Land Suitability Mapping for Implementation of Precision Farming

Norasmanizan Binti Abdullah School of Environmental Engineering,

Universiti Malaysia Perlis, Kompleks Pusat Pengajian Jejawi 3,

02600 Arau, Perlis, Malaysia E-mail: izan1510@yahoo.com

Mohd Sabri Hussin School of Manufacturing Engineering,

Universiti Malaysia Perlis, Kompleks Pusat Pengajian Jejawi 3,

02600 Arau, Perlis, Malaysia E-mail: mohdsabri@unimap.edu.my

Ayu Wazira Azhari School of Environmental Engineering,

Universiti Malaysia Perlis, Kompleks Pusat Pengajian Jejawi 3,

02600 Arau, Perlis, Malaysia E-mail: ayuwazira@unimap.edu.my

Mahmad Nor Jaafar School of Bioprocess Engineering,

Universiti Malaysia Perlis, Kompleks Pusat Pengajian Jejawi 3,

02600 Arau, Perlis, Malaysia E-mail: mahmad@unimap.edu.my

Abstract— This paper studies the potential of satellite imagery in mapping of soil compaction and land suitability using Remote Sensing and GIS method. GIS and remote sensing, play an important role in environmental study, monitoring and remediation which can be used for a great variety of application. Soil compaction is important part in crop growth and the major causes for reduced crop yield worldwide. The aim of this study is producing the potential of land suitability map through measuring soil compactness for implementation of precision farming. In addition, it analyzes the correlation between soil compaction data and pixel value of satellite imagery. For this study, soil penetrometer tool are used to record data manually while digital analysis uses satellite images. A Landsat 5 TM and SPOT 5 images of Perlis were acquired to illustrate the use of Remote Sensing to identify areas of possible soil compaction. Soil sampling data was collected and simultaneously recorded the coordinate using GPS handheld. The correlation was analyzed in identifying areas of potential soil compaction using linear regression and polynomial. The farmers can used the product of soil suitability map to identify areas suitable for particular crops. Furthermore, to increase crop production through the use of integrated remote sensing, GIS, GPS and sensor technologies simultaneously maintaining the quality of the environment.

Keywords- Soil Compaction, Remote Sensing, GIS, GPS, Precision Farming

I. INTRODUCTION

In Precision Farming concept, each crop production input is managed to reduce wastage of input, increase the profits and at the same time maintaining the good quality of the environment. [1] Numerous technologies have been applied to make agricultural products safer and to lower their adverse

impacts on the environment, a goal that is consistent with sustainable agriculture. This review highlights on remote sensing technology and describes how it can be used as an effective tool in precision agriculture. [2]

The use of remote sensing imagery may also provide important information that can reduce or eliminate the need for measuring soil cone index. To characterize the state of soil compaction in a field using soil cone index and to determine if remotely sensed images could identify compacted regions that require deep tillage. The results of linear correlation determinations between the various indices associated with the cone index measurements and the LANDSAT reflectance data as recorded in the grid cells of Ellis Field 31. While much better correlation must be established before remote sensing measurements can be used to with confidence to identify areas of likely compaction. However, additional strategies should be explored in a search for more useful remote imagery for identifying areas of potential soil compaction. [3]

Soil compaction is an important physical limiting factor for the root growth and plant emergence and is one of the major causes for reduced crop yield worldwide. The system may also incorporate GPS data to create geo-referenced soil maps. The data could be automatically stratified to determine soil compaction distribution at different layers of 5, 10,.…, 40 cm depths.[4]

The LU/LC map was produced by processing and interpreting a Landsat TM image. Soil map, topographic maps and ground truth data were used as ancillary data for image analysis. After image interpretation, the thematic map resulted in the following LU/LC classes: urban/developed, water,

agriculture, pasture and forest. A triangular irregular network (TIN) was interpolated from contour lines and elevation points (VIPs) considering drainage as breaklines. The slope map was created from DEM with 30 meters cell resolution and it was recorded in 5 classes. The soil map was converted into a grid. The combination of both LU/LC and LUC maps allowed the production of land-use suitability (LUS) map in which discrepancies could be detected as a result. This information can be used for decision-making, management and monitoring agricultural areas in order to avoid soil sustainability losses. [5]

