|
|
The Penn State Apple Orchard Consultant(PSAOC)
was developed on a Macintosh computer and employs a frame-based expert system tool,
Pennshell, which was written in the C programming language. Although parts of PSAOC were
built directly from the C language, Pennshell is designed as a 'tool box' of often-used
functions so that little direct coding is necessary. Each frame in PSAOC stores knowledge
about a particular object (the phenology of the orchard, for example). The status (or
value) of the frame (e.g. pink stage of phenology) helps to determine the final
recommendation. Frames can be one of two types, independent or dependent. As its name
implies, a dependent frame must make use of other frames or specifically built functions
to determine its status. An independent frame's state (i.e. its value) remains fixed and
is not dependent on other frames and its state is determined by querying the user.
Phenology, cultivar, disease status, last spray date and last pesticides sprayed are
examples of independent frames. The states of the independent frames and other functions
built in the C lanuage are used to determine the states of dependent frames, i.e. disease
potential, cultivar susceptibility and infection. Each frame contains five cells. An action is performed when any cell is
called directly or indirectly by the user of the system. The ABOUT cell is a description
of the object for which the frame was built. The EXPLANATION cell contains reasons, based
on the frame's status, for a certain recommendation provided by the expert system. The
RESPONSE cell performs an action based on the state of the frame. The HELP cell can be
used to help the user understand a particular question asked by the system. The GETVALUE
cell contains the information used to determine the state of the frame. Only the GETVALUE
cell of any particular frame has to have something built into it for the frame to become activiated.
|
|
|
|
|
The pest management portion of PSAOC is
composed of 3 parts; 1) the orchard profile which is composed of the variables that
describe the orchard, 2) the pest rating modules, which determine the pests and level of
severity and 3) the chemical management modules, which determine the chemicals that are
appropriate for the given circumstances, the rates of those chemicals and the spray
interval for the next pesticide application. The compatibility of the chemicals and
days-to-harvest limitations are also determined. |
|
|
|
|
Since the information needed to assemble a
meaningful expert system is derived from many areas, a team approach to knowledge base
development was taken. The team includes experts from, plant pathology, entomology,
horticulture, agricultural engineering, agricultural economics, and agricultural
meteorology. Expert systems are best conceived as a whole but then broken down into
smaller subunits for the actual development. For example, PSAOC covers the range of
problems encountered by a fruit grower but it was built as a series of modules (pest
management, leaf analysis, tree spacing etc.). Each module may be subdivided several more
times to arrive at the point of simplification is reached where the information is
manageable. For instance, the pest management module of PSAOC includes lower level modules
encompassing apple scab, powdery mildew and cedar apple rust potentials, insect
thresholds, chemical, chemical rate and spray intervals. There are modules below these
which describe infection periods, chemical residue levels, etc. These modules, which were
built separately, interact to derive an integrated disease and insect recommendation. The
relationship and number of modules in an area is determined by the experts who are
designing the system. To efficiently utilize information put into the system by the user,
the system stores the orchard description supplied by the user for use by all modules within the system. |
|
|
|
|
Upon operating the expert system, the user
views the start-up screen. From this point the user can gain access to any module within
the system. The pest management program can either be initiated directly from the profile,
in which case all profile information will automatically be loaded into the program, or
else the user will be asked if a profile needs to be loaded. Typically, an orchard
contains many "blocks" or "management units" which tend to be managed
in a slightly different manner. Each block would have its own profile. The user can either
choose a previously defined profile or create a new one. The user has the option of
looking at an individual pest problem or running the IPM module, which considers the
entire orchard block as a system where disease and insect recommendations are integrated.
For disease, the program first determines the disease potential in the block for apple
scab, powdery mildew, cedar apple rust and summer diseases. The system identifies the
fungicides that are available to control scab under the circumstances and asks the user to
identify preferences. Once the primary scab fungicide is selected by the user, the system
lists the combination fungicides which are recommended to prevent resistance build-up and
provide additional control of powdery mildew or rust if necessary. The fungicide
recommendation is given as part of the IPM recommendation after running the insect
modules. In the insect module the system determines if the insect and mite populations are
over thresholds that will require control. It then calls the chemical management module to
establish pesticide application priorities. If the mite population is over threshold and
predators are not sufficient to control the mites, miticide rates are determined.
Insecticides and rates are then determined for the primary insect over threshold (i.e.
most damaging). However, users are given the option of selecting a different primary
insect pest. If there are several insecticides that will control the primary insect pest,
the user is given the option of making the selection. If the primary insect control
material is effective for all secondary insects, no more insecticide compounds will be
considered. Otherwise, the module will determine other compounds and rates to control the
secondary insects. The recommendation follows for each chemical selected along with the
rates and application timing required (Figure 10). In addition, spray incompatibility
warnings are displayed. After the recommendation has been viewed the user is given the
option of making other chemical selections which result in a new IPM recommendation. At
this time, the user may also ask for an explanation of how the recommendation was derived.
