Precision Farming Approaches to Small-Farm Agriculture

Sakae Shibusawa
Faculty of Agriculture,
Tokyo University of Agriculture and Technology
3-5-8 Saiwai-Cho, Fuchu, Tokyo 183-8509, Japan, 2002-08-01

Introduction

Precison farming provides a new solution using a systems approach for today's agricultural issues such as the need to balance productivity with environmental concerns. It is based on advanced information technology. It includes describing and modeling variation in soils and plant species, and integrating agricultural practices to meet site-specific requirements. It aims at increased economic returns, as well as at reducing the energy input and the environmental impact of agriculture.

This Bulletin describes the concept of precision farming, and also its use in rural areas, including those with small-scale farms with diverse kinds of land use. A real-time soil sensor developed in our laboratory is also introduced.

What Is Precision Farming?

The term "Precision Farming" or "Precision Agriculture" is capturing the imagination of many people concerned with the production of food, feed, and fiber. It offers the promise of increasing productivity, while decreasing production costs and minimizing the environmental impact of farming (NRC 1997, SKY-Farm 1999).

Elements of the Technology

There are three fundamental elements in this technology (Fig. 1) (Shibusawa 2000, 2001).

Describing variability is the key concept. In particular, it is based on variation within each field. Variability should be understood in at least three aspects: spatial, temporal and predictive.

Variable-rate technology (VRT) is used to adjust the agricultural inputs according to the site-specific requirements in each part of the field. If machines are used, this requires variable-rate machinery. On small farms, inputs can be applied manually. Variable-rate applications need:

• Correct positioning in the field;
• Correct information at the location; and
• Timely operations at the site concerned.

Decision support systems offer a range of choices to farmers with respect to trade-off problems where conflicting demands must be taken into account, such as productivity and protection of the environment. This approach helps to optimize the whole farming system.

Information Levels

There are four levels or stages in the quality of information. The lowest level is data, followed by information, knowledge, and finally wisdom (Fig. 2) (Shibusawa 1999, 2000). The "data-stage" means a mass of signals and numerical values, which have no practical value in themselves. The "infor-mation-stage" provides some meaning from a set of data, such as levels of excessive, appropriate or deficient fertilizer use. The "knowledge-stage" implies that the information is individualized in some logical way which can enable someone to make a decision, such as application guidelines. The "wisdom-stage" belongs to an individual who finds an original solution, such as the approach of the skilled farmer.

Information technology tends to be powerful in levels up to the knowledge-stage. The wisdom-stage requires the intellectual and creative activities of farmers and researchers, if there is to be a break-through in accumulated knowledge. Precision farming needs all stages of information in the agricultural production system, and also requires good linkage between the stages. In particular, information technology should be closely linked to farmers.

Scenarios

Developing system technology for precision farming is shown in Fig. 3 (Shibusawa 1999, 2000). First of all, we need to describe and understand the variability within and between fields. Field sensors with GPS and monitors for machine application make this easier. The next stage is to develop machines which can be operated by remote control.

There are three steps in technology development, and three strategies for precision farming, as shown in Fig. 4 (Shibusawa 1999, 2000). Step 1 is based on conventional farming technology, with intensive mechanization to reduce the labor input. Step 2 involves the development of mapping techniques, VRT machines, and introductory DSS on the basis of information technology. Step 3 implies the maturity of wisdom-oriented technologies.

Scenario 1 is based on a "high-input and high-output" conventional strategy. Scenario 2 has a strategy for "low-input but constant-output", and Scenario 3 aims at "optimized input-output" as the goal of precision farming. Advanced technology levels allow us to choose freely between these three scenarios. Effective regulations will encourage progress in precision farming.

In Japan, technology at the moment is at the level between Step 1 and 2, while the farming strategy is changing from Scenario 1 to 2. The shift from Step 2 to 3 involves a drastic change in the farming system. In the mature stage of Step 2, each field has information added which makes possible the best management of that field. In Step 3, all factors of the farming system are well-organized for precision farming. This allows us to manage regional variability, as well as the local variability within a single field.

What Can Precision Farming Achieve?

In general, a farming system involves five factors (Shibusawa 1999, 2000). These are: plant variety, field features, technology, regional infrastructure, and the motivation/intentions of the farmer. Better integration of these five factors can creat a competitive farming system which suits local conditions.

