[page 2↓]

2.  State of Art

2.1. Precision Farming in General

The variability of soils, and hence the variability of crops, was probably known and used all the time by the farmers as it is still done by the Philippine rice farmers some of whom differentiated fertilizer inputs according to the visual impression of the crop. In the early part of the 20th century, scholars were already studying variability in soil properties, such as nutrient status and organic matter levels (ROBINSON & LLOYD 1915). Researchers continued to report on soil and yield variability (FAIRFIELD SMITH 1938, REED & RIGNEY 1947, JAMES & DOW 1972, BURGHARDT 1984, SCHMIDT 1985, KNEIB 1985). The renown German MITSCHERLICH (1933, p 2) was the first to call the soil a “heterogeneous mixture”, detectable even on a small scale (MITSCHERLICH 1950a, p 14), thus biasing plot trials. He found soils being highly different at a distance of 10 m (MITSCHERLICH 1950b, p 297), and concluded the value of small scale maps “illusory” for practical farming at that time. The mechanisation of agriculture and the trend towards larger implements and fields led agriculture in the direction of large-scale farming.

In the last two decades, technological advances and the pressure of public opinion about environmental effects re-vitalised the idea of defining smaller management units, and applying agrochemicals based on the individual characteristics of those units, in the concept now generally referred to as precision farming or site-specific agriculture. The concept of precision farming is not a product of engineering technology, but rather has its foundations in the application of sound agronomic principles on an area-to-area, point-to-point basis. The practical implementation of precision farming concepts, however, is dependent on technological developments to provide, manage, and utilise the information on field heterogeneity.

Depending on the way of processing, these information can be managed either on an off-line basis, where “the input data ... are collected and held until a convenient later time for processing”, or on an on-line basis, where “the input data are processed as they become available” (WEBSTER 1999a). Based on the needed response time, on-line sensors are further called real-time sensors, “if response time is an important determination of correct functioning” (WEBSTER 1999b).

The practical implementation of site-specific agriculture, and therefore, the technological development is not limited to a specific research area. Furthermore, many research areas are interacting with each other on site-specific topics. Therefore, this state of art will also mention major works in the more important areas of site-specific agriculture.


[page 3↓]

Finding a location precisely again or fixing data precisely to a location, forms the backbone of precision farming, and without such a system, precision farming is unthinkable. Most of the current precision farming efforts use the Global Positioning System GPS, or differential Global Positioning System DGPS, to provide location data (LEICK 1990, AUERNHAMMER & MUHR 1991, MUHR et al. 1994, STAFFORD et al. 1996, OTT 1998, HALE & WENDTE 1999).

Soil and yield were the first to come into the focus of precision farming research. For obtaining soil data and quantifying specific soil properties, two approaches are in use: traditional soil sampling, and the development of sensors for either remote sensing or machine-based sensing. Traditional soil sampling is highly accurate yet time-consuming (ALTMANN & HAASE 1984, DAHIYA et al. 1984, ARMSTRONG 1986, DOBERMANN 1994, DOBERMANN et al. 1994, DOBERMANN et al.1995, DOBERMANN et al.1997) which is inadequate for soil parameters that change rapidly both spatially and temporally, and hinders the impact in practical farming. Therefore are the technicians focussing on remote sensing of soil properties – which are mostly hindered by weather conditions (MORAN et al. 1998, DIKER & BAUSCH 1998), and machine-based sensors. Several machine-based soil sensors exist, such as a real-time nitrate sensor (ADSETT & ZOERB 1991, BIRRELL & HUMMEL 1997), automated soil sampling collection systems (MC GRATH et al. 1995, LÜTTICKEN 1998a, COLBURN 1998, HALE 2000), sensing of soil slope (GEBBERS & SCHMIDT 1999), sensing of soil organic matter content via reflectance (PAGE 1974, KRISHNAN et al.1980, GRIFFIS 1985, SUDDUTH et al. 1990, SHONK et al. 1991), estimating soil moisture (KANO et al. 1985), soil compaction (DOMSCH & WENDROTH 1997) and soil electromagnetic sensing (WILLIAMS & HOEY 1987, KACHANOSKI et al. 1988, RHOADES et al. 1989, SHEETS & HENDRICKX 1995). These approaches face the problem of changing weather conditions both in the season and between the years (LAMB et al. 1997) and a sometimes low correlation between yield and soil nutrient levels (BIRRELL et al. 1993), and soil parameters can not be used for a site-specific crop protection, since there is almost no relationship between the soil and leaf fungi infestation.