The system then reviews each aspect of the decision making process. The user can also
request detailed information about any one of the chemicals or pests included in the
recommendation. The user also has the options of printing the recommendation, explanation,
or the profile information. |
|
|
|
|
At many places in the expert system the user
can obtain educational information that gives further explanation and background about a
disease, an insect or a chemical. This instruction allows the user to more accurately
answer a given question. For instance, in Figure 3 if the cursor is placed on the word
scab and "clicked" (mouse button depressed), the instructions for scouting for
apple scab and a description of the disease cycle of apple scab appears along with
illustrations of fruit and leaf symptoms. Instructions for scouting the diseases and
insects vary as the season progresses. |
|
|
|
|
The orchard profile describes the orchard. The
characteristics that describe the orchard are separated into long term characteristics
that do not change within a season and temporal characteristics that have the potential to
change within the season. The grower need only update the long term characteristics once a
year but temporal characteristics are prompted for update daily (Figure 3 & 4). Data
entry is accomplished in one of three ways; the user can highlight active zones on the
screen (Figure 3), the user can scroll to the correct response (Figure 4) or the user can
enter the data directly from the keyboard. For example, when entering the fungicide
material and rate the user applied in the last fungicide application (Figure 4), the
button can be clicked in front of the chemical name to indicate which chemical was
applied. The rate applied can be entered by scrolling to the correct amount or it can be directly entered in the box to the right
of the scroll bar from the keyboard. Information within the orchard profile can be
utilized by all modules of the system at any time. |
|
|
|
|
Different components (modules) of the expert
system are called by a menu interface. For the IPM module in PSAOC the menu items include
orchard profile, scouting, weather, diseases - apple scab (Venturia inaequalis),
powdery mildew (Podosphaera leucotricha), cedar apple rust (Gymnosporangia
juniperi-virginianae), summer diseases (Botryosphaeria obtusa, Botryosphaeria
dothidea, Glomerella cingulata, Gloeodes pomigena, Zygophiala
jamaicensis) and all diseases, insects (codling moth, European red mites, green apple
aphid, green fruit worm, gypsy moth, plum curculio, rosy apple aphids, spotted tentiform
leaf miner, tufted apple bud moth, white apple leafhopper, all insects), integrated pest
management (IPM) and horticultural modules (leaf analysis, tree spacing, irrigation
scheduling, and weed control). The user can choose one pest, all of the diseases, all of
the insects or receive an integrated insect and disease control recommendation by
selecting IPM. Once the menu selection is made, the program executes the GETVALUE cell of
the particular frame called by the menu selection. This frame is dependent (disease
potential, for example) and is assigned a value based on the status of other frames
(independent or dependent), whose values are obtained from profile information or user
interaction. In the RESPONSE cell another dependent frame (e.g. chemical selection) is
called to determine its state and response. More dependent frames or custom built
functions are called until all the necessary information has been obtained to offer a recommendation. |
|
|
The knowledge provided by PSAOC can be
substituted for the routine spraying practices that might have occurred without this
knowledge. Thus, the ecosystem is spared the application of unnecessary pesticides, while
the grower realizes an economic savings derived from not applying pesticides in certain
situations. Moreover, the yield and quality of the crop are maintained because pest
problems are managed with a profitability objective.
PSAOC is a potentially effective tool for apple production for at least seven reasons:
|
 |
 |
Delivery IPM derived information and solutions
to pest management problems |
|
Providing this information in a very
up-to-date and site specific fashion unattainable by traditional information delivery
systems |
|
This information is always readily available
to any grower having access to a computer and the software, relieving dependence upon the
accessibility of literature or human experts, thus enabling the grower to make critical,
timely decisions whenever necessary |
|
Used effectively, it provides the apple grower
with the opportunity to reduce or in many cases optimize the use of chemicals thus
reducing the negative impacts of apple production on the ecosystem and human health |
|
The PSAOC is an educational tool. As the
grower continues to use the PSAOC expert system, he receives instruction within the expert
on IPM procedures but he also observes and comes to understand how IPM strategies are
implemented by the expert system |
|
Increase grower profits |
|
As additional methods of production management
are developed, these can be easily incorporated into the PSAOC |
However, it remains to be seen whether apple
producers will successfully adopt this new agricultural innovation on a widespread basis.
The adoption of computer technology by growers is predicated on a linkage between a
particular farm operation and the access conditions of the particular technology (3).
These access conditions are determined, in part, by the development of the technology and
by private and public diffusion infrastructures. The development of diffusion strategies
that consider grower needs and capabilities relative to specific access conditions will
accelerate the adoption of these new technologies.
Because of its interactive nature and potential impact on
farm decision making, PSAOC was designed with the aid of several commercial orchardists
and its impact on farm decisions was assessed during a two year field test (11). Positive
economic and behavioral changes were noted that promoted improved profitability, increased
orchard monitoring and more efficient pesticide use. Several recommendations were
constructed from this study that would enhance the adoption of expert systems technology.
They are:
|
Training and basic orientation to computer use
for farming operations in general, and agricultural expert systems in particular. These
trainings should be held on a very localized basis and taught by persons familiar with
expert systems software and the cropping system being discussed |
|
Training which links expert system technology
with an overview of the gradual modification of existing production systems to incorporate
modern efficient methods. This training should focus on the societal level needs and
responsibilities for reducing pesticide use as well as the long-term farm level benefits
for doing so |
|
Establishment of a "local experts"
network to provide a resource for growers experiencing difficulties with the computer or
expert system |
|
Continual updating of system capabilities, so
that recommendations remain scientifically current and appropriate |
|
Training of extension specialists and agents
to familiarize them with the possibilities and potentials of the system |
|
Begin the process by delineating the criteria
and goals for modern crop production attainable with expert systems as a tool. In this way
scientists will be better able to begin to design production systems for agricultural
operations of all sizes that provide more flexibility in responding to dynamic production
conditions, thus enabling time and spatially specific recommendations of the expert system
to be better implemented. In the long run this may be the greatest contribution of
agricultural expert systems development toward a more efficient system of agriculture |
|