The most common date for land use analysis are satellite data. They become very popular in recent years because of their better spatial and spectral resolution and their capacity to generate multitemporal products more cheaply than aerial photos. Besides that, land suitability analysis needs thematic maps such as soil, slope and rainfall maps. [6]

The land potential of the study area was categorized as very high, high, moderate, low and very low by adopting the logical criteria. These categories were arrived at by integrating the various layers with corresponding weights in a Geographical Information System (GIS). [7] The spectral signature of an object is usually represented by characteristics reflectance curves in the part of the solar spectrum, covered by satellite sensor (e.g. Landsat, SPOT). Soil and vegetation spectral signatures like those of other objects, vary according to parameters such as moisture and state of vegetation etc. [6]

The objectives of this project are to evaluate the potential of

satellite imagery in mapping purpose, classifying natural features based on spectral reflectance, land suitability analysis, and identify areas of possible soil compaction. Besides that, to explore the potential of Remote Sensing, GIS technologies, GPS in land suitability study for implementation of precision agriculture and land management.

II. MATERIALS AND METHODS

A. Data Acquisition i. Penetrometer (Soil compacter)

Fig. 1 shows the picture of soil penetrometer that use to indicate soil compactness and strength with various depths.

Figure 1.Penetrometer (Soil Compacter)

ii. Satellite images

Fig. 2 shows the area that involved in this study which located at Agrotechnology Research Station, Sg. Chucuh, Padang Besar, Perlis. As can be seen from Fig. 2 which is SPOT 5 image (A) and Landsat 5 TM image (B) was used for this study.

LEGEND 16 Acre 56 Acre 253 Acre (B)

(A)

Figure 2.Study Area

iii. Handheld GPS Use to locate & record coordinate during field

sampling. The data was uploaded and display using Garmin Map Source. iv. Ancillary Data

There are several maps that needed for this study such as Topographic map, Soil Map, Agriculture Rainfall Index Map and Land Use Map. v. Software

ERDAS Imagine 9.2 was used in processing satellite images data and ArcGIS 9.3 was used as analyzing tool. Besides that, SPPS 16.0 and Microsoft Excel were used for statistical analysis.

B. Map Accuracy Standard The U.S. Geological Survey (USGS) publishes maps and

other products at high levels of accuracy. The accuracy level is important to engineers, surveyors, highway officials, and land-use planners who use for mapping purpose. The values are used to determine the RMSE of the X, Y coordinate, as follows:

RMSE planimetric = ( ) ( )∑⎥⎥⎦

⎤

⎢⎢⎣

⎡ −+−n

iiiii

nyyxx 2*2* (1)

Where n is the number of checkpoints; ( ii yx , ) are coordinate of the point measured and ( ** , ii yx ) are coordinate of checkpoint i respectively.

C. Digital Image Processing The digital image processing is the task of processing and

analyzing the digital data using some image processing algorithm. There are several parts of digital image processing that involves in this study.

i) Digital Image Pre-Processing ii) Digital Image Enhancement iii) Digital Image Classification iv) Data merging and GIS integration

For this study, pre-processing part involves image

rectification, reprojection and registration. Image pre-processing must be done before do another processing such as enhancement, manipulation, interpretation and classification of satellite images. Rectification and reprojection of satellite imagery is performed to MRSO or Malayan Rectified Skew Orthomorphic which is coordinate systems commonly used in Peninsular or West Malaysia.

Before starting visually interpreting satellite images, some image enhancement techniques are applied to improve and enhance features. This processing involves image data fusion, image stacking, Wetness Index and NDVI (Normalized Differential Vegetation Index).