Crop yield and its within-field variation is an important input for site-specific decision making and is in itself an integrator of many varying crop and soil parameters, thus becoming a main focus for engineers to develop crop yield sensors integrated in harvesters, such as grain flow sensors (WAGNER & SCHROCK 1987, PETERSON et al. 1989, SEARCY 1989, PANG & ZOERB 1990, STAFFORD et al. 1991, VANSICHEN & DE BAERDEMAEKER 1991, BORGELT & SUDDUTH 1992, KLEMME et al. 1992, PFEIFFER et al. 1993, STOTT et al. 1993, IIDA et al. 1998), or sensors for determining silage yield (VANSICHEN & DE [page 4↓]BAERDEMAEKER 1993, BALASTREIRE et al. 1997). Here is the yield data the decision base for the future treatments, but correlation between seasons might be low (BIRRELL et al. 1993).

Variable-rate technology for application control to differentiate crop inputs automatically while traversing a field has been in development for some time to control the amount of liquid and dry fertilizers, herbicides, seeds, and manure (CAHN 1995, OLIESLAGERS et al. 1995, GOENSE 1998, LÜTTICKEN 1998b, THIRION et al. 1998, SEYMOUR et al. 2000). Control of field sprayers was at first a mere on/off control and later became a variable rate control (FELTON & MC CLOY 1992, BALSARI & TAMAGNONE 1997, SCARR et al. 1997, VIERI & SPUGNOLI 1997). Without variable rate technology, an advance in site-specific agriculture is handicapped, because a decision to differentiate the application rate needs the techniques to apply these different rates.

The economic situation of precision farming is mainly unexplored. REETZ & FIXEN 1995 reported a net income of US-$ 43 per hectare, or a total of US-$ 2 798 for a 65 ha farm. And according to SCHMERLER & BASTEN 1999 ranges the economic benefit of site-specific management between + 15 EURO ha-1 and + 50 EURO ha-1. These results enforce the development of sensors that are able to scan at low costs per hectare to leave the benefit with the farmers. Only low-cost sensors can bear this economic pressure for sensing temporal and spatial variability, or sensors whose measurements have a low dependency on time, and thus, are suitable for various management dates or even seasons. Real-time sensors can provide a sampling intensity several magnitudes higher than traditional methods, and therefore, can tolerate a much higher analysis error while providing greater overall accuracy in mapping variability. These sensors can potentially reduce costs while improving the accuracy of the data collection process. In site-specific application of nitrogen top-dressing several machine-based real-time sensors were developed using image vision (SINGH et al. 1996) or leaf reflectance technology (SOLIE et al. 1996, STONE et al. 1996, REUSCH 1997, STAFFORD & BOLAM 1998, GÜNTHER et al. 1999, NIELSEN et al. 1999, WOLLRING & REUSCH 1999).

2.2. Site-Specific Plant Protection and Biomass Sensing

Differentiated herbicide application rates can be derived from previously mapped weed information (BROWN et al. 1990, WARTENBERG 1996, NORDMEYER et al. 1997, WILES & SCHWEIZER 1998, SÖKEFELD et al. 1999), or through on-board weed sensing by locating green plant material on bare ground (HAGGAR et al. 1983, FELTON & MC CLOY 1992, MERRITT et al. 1994), and a sensor/nozzle unit developed by PATCHENInc., California (HANKS 1995). More sophisticated sensors are under development (BILLER et al. 1997, VRINDTS & DE BAERDEMAEKER 1997, BROULIK et al. 1998, FEYAERTS et al. 1998, [page 5↓]HEISEL & CHRISTENSEN 1998, CHAPRON et al. 1999, MOSHOU et al. 1999), of which some are partly supported by computer models (ORIADE et al. 1996).

In site-specific differentiated application of growth-regulators, no on-line sensor so far is able to sense the crop parameters commonly used for applying site-specific differentiated amounts of these regulators, such as crop density. Only a few publications exist considering the application of growth-regulators according to soil maps and crop densities compiled by a German research group (SCHULZKE 1982, SCHÄDLICH et al. 1985, SCHÄDLICH et al. 1986, SCHULZKE & THIERE 1986, SCHULZKE et al. 1986). Just recently was the here presented pendulum-meter used by a research group to measure the lodging susceptibility of wheat (UDOH et al. 2000).