Precision farming uses field maps, variable-rate technologies and a decision support system. As shown in Fig. 5, generating the field maps is in itself an important source of information. Variable-rate technology not only increases productivity by re-organizing the three factors of technology, plants and fields, but also creates a better linkage with the regional infrastructure, e.g. by following environmental regulations. A decision support system provides the best technology, taking into account the aims and motivation of farmers as well as environmental factors. In other words, precision farming brings about an innovation in the whole system of agriculture.

Precision Farming for Small-Scale Farms

Whether precision farming is feasible for small-scale farms is a leading issue for agricultural scientists and politicians in Japan. It should be noted that precision farming is characterized by variable management. A key point in precision farming is understanding variability in the field.

There are at least two types of variability. One is within-field variability, the other is between-field or regional variability. Within-field variability focuses on a single field, and the one plant variety being cultivated. Between-field variability considers each field as a unit on a map.

We need to consider what kind of variability is involved when we consider precision farming for small farms. Whether farms are large or small, precision farming should mean improved farm management. It should give a higher economic return with a reduced environmental impact.

On a single small farm, the farmer can understand fairly well what is going on in each field. This makes possible variable-rate applications to meet site-specific requirements, using the farmer's knowledge and skills. When it comes to an area of a few dozen hectares, containing many small fields, precision farming has to coordinate diverse types of land use and many farmers with different motivations, as shown in Fig. 6.

Regional precision farming must manage a hierarchy of variability: within-field, between-field and between-farmers. High-tech approaches, such as a yield meter with GPS, are available for regional precision farming covering many small farms. Moreover, measures to conserve or improve the environment should be undertaken on a similar scale.

From the point of view of development in a rural area which includes small farms and local companies, precision farming offers the possibility of developing a new kind of industry, by fusing agriculture to various kinds of industrial activity (Fig. 7).

If the multi-functions of agriculture are re-evaluated using information-added fields, value-added space of this kind can be seen as providing new resources, such as new biological materials, open-air classrooms and green tourism.

Real-Time Soil Sensing

In Japan, paddy fields occupy half of the agricultural area, producing about ten million metric tons of grain every year. Paddy rice production is a kind of hydroponic system, with well-organized irrigation and drainage facilities. In Japan, it is also highly mechanized.

Paddy rice production is very productive, but environmental concerns have become national issues in Japan. The problem is how to manage precisely the paddy production system while considering the environmental impact. Solving this problem requires a good understanding of what is happening in the paddy field.

Shonk et al. (1991) developed a portable soil organic matter sensor with photodiodes using a single wavelength. It gave good results in predicting soil organic matter in the range 1.5 - 6%. Sudduth and Hummel (1991) investigated the feasibility of spectral reflectance to sense soil organic matter. A portable NIR spectrophotometer was designed to evaluate soil organic matter, CEC and moisture content in a ploughed soil at a depth of 3.5 - 5 cm (Sudduth and Hummel 1993a,b). This approach can be useful to get information about the field surface, but we still need in situ soil sensing in the zone of root development for practical use in crop management. Shibusawa et al. (2000) have developed a real-time soil spectrophotometer with an RTK-GPS to sense underground soil parameters at depths of 15 - 40 cm.

The objective of this work was to use the soil spectrophotometer to generate detailed soil maps of the paddy field, for the implementation of precision rice farming.

Soil Spectrophotometer

The soil spectrophotometer used in this study is shown in Fig. 8. It was designed to collect data on soil reflectance at depths of 15 to 40 cm. The sensor system was composed of three main units: the external housing, the soil penetrator and probes, and the external sensing and monitoring devices.

The penetrator tip with its flat edge cuts the soil in a uniform way. The plane edge behind it smooths the soil to produce a uniform surface. Inside the housing are seven micro optical devices. Two optical fiber probes, using light energy at wave-lengths of 400 - 2400 nm, are used for illumination, giving an illuminated area of about 50 mm diameter on the soil surface. Two additional optical fiber probes are used to collect soil reflectance in the visible and NIR ranges. One fiber bundle passes reflected energy in the 400 - 900 nm wavelength range, while the other optical probe carries reflected energy in the 900 - 2400 nm wavelength range. A micro CCD camera is adjusted to monitor a 75-mm focus point on the soil surface.