Digital image classification techniques are less time consuming than visual techniques. Digital satellite image can be classified digitally using supervised or unsupervised of image classification. The latest land use maps can be produced using this method for land suitability analysis and mapping purpose.

Besides that, ArcGIS software is very useful as analyzing tool, map overlaying and digitizing tool. Data integration involves combination data in different sources such as map, satellite imagery, aerial photograph, database and so on. Fig. 3 and Fig. 4 show the potential of GIS in data integration and data merging.

Figure 3.Data integration from different sources

Figure 4.GPS sampling points overlay with Spot 5 image

D. Spectral Reflectance Different materials reflect and absorb differently at

different wavelengths. Thus, the targets can be differentiated by their spectral reflectance signatures in the remotely sensed images.

III. RESULTS AND DISCUSION

A. Pre-processing Fig. 5 and Table 1 show the results of RMS error after

geometric correction using image-to-image transformation method. Geometric correction is the process of resampling a data set to remove undesirable or misleading distortions. Spot 5 images was used as reference image to geocode with Landsat 5 TM.

Figure 5.Graph of RMS Error

RMSE is the square root of the average of the set of squared differences between dataset coordinate values and coordinate values from an independent source of higher accuracy for identical points. Accuracy is reported in ground distances at the 95% confidence level. The acceptable of total RMSE should be less than 1 pixel.

TABLE I. RMS ERROR Pre-processing part is important before run another

processing and analysis. Fig. 6 shows the result of mosaicking which is join two images together. In order to analyze remotely sensed images, the different images representing different bands must be stacked.

Figure 6.Image Mosaic

B. Spectral Profile Fig. 7 shows the potential of satellite image data in

classifying the natural features with different band wavelength. The spectral profile shows the band wavelength of band 1- band 4 for Spot 5 image with different location of GPS points.

Figure 7.Spectral profile of Spot 5

C. Regression Method Table II and Table III show the results of correlation

between soil compaction data and pixel value of satellite images using the linear correlation methods. It shows that several bands of SPOT 5 and Landsat TM representing low correlation. However, the highest correlation of Landsat TM is 0.686 (Band 3) and for Spot 5 is 0.6041 (Band 4). From the

Table II, shows Band 1 and Band 2 of Landsat TM also

give high correlation but the correlation decrease with depth of ground level. It can be concluded that, satellite imagery are limited in penetrating ground surface.

TABLE II.LINEAR CORRELATION BETWEEN PIXEL VALUE AND

SOIL COMPACTION OF LANDSAT 5 TM

R² for (5cm)

ground surface R² for (10cm) ground surface

R² for (15cm) ground surface

Band 1 0.641 0.474 0.463 Band 2 0.604 0.399 0.358 Band 3 0.686 0.561 0.534

Band 4 0.002 0.185 0.206

Band 5 0.037 0.516 0.490 Band 6 0.018 0.167 0.186 Band 7 0.281 0.304 0.328

TABLE III.CORRELATION BETWEEN PIXEL VALUE AND SOIL

COMPACTION OF SPOT 5

R² for (5cm)

ground surface R² for (10cm) ground surface

R² for (15cm) ground surface

Band 1 0.106 0.018 0.001 Band 2 0.082 0.026 0.007 Band 3 0.272 0.047 0.017 Band 4 0.604 0.498 0.441

Fig. 8 shows the correlation between pixel value and soil compaction data of SPOT 5 image and Landsat 5 TM using Microsoft Excel for 5 cm level of ground surface.

(a)

X Residual Y Residual Total RMSE Control

Point Error 0.4939 0.4118 0.6431

(b)

Figure 8.Linear Correlation between pixel value and soil compaction (a) Linear Correlation of Landsat TM (b) Linear Correlation of Spot 5

Table IV and V show the results of correlation using

polynomial method. The highest correlation of Landsat TM is 0.663 (Band 3) and for Spot 5 is 0.752 (Band 4).