The potential of site-specific differentiated application of fungicides is still untouched due to the lack of sensors that are able to sense either the crop parameters, influencing growth of fungi, or sensing fungi on or in the plants directly. Only a double-step approach based on soil maps and then weather data, and later processed in decision support systems, is working up to date (SECHER et al. 1996a, AUDSLEY et al. 1997, SECHER et al. 1997, VOLK 1998, BJERRE 1999). Further practised is the traditional field sampling of fungi infestation or plant parameters (SECHER et al. 1995, SECHER et al. 1996b, KUTCHER et al. 1998). The exception is a study conducted to use remotely sensed digital imagery to evaluate Sclerotinia sclerotiorum infestations in soybeans (DUDKA et al. 1998), and a publication suggesting a spectral reflectance technique for disease diagnosis on cucumber leaves (SASAKI et al. 1998). Both methods are not working so far in cereals. All decision support systems for plant protection, such as PRO_PLANT® in Germany or the Danish PC-PLANT PROTECTION®, take crop density into account for their advice on the use of fungicides and growth-regulators, and this works usually on a field scale or even a regional scale. Therefore, a sensor for determining crop density is highly desirable, but so far such a sensor has not been reported.

As long as there is no way to sense crop density, other potentially useful plant parameters such as crop biomass have to be used. Determination of crop biomass by using radioactive radiation (KÜHN & SCHÄTZLER 1976, FREYTAG & JÄGER 1979, FREYTAG et al. 1983, UNGER & QUILITZSCH 1984) is not advisable. Remote sensing of crop properties is also possible, but mostly hindered by weather conditions and a narrow time schedule (ASRAR et al. 1985, LORD & DESJARDINS 1985, WILLIS et al. 1998, ZHANG et al. 1998). Additionally, estimation of crop biomass is possible by subjective visual assessment (CAMPBELL & ARNOLD 1973, PIGGOT & MORGAN 1985), through off-line point-to-point data measured by disk-meters or plate-meters (table 1), and capacitance-meters (CAMPBELL et al. 1962, BACK 1968, ANGELONE et al. 1980, VICKERY et al. 1980, MICHELL & LARGE 1983, THOMSON [page 6↓]1983, ROBERTS et al. 1984, STOCKDALE 1984, WEBBY & PENGELLY 1986, BIRRELL & THOMPSON 1987, CROSBIE et al. 1987, CURRIE et al. 1987, GREATHEAD et al. 1987, L’HUILLIER & THOMSON 1988, CARLIER et al. 1989, PIGGOT 1989, BRYAN et al. 1990, GONZALEZ et al. 1990, GABRIELS & VAN DEN BERG 1993, HIRATA et al. 1993, LIU & HIRATA 1995, MURPHY et al. 1995). Disk-meters and plate-meters measure biomass by the resting height of disks / plates on crops, and capacitance-meters use the electric conductivity.

Table 1: Publications on disk-meters and plate-meters.

Authors

Year

Crop / Predominant Plant

Object

Utilisation Type

POWELL

1974

grasses & herbage

DM

pasture

CASTLE

1976

grasses

DM

pasture

EARLE & MC GOWAN

1979

Lolium perenne

DM

pasture

BAKER et al.

1981

grasses & herbage

DM

meadow

MICHELL & LARGE

1983

Lolium perenne

DM

pasture

FARIAS & THOMAS

1984

Lolium multiflorum

DM & FM

meadow

SHARROW

1984

grasses & herbage, and Avena sativa

FM

pasture

STOCKDALE

1984

grasses & herbage

DM

pasture

GIBB & RIDOUT

1986

Lolium perenne & Trifolium repens

sward height

pasture

PALAZZO & LEE

1986

Festuca arundinacea & Lespedeza cuneata

DM

meadow

PIGGOT

1986

grass

DM

pasture

SCRIVNER et al.

1986

Lolium multiflorum & Trifolium subterraneum

DM

pasture

FULKERSON & MICHELL

1987

Lolium perenne & Trifolium repens

DM

pasture

ILLIUS et al.

1987

grass

sward height

pasture

KARL & NICHOLSON

1987

grass & herbage

DM

meadow

PETERSON & HUSSEY

1987

Cynodon dactylon

DM

pasture

GIBB & RIDOUT

1988

Lolium perenne & Trifolium repens

sward height

pasture

L'HUILLIER & THOMSON

1988

Lolium perenne & Trifolium repens

FM

pasture

MARTINEZ et al.

