMSE = 0.008 using linear algorithm from relationship P and K, and RMSE = 0.007
using linear algorithm from relationship N and K. The lower the RMSE, the better the result will be.
Figure 4.1 (a) shows the nitrogen prediction map based on actual values. The map illustrates smallest
value of predicted N based on actual values is 0.037 % while the highest value of predicted N is 0.390 %.
Moreover, the maps illustrates most area of the predicted values are between 0.108 – 0.178 % while least area
predicted values are between 0.319 – 0.390 %. While Figure 4.1 (b) shows the nitrogen prediction map based on
predicted values from regression. The map illustrates smallest value of predicted N based on predicted value
from regression is 0.075 % while the highest value of predicted N regression is 0.371 %. Moreover, the maps
illustrates the most area of the predicted values are between 0.161 – 0.204 % while least area predicted values
are between 0.247 – 0.371 %.

Figure 4.2 (a) shows the phosphorus prediction map based on actual values. The map illustrates smallest
value of predicted P based on actual values is 0.069 ppm while the highest value of predicted P is 0.304 ppm.
Moreover, the maps illustrates the most area of the predicted values are between 0.163 – 0.210 ppm while least
area predicted values are between 0.257 – 0.304 ppm. While figure 4.2 (b) shows the phosphorus prediction map
based on predicted values from regression. The map illustrates smallest value of predicted P based predicted
value from regression is 0.104 ppm while the highest value of predicted P regression is 0.168 ppm. Moreover,
the maps illustrates the most area of the predicted values are between 0.117 – 0.130 ppm while least area
predicted values are between 0.143 – 0.155 ppm.

Figure 4.3 (a) shows the Potassium prediction map based on actual values. The map illustrates smallest
value of predicted K based on actual values is 0.090 ppm while the highest value of predicted K is 0.469 ppm.
Moreover, the maps illustrates the most area of the predicted values are between 0.165 – 0.241 ppm while least
area predicted values are between 0.317 – 0.393 ppm. While, Figure 4.3 (b) shows the Potassium prediction map
based on predicted values from regression. The map illustrates smallest value of predicted K based predicted
value from regression is 0.183 ppm while the highest value of predicted K is 0.296 ppm. Moreover, the maps
illustrates the most area of the predicted values are between 0.206 – 0.228 ppm while least area predicted values
are between 0.273 – 0.296 ppm.

(a) (b)

Figure 4.1: (a) N Prediction Map based on Actual values and (b) N Prediction Map based on Predicted Values
from regression
(a) (b)
Figure 4.2: (a) P Prediction Map based on Actual values and (b) P Prediction Map based on Predicted Values
from regression

(a) (b)
Figure 4.3: (a) K Prediction Map based on Actual values and (b) K Prediction Map based on Predicted Values
from regression
Based on the result above, the Nitrogen prediction maps values from regression can be useful for
predicted. This is supported by parameter given by Department of Agriculture Teluk Chengai, Kedah which
stated that the best parameter for Nitrogen is between 0.150 – 0.260 % and 0.270 – 0.400 %. According to
Yussnazalida Yusop (2012), the value of 0.183 – 0.296 is considered high for potassium classification. Plus, she
said that, higher potassium will help the plant growth.

4. CONCLUSION

Precision agriculture had introduced better management in farming practice. The relationship of soil
nutrients is carried out to generate the linear regression so that the laboratory costs are reduced. The relationship
between N and P is R2 = 0.419 with RMSE= ± 0.0876, the relationship between P and K is R2 = 0.058 with
RMSE = ± 0.1509, and the relationship between N and K is R2 = 0.102 with RMSE = ± 0.0252. The verification
made using linear algorithm relationship between N and P, P and K, and N and K which using predicted and
measured values contributed RMSE = ±0.010, RMSE = ±0.008 and RMSE = ±0.007 respectively. While, it was
found that the range for prediction map based on actual values are between 0.037 – 0.390 % for N, 0.069 – 0.304
ppm for P and 0.090 – 0.469 for Potassium and the range for prediction map based on predicted values from
regression are between 0.075 – 0.371 %, 0.104 – 0.168 ppm, and 0.183 – 0.296 ppm. In conclusion, the
relationship soil elements of N, P and K is done in order to improve soil fertility practice towards managing
better farming practice which also helps rise up the crop yield. Analyzing all of these elements which is N, P and
K might cost the farm unit. Therefore, by generated these linear algorithms, it expected to reduce lab works
which can be use in seek out the predicted values.

BIBLIOGRAPHY

Agricultural Research and Extension Council of Alberta. (2010). Precision Farming and Variable Rate
Technology. Alberta: Agricultural Research and Extension Council of Alberta.

Aida Nasirah. (2009). Mango Harum manis. Retrieved May 16, 2013, from http://adrea-
cranberries89.blogspot.com/2013/05/stid-part-2-mango-harum-manis.html

Chang, K. T. (2012). Introduction to Geographic Information Systems. New York: McGraw-Hill.

Che Ammar Faizal Che Hamzah. (2011). Fruit Recognition System (Harum Manis Mangoes). Melaka:
University Teknikal Malaysia.

Conjecture Corporation. (2013). Top 5 fact about mangoes. Retrieved April 5, 2013, from wiseGEEK:
http://www.wisegeek.com/what-is-a-mango.htm

Dubey, A. K., Singh, A.K., Srivastav, M. (2007). Salt stress studies in mango. Agricultural Reviews , 75-78.

Elprince, A. M. (2009). Prediction Soil Fertilization Maps Using Logistic Modeliing and a Geographic
Information System. Department of Soils and Agricultural Chemistry , 2032-2042.

ESRI. (2001). ArcGIS Geostatistical Analyst: Statistical Tools for Data Exploration, Modeling and Advanced
Surface Surface Generation. New York.

ESRI. (2013, Jun 24). ArcGIS Help. Retrieved January 4, 2013, from ArcGIS Resource Center:
http://help.arcgis.com/en/arcgisdesktop/10.0/help../index.html#/Introduction_to_the_ArcGIS_Geostatist
ical_Analyst_Tutorial/0031000000nz000000/

Genu, A. M. & Dematte, J.A.M. (2009, August 17). Prediction of soil chemical attribute using optical remote
sensing. Retrieved December 6, 2013, from Scielo: http://www.scielo.br/scielo.php?pid=S1807-
86212011000400024&script=sci_arttext

Haifa Group. (2011). Controlled release fertilizer in precision agriculture. Retrieved November 24, 2013, from
Knowledge Center & Community: http://www.haifa-
group.com/knowledge_center/articles/haifa_articles/controlled_release_fertilizer_in_precision_agricult
ure.aspx

Yussnazalida Yusop. (2012). Mapping of Soil Fertility for Harum Manis Tree (MA 128) using GIS Approach.

You might also like