TABLE IV.POLYNOMIAL CORRELATION BETWEEN PIXEL

VALUE AND SOIL COMPACTION OF LANDSAT 5 TM

R² for (5cm)

ground surface R² for (10cm) ground surface

R² for (15cm) ground surface

Band 1 0.659 0.479 0.463 Band 2 0.637 0.400 0.370 Band 3 0.663 0.586 0.544 Band 4 0.158 0.276 0.251 Band 5 0.040 0.185 0.213 Band 6 0.116 0.348 0.306 Band 7 0.429 0.304 0.328

TABLE V.POLYNOMIAL CORRELATION BETWEEN PIXEL VALUE

AND SOIL COMPACTION OF SPOT 5

R² for (5cm)

ground surface R² for (10cm) ground surface

R² for (15cm) ground surface

Band 1 0.201 0.142 0.171 Band 2 0.116 0.076 0.101 Band 3 0.295 0.123 0.138 Band 4 0.752 0.540 0.441

Fig. 9 shows the better results of correlation between

pixel value of SPOT 5 image and soil compaction data using polynomial method.

(a)

(b)

Figure 9.Polynomial correlation between pixel value and soil compaction (a) Polynomial Correlation of Landsat TM (b) Polynomial Correlation of SPOT 5

D. Soil Compaction Map This map was created manually using soil compaction data

that collected using soil cone penetrometer for 5cm level of ground surface. Fig. 10 shows the soil compaction map was mapped using Arc Map.

Figure 10.Soil Compaction using Manual Method

IV. CONCLUSION Image processing is a vital part of remote sensing

operations. There are several parts that involved in digital image processing such as image rectification, image enhancement, image classification and GIS integration. The Root Mean Square Error (RMSE) is important to estimate positional accuracy. Satellite imagery and GIS have potential in doing spectral analysis, land suitability and for mapping purpose. From the correlation analysis, it shows that have correlation between soil compaction and pixel value. Several methods of regression were applied to find out the best correlation. In this study, polynomial and linear method was applied in the correlation analysis.

ACKNOWLEDGMENT This work was conducted in the Agrotechnology

Research Station, UniMAP Sg. Chucuh Campus, located at Padang Besar, Perlis. The authors would like to thank the individuals who were involved in making this work possible. Special thanks to supervisor, family and husband for their support.

REFERENCES [1] N. N. Mahmood, D. Ahmad , 2001. “Use of Remote Sensing and GIS

technologies in Precision Farming for rice crop in Malaysia,” Remote Sensing News, Newsletter of the Malaysian Centre for Remote Sensing.

[2] S. Liaghat. and S.K. Balasundram, 2010. “A Review: The Role of Remote Sensing in Precision Agriculture,” American. Journal of Agricultural and Biological Sciences 5 (1): 50-55.

[3] L.G. Wells, S.A. Shearer, J.P. Fulton and L.W. Murdock, 2000. “Assessment of Remote Sensing for Implementation

of Precision Tillage,” ASAE Paper No. 001084. Annual International Meeting, Midwest Express Center, Milwaukee, Wisconsin, July 9-12.

[4] Amacher, Michael C. O'Neill, Katherine P. 2004. “Assesing Soil Compaction on Forest Inventory and Analysis phase 3 field plots using a pocket penetrometer,” Research Paper RMRS-RP-46WWW. Fort Collins, CO: U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station. 7 p.

[5] G. Ippoliti, M. O. Mortara, A. C. Rezende, M. S. Simoes, M. V. Filho. 2000. “GIS and Remote Sensing Techniques Applied for Land-Use Suitability Study,” International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7.

[6] E. H. Harahsheh, 1994. “Agricultural Applications of Remote Sensing and Geographic Information System in Land-use and Land Suitability Mapping,” GISdevelopment.net, AARS

[7] M.H. Al-Mashreki, J.B.M. Akhir, S. Abd Rahim, K.M. Desa and Z.A. Rahman, 2010. “Remote sensing and GIS application for assessment of land suitability potential for agriculture in the IBB governorate, the Republic of Yemen,” Pak. J. Biol. Sci., 132: 1116-1128. .