1988

Cynodon dactylon

DM

pasture

BRYAN et al.

1989

grasses & herbage

DM

pasture

CARLIER et al.

1989

grasses

DM

pasture & meadow

GIBB et al.

1989

Lolium perenne & Trifolium repens

sward height

pasture

GREEN et al.

1989

Festuca arundinacea & Lolium perenne

DM

pasture

KRUMIN'SH

1989

Dactylis glomerata

DM

meadow

LACA et al.

1989

grass & herbage

DM

pasture

MARTINEZ et al.

1989

grass

DM

pasture


[page 7↓]

Table 1 continued

Authors

Year

Crop / Predominant Plant

Object

Utilisation Type

MARTINEZ-EXPOSITO et al.

1989

Neonotonia wightii

DM

pasture

PIGGOT

1989

grass & herbage

DM

pasture

GONZALEZ et al.

1990

Cynodon dactylon

DM

pasture

MOULD

1990

grasses

DM

pasture & meadow

GOURLEY & MC GOWAN

1991

Lolium perenne & Trifolium repens

DM

pasture

HUTCHINGS

1991

grass

sward height

meadow

AIKEN & BRANSBY

1992

Festuca arundinacea

FM

pasture

CRAWFORD

1992

Lolium multiflorum

DM

meadow

MOULD

1992

grass

DM

pasture

GABRIELS & VAN DEN BERG

1993

Lolium perenne

DM

pasture & meadow

HATCH & TAINTON

1995

grass & herbage

DM

savannah

MURPHY et al.

1995

grasses & herbage

DM

pasture

VIRKAJÄRVI & MATILAINEN

1995

Phleum pratense

FM

pasture

BROCKETT

1996

grass

DM

biomass production

REEVES et al.

1996

Pennisetum clandestinum

DM & FM

pasture

TEAGUE et al.

1996

Bothriochloa ischaemum

sward height

pasture

HARMONEY et al.

1997

grasses & herbage

DM

pasture

RAYBURN&RAYBURN

1998

grasses & herbage

DM

pasture

MOSIMAN et al.

1999

grasses & herbage

DM

pasture & meadow

Additional methods of biomass sensing were hand-held devices for red or infrared reflectance (DRAKE 1976, BUERKERT et al. 1995) and hand-held radiometers (PEARSON et al. 1976a, AASE & SIDDOWAY 1981, WALLER et al. 1981, HARDISKY et al. 1984, ASRAR et al. 1985), a crane-based microwave backscattering technique (WIGNERON et al. 1995), and leaf area indices LAI measurements (DOBERMANN & PAMPOLINO 1995, HARMONEY et al. 1997). Sensing of crop biomass was reported by remote sensing through radar (STEINGIESSER 1997, DEGUISE et al. 1998), computer image analysis (THAMM 1986, EVERS et al. 1987, GROSS et al. 1987, THAMM 1989), and infrared reflectance (PEARSON et al. 1976b, TUCKER et al. 1981). Machine-based biomass sensing was conducted by using visible and near infrared reflectance (TUCKER 1979, DUSEK et al. 1985, LORENZEN & JENSEN 1988, RICHTER 1988, JENSEN et al. 1990) and mechanical scanning (EHLERT & SCHMIDT 1996, EHLERT & JÜRSCHIK 1997, EHLERT 1998). Several researchers used LAI or NDVI measurements as a base for decision making in plant protection (CHRISTENSEN et al. 1997, PAICE et al. 1999). Unfortunately, most methods are not suitable for on-line sensing, and taking samples is time-consuming for providing application maps, and is cost-intensive when used on a small field scale. Of the two most promising on-line methods, the mechanical scanning and the NDVI-scanning, the second is a much more expensive measurement system than the first.


© Die inhaltliche Zusammenstellung und Aufmachung dieser Publikation sowie die elektronische Verarbeitung sind urheberrechtlich geschützt. Jede Verwertung, die nicht ausdrücklich vom Urheberrechtsgesetz zugelassen ist, bedarf der vorherigen Zustimmung. Das gilt insbesondere für die Vervielfältigung, die Bearbeitung und Einspeicherung und Verarbeitung in elektronische Systeme.
DiML DTD Version 3.0Zertifizierter Dokumentenserver
der Humboldt-Universität zu Berlin
HTML generated:
13.09.2004