The sensor unit's housing includes the core devices of the system, such as a 150 W halogen lamp, a spectrophotometer (Carl Zeiss Ltd.), a FA computer (IBM, PC/AT, Pentium MMX), a RTK-GPS (Trimble MS740) receiver, etc. The spectrophotometer has a 256-channel linear photodiode array to quantify the reflected energy in the 400 - 900 nm wavelength range. A 128 channel linear photodiode array is used to quantify the reflected energy in the 900 - 1700 nm wavelength range. Data scanning time is just over four microseconds. Integration of scanned data is carried out for each individual scan to get average values.

A video data recorder on the tractor displays images of the soil surface during the experiment. The displayed images are used to monitor operations in case of emergency, such as blockages or obstacles. The images also provide information about which data should be omitted from data analysis. The liquid crystal monitor serves as a touch control panel, and a mouse and keyboard are also available for accessing the FA computer.

Field Tests and Results

The experiment was conducted in a 0.5 ha paddy field on the Experimental Farm of Kyoto University, Japan in December 1999. The soil texture of the fields was 47% sand, 30% silt and 23% clay. The working speed was about one kilometer per hour. Scans were at approximately one-meter intervals. It took approximately 20 microseconds to do the scan, and three seconds to record the data. The spectrum data was collected at intervals of about 5 m, which gave more than 800 locations for soil reflectance data. The working depth was 200 - 250 mm.

For calibration purposes, 25 soil samples were collected at the same location and depth as the scanning points and analyzed in the laboratory for moisture, organic matter content, nitrate (NO₃-N), pH and EC. Fifteen samples were used for calibration, and the remaining ten samples were used for validation. A standard moisture content was obtained, by keeping the samples 24 hours in an oven at 110^oC. Soil organic matter content was evaluated as the loss after four hours of combustion in an oven at 800^oC. NO₃-N, pH and EC were analyzed in the clear layer at the top of muddy water, using portable ion meters. The test muddy water was provided by diluting 5 g dried soil with 25 g distilled water, stirring the mixture for 30 minutes, and then leaving it to stand for 24 hours.

For the spectral reflectance, four stages were followed (Marten and Naes 1987). The first stage was linearization with a Kubelka-Munk transform, while the second stage was elimination of optical interference in the spectral data with a multiplicative scatter correction. The third stage was to reduce the number of wavelengths used for calibration with correlation analysis, including derivative operation. The final stage was the calibration stage using the stepwise multiple linear regression analysis with S-Plus Data Analysis Software. The calibration model was quantified using the standard error for calibration (SEC), standard error for prediction (SEP) and coefficient of determination (R²).

Semivariance analysis was performed using the GS+ Geostatistics Software, and soil maps were obtained by the block kriging method.

Results of calibration and validation analysis (Table 1) produced higher scores of R² and fewer errors for the respective soil parameters. The second derivatives of light absorption tended to provide best-fit prediction models (I Made Anom et al. 2001).

With the prediction models, values for soil parameters were evaluated at 860 locations in the field. The means and standard deviations were 48.4% and 6.5% for moisture content, 9.51% and 1.06% for organic matter (OM) content, 42.1 mg/100g and 11.0 mg/100g for NO₃-N content, 6.83 and 0.39 for pH, and 173.1 µS cm and 69.6 µS/cm for EC. Based on these values, semivariance analysis (Table 2) was performed. Within the experimental field, the soil OM content had the lowest spatial correlation (29.20 m), followed by the NO₃-N content (34.50 m), moisture content (38.60 m), the pH (40.40 m), and the EC (46.60 m).

With the results of the semivariance analysis, the soil parameter maps were then developed (Fig. 9). The maps were interpolated by block kriging with 10-neighborhood interpolation. Errors of kriged to observed values were estimated over 40 grids, each 10 m square. Error means and standard deviations were -0.35 and 2.25 for moisture content, -0/01 and 0.46 for SOM content, 0.08 and 2.72 for NO₃-N content, -0.04 and 0.22 for pH, and 3.31 and 12.26 for EC (I Made Anom et al. 2001).

The distribution of variability in soil parameters shows some stripes running east-west. For example, there is a belt with a high moisture content and a high OM content in the eastern part. Other belts have a high NO₃-N, a high EC or a low pH. The irrigation inlet was located at the north-west, and the drainage gate at the south-east. This may have produced the striped effect, since water flowed from north to south.

Conclusion

Precision farming implies a management strategy to increase productivity and economic returns with an reduced impact on the environment, by taking into account the variability within and between fields. Variability description, variable-rate technology and decision support systems are the key technologies for precision farming. Precision farming on a regional level is one way to apply this approach to small-farm agriculture. It may not only improve farm management, but may also promote the development of rural areas.

