Mr Greg Beeston Dr Stephen Gherardi Dr David Cooperative

Document Sample
Mr Greg Beeston Dr Stephen Gherardi Dr David Cooperative Powered By Docstoc
					             A Near Real Time Crop and Pasture Package: Integrated Remote Sensing Technologies for
                                          Improved Farm Management

 Project Leader                             A/Prof. Graciela Metternicht: Dept. of Spatial Sciences, Curtin University of Technology, Email:
 Research Team                              Dr Andrew Rodger, Research Associate, Email:
                                            Dr Tom Schut, Research Associate, Email:
                                            Mr Greg Beeston, Dr Stephen Gherardi, Dr David Stephens, Mr Damien Shepherd: Department of Agriculture Western Australia (DAWA)
                                            Dr Richard Smith: Satellite Remote Sensing, Department of Land Information Western Australia (DLI)
 Project Participants                       Universities: Curtin University of Technology
                                            SMEs: Intergraph, SKM, Wesfarmers CSBP
                                            Public sector agencies: Dept. of Agriculture Western Australia, Dept. of Land Information Western Australia, CSIRO Livestock Industries

Objectives                                     In partnership with commercial companies, deliver a „farm package‟ for pasture and crops that can be used by producers to make informed tactical and
                                               strategic decisions at the paddock and farm level.
Outcomes                                       Image-maps, models and customised software for the commercialisation of a „near-real time‟ farm package of pastures and crops, containing information
                                               on growth rate, crop yield forecast, biomass and pasture quality.

                                                                                                   Specific Objectives:
Crop Yield/Biomass Estimation in Near Real Time (NRT)                                                                                                   Estimation of paddock-level pasture quality characteristics
from Remotely Sensed Data       Current atmospherically compensated
                                                                                                                                                        at monthly intervals in Mediterranean region of WA.
                           Reformat                                        MODIS data (surface reflectance) for the
MODIS Reflectance          & Cull DLI
                                                 Add data to the DAWA
                                                                                                                                                                                                                     The sheep industry in Western Australia has
DATA: DLI Supplied          MODIS
                                                  Database of MODIS        West Coast of Australia is processed and
                                                        imagery                                                                                                                                                      improved pasture utilization by adoption of
                             data                                          supplied by DLI. Historic MODIS data is
                                                                                                                                                                                                                     rotational grazing and monitoring pasture
                                                                           being added to the archive, and will be
                                                                           used to find and retrieve “pure” pixels, or
 At given location
  identify historic      Extract spectral data      Calculate fractional   spectral end-members, of soils and crop.
                                                                                                                                                                                                                     However, pasture quality characteristics
    maximum &           & define end-members            Coverage‟s         The end-members will allow a fractional
   minimum NDVI                                                                                                                                                                                                      (botanical composition, nitrogen, fibre,
                                                                           coverage of crop at the paddock-level to
                                                                                                                                                                                                                     pasture water content, carbohydrate, in vitro
                                 Historic or                               be estimated, and related back to a crop-
                                                   Compute Regression                                                                                                                                                dry matter digestibility, organic matter and
                                                       Equations           growth model.
 At the same location
                                 Yield data                                                                                                                                                                          lignin/cellulose) change strongly within the
    With NRT data                                                          The crop-growth model and fractional                                                                                                      growing season.
                          Calculate fractional       Calculate Target      coverage knowledge are used to produce
 Extract spectra
                               coverage                   Yield            a crop-biomass for a given paddock.
Figure 1: Generalised flow for estimation of crop yield                                                                                                 90                                                       25                                   5                                                          31.3
are a given target.                                                                                                                                                                            1992-1993
                                                                 Biomass (kg/ha) = crop-growth yield (kg/ha) *
                                                                                                                                                        80                                                                                                                                 2001-2002

                                                                                                                                                                                                                      Metabolic energy (MJ/day/DSE)
                                                                                                                                                                                               DSE Maintenance   20                                   4                                                          25.0
                                                                                                                                                                                                                                                                                           CP% DSE Maintenance
                                                                                                                         In-vitro digestibility (%) .

                                                                 Fractional Coverage

                                                                                                                                                                                                                                                                                                                        Crude protein (%)

                                                                                                                                                                                                                               N content (%)
                                                                                                                                                                                                                 15                                   3                                                          18.8


                                         Selected field sites will be used to test the methodology. Such                                                                                                         10                                   2                                                          12.5

                                         sites include the DAWA Merredin research station, Western                                                      50

                                         Australia. This station grows a variety of crops, which includes                                               40
                                                                                                                                                                                                                 5                                    1                                                          6.3

                                         wheat, lupins, and field peas. Preliminary field trips are being
                                         used to estimate the viability of the proposed fractional                                                      30                                                       0                                    0                                                          0.0

                                         coverage method. This method, like the use of a NDVI is only
                                                                                                                                                          May    Jul   Sep   Oct   Dec   Feb    Mar       May                                          May   Jul   Sep   Oct   Dec   Feb         Mar      May

                                         feasible while the crop is below 100% coverage. Beyond this
                                         point another method will be used. It is assumed that the                                             Assessment of the feeding value enables farmers to limit costly supplement feeding
                                         combination of the estimated biomass for a given paddock, as                                          by differentiating between groups of animals and adjustment of nutrient supply.
                                         a function of time, will allow a paddock specific biomass-                                            Remote sensing provides means to measure the chemical and botanical quality of
                                         model to continue the role of estimating the current biomass.                                         pastures.
Figure 2: Merredin Research Station (Western Australia):-31.49S 118.226E

                                                                                                                                                                                                                                                                                             Dry Matter
                                                                                                                                                                                                                                                                                              Yield Map

                                                                                                                                                                                                                                                                                            Protein Map
Figure 3: From field to end-member to fractional coverage                                                                                                                                                                                                    Processing
Figure 3 shows the concept of end-member and fractional coverage. End members from
selected field measurements are used to estimate the coverage of each crop within the
field spectrometers field-of-view (FOV). The results are encouraging at the large spatial                                                                                                                                                                                                   Energy Map
scale but are yet to be confirmed at the smaller MODIS spatial scale.
                                                      A technique allowing automatic selection of end-members
                                                      from MODIS imagery is required. The spectra at left show
                                                      a selection of “soil” end-members automatically extracted
                                                      from MODIS imagery of the Merredin station. Such end-                                                  Within this project, spectral, spatial and bi-directional information is used to
                                                      members, when combined with a vegetation end-member                                                    establish relationships between RS and pasture quality. Experiments with varying
                                                      will be used to estimate the fractional coverage of a given                                            botanical composition and nutrient (N, P, K, S) supply are used to study the
                                                      paddock, and hence biomass.                                                                            robustness of these relationships.

Figure 4: Soil end-members automatically extracted from temporal MODIS data at different paddock locations at
the Merredin research station.

                                        Cooperative Research Centre for Spatial Information