A real-time soil spectrophotometer will be commercially available in a few years.

References

• I Made Anom, S.W., S. Shibusawa, A. Sasao, S. Hirako. 2001. Parameters Maps in Paddy Field Using the Real-time Soil Spectrophotometer. Journal of the Japanese Society of Agricultural Machinery, 63, 3: 51-58.
• National Researach Council (NRC) (1997). Precision Agriculture in the 21st Century. (Committee on Assessing Crop Yield: Site-Specific Farming, Information Systems, and Research Opportunities), National Academy Press, Washington DC., USA, pp. 149.
• Marten, H., T. Naes. 1987. Multivariate calibration by data compression. Near-infrared Technology in the Agricultural and Food Industries, P. Williams and K. Norris (Eds.). American Associatiion of Cereal Chemists, Inc. St. Paul, Minnesota, USA, pp. 57-87.
• Shibusawa, S. 1999. Environmentally-friendly agriculture and mechanization trend in Japan: Prospects of precision farming in Japan. Proc. International Symposium on Farm Mechanization for Environmentally-Friendly Agriculture, Seoul, Korea. The Korean Society for Agricultural Machinery, pp. 53-80.
• Shibusawa, S. 2000. Precision farming Approaches for model and control. Proceedings of the XIV Memorial CIGR World Congress, Nov. 28-Dec. 1, Tsukuba, Japan, pp. 133-141.
• Shibusawa, S. 2001. Precision farming: Approaches for small-scale farms. Proc. 2nd IFAC-CIGR Workshop on Intelligent Control for Agricultural Applications, 22-24 August, 2001, Bali, Indonesia, pp. 22-27.
• Shibusawa, S., A. Sasao, S. Sakai. 1999. Local variability of nitrate nitrogen in a small field. Precision Agriculture '99, Proc. 2nd Euro. Conf. Precision Agriculture, J.V. Stafford (Ed.). UK, pp. 377-386.
• Shibusawa, S., H. Sato, S. Hirako, A. Otomo, A. Sasao. 2000. A revised soil spectro-photometer. Proc. 2nd IFAC/CIGR International Workshop on BIO-ROBOTICS II, Nov. 25-26, Osaka, Japan, pp. 225-230.
• Shibusawa, S. et al. 2001. Soil mapping using the real-time soil spectrophotometer. Proc. (on CD-ROM) 3rd European Conference on Precision Agricultaure, Montpellier, France, June 18-20, 2001.
• Shonk, J.L., L.D. Gaultney, D.G. Schulze, G.E. Van Scoyoc. 1991. Spectroscopic sensing of soil organic matter content. Transactions of the American Society of Agricultural Engineers. 34, 5: 1978-1984.
• SKY-farm. 1999. Opportunities for precision framing in Europe: Updated report 1999. 126 pp.
• Sudduth, K.A., J.W. Hummel. 1991. Evaluations of reflectance methods for soil organic matter sensing. Transactions of the American Society of Agricultural Engineers. 34, 4: 1900-1909.
• Sudduth, K.A., J.W. Hummel. 1993a. Portable, near-infrared spectrophotometer for rapid soil analysis. Transactions of the American Society of Agricultural Engineers. 36, 1: 185-193.
• Sudduth, K.A., J.W. Hummel. 1993b. Soil organic matter, CEC, and moisture sensing with a portable NIR spectrophotometer. Transactions of the American Society of Agricultural Engineers. 36, 6: 1571-1582.
  

Index of Images

• 

  Figure 1 How Precision Farming Works

• 

  Figure 2 Level of Information

• 

  Figure 3 Development of Precision Farming Technologies

• 

  Figure 4 Development of Technology Level and Farming Strategies

• 

  Figure 5 Precision Farming Technologies Make Innovations in the Agricultural System

• 

  Figure 6 Regional Precision Farming (PF) for Small Farms

• 

  Figure 7 Role of Precision Farming in Regional Development

• 

  Table 1 Results of Calibration and Validation for Soil Parameter Prediction

• 

  Figure 8 Revised Soil Spectrophotometer

• 

  Figure 9 Soil Parameter Maps of a 0.5 Ha Paddy Field Using Soil Reflectance Collected by the Real-Time Soil Spectrophotometer. 860 Data Points at Depths of 200 to 250 MM Depth.

• 

  Table 2 Summary of Semivariance Analysis

Download the PDF. of this document, 567,468 bytes (554 KB).