COMPUTER SYSTEMS AND METHODS FOR SELECTING SEED VARIETIES
The present invention relates to computer systems and methods for assisting users with selecting seed varieties and developing customized planting strategies for particular crop growing operations and, more specifically, to such systems and methods where outcome predictions and recommendations are made for particular seed varieties based upon numerous parameters input by a user which relate to a growing operation of interest. Background of the Invention
Due to the large number of commercially available seed varieties, and the numerous variables which bear upon the success of any given seed variety in any given crop growing operation, it is becoming increasingly difficult for growers to select the best seed varieties for their particular operations. As is well known, the outcome of using any particular seed variety depends upon a large number of factors such as, for example, the geographic region where the seed will be grown, the soil type, the planting date, and the irrigation levels or rainfall, the prior crop grown, pesticide treatments, etc. Thus, it is almost impossible for a grower to intuitively and effectively select the best seed variety for any given set of conditions.
To choose their seed varieties, growers frequently use basic fact sheets which rank various seed varieties (e.g., on a scale of 1 to 10) according to certain characteristics such as expected yields, disease resistance, etc. However, such rankings, by their very nature, provide only limited information. Although one seed variety may be ranked higher than others with respect to, for example, disease resistance, the difference in the rankings is seldom if ever quantified. Moreover, there are likely many other factors related to a particular growing operation which will determine or at least affect the success of using a particular seed variety, but that are
not included in the basic fact sheet. Further, because such fact sheets are normally provided by seed producers, and only list seed varieties from a given source (or commonly owned sources), the unbiased nature of these fact sheets is suspect, and the ability to make a meaningful comparison of a fact sheet from one producer with a fact sheet from another is limited.
Perhaps most importantly, there is currently no means available for growers to effectively evaluate what is usually their primary objective — maximizing profits — when selecting seed varieties and developing planting strategies. For example, it may be that, for a particular growing operation, the greatest profits could be realized by using a particular seed variety in combination with a pesticide treatment, an early planting date, tillage system, and several other factors. But the grower presently has no way to identify this optimal planting strategy from the myriad possibilities, which would also require consideration of numerous cost parameters, including seed costs, operational costs, etc. Thus, while the number of variables and choices faced by the grower is large, and the interactions between those variables and choices are complex, the information to which the grower has access regarding the consequences of those choices is limited. Consequently, it is extremely difficult for many growers to identify optimal planting strategies. Summary of the Invention In order to solve these and other needs in the art, the inventors hereof have succeeded at designing and developing computer systems and methods which enable growers to identify optimal planting strategies for their growing operations. In addition to allowing growers to consider planting strategies that will maximize crop yield, a trait of particular interest to most growers, the present invention also provides growers with a tool for identifying and considering those strategies that will maximize profits, or any other agronomic variable of interest. The present invention also enables growers to specifically investigate the sensitivity of their profits to changes in their planting strategies, and to their tolerances for risk.
Preferably, a computer system according to the present invention includes an interface through which a user can input several parameters related to a growing operation, as well as certain preferences of the user and the user's tolerance for risk. Using these parameters, the system can reduce what may be a large list of commercially available seed varieties into a smaller, more manageable list of seed
varieties that are compatible with the parameters input by the user. From this list, the user can preferably select a seed variety to obtain predictions of the outcomes to result from using that seed variety in the growing operation of interest. These predictions may include predictions of certain crop traits levels, profits, etc. which are preferably displayed in the form of probability distribution functions, thereby allowing the user to consider the likelihood of realizing a particular level of yield, profit, etc. Alternatively, the preferred system can perform an optimization routine in which the system evaluates each seed variety that is compatible with the growing operation of interest to identify those varieties that are projected to produce the highest profits, yields, or other results of interest to the user. The selection of optimal seed varieties can also be based upon the user's risk tolerance. The preferred system also allows the user to develop planting strategies at the subfield or plot level, and then to aggregate projected profits, yields, etc. for a particular field having multiple plots, a particular farm having multiple fields, or a particular firm having multiple farms, such that the maximal average profitability and yield, and the maximal minimum profitability, can be determined not only at the plot level, but also at the field, farm and firm levels. This computer system is preferably implemented over a publicly distributed computer network, such as the Internet, to provide growers with a convenient, interactive decision support tool which aids them in making complex planting decisions. In accordance with one aspect of the present invention, a computer- implemented method for assisting a user with selecting at least one seed variety from a plurality of seed varieties for use in a growing operation comprises the steps of providing an interface through which the user can input at least one and preferably several parameters related to the growing operation, and determining which of said plurality of seed varieties are compatible with the growing operation using the parameter(s) input by the user. One parameter input by the user is preferably an anticipated crop end use and/or a desired crop trait (or level thereof). Preferably, the user is permitted to access a list of the plurality of seed varieties, and to select from that list a particular seed variety of interest to the user. Upon inputting a parameter incompatible with the selected seed variety, the user's selection of the particular seed variety is preferably discarded.
In accordance with another aspect of the present invention, a computer system for assisting a user with selecting at least one seed variety from a plurality of seed
varieties for use in a growing operation includes an interface through which the user can input at least one and preferably several parameters related to the growing operation. The computer system is preferably configured to determine which of the plurality of seed varieties are compatible with the growing operation using the parameter input by the user, which may be anticipated crop end use, a desired crop trait, etc.
In accordance with yet another aspect of the present invention, a computer- implemented method for agricultural applications comprises the step of predicting at least one outcome to result from using a particular seed variety in a growing operation using at least a quantitative model, and preferably using one or more parameters related to the growing operation which are input by the user. These parameters may be field parameters, operational parameters, cost parameters, desired crop characteristics, etc. In one preferred embodiment, the quantitative model is a statistical model, the outcome of interest is modeled as a normal random variable, and a multiplicity of realizations are generated from a probability distribution for the random variable. Preferably, the predicted outcome, which may be an economic or crop trait outcome, is displayed in the form of a probability distribution function or a probability interval. In the latter case, the length of the probability interval is preferably defined by a probability value input by the user. In accordance with yet another aspect of the present invention, a computer system for agricultural applications is configured to predict at least one outcome to result from using a particular seed variety in a growing operation using at least a quantitative model, and preferably using one or more parameters related to the growing operation which are input by the user. The predicted outcome is preferably an economic outcome such as crop selling price, crop revenue or crop profit, or a crop trait such as yield, Hagberg falling number, specific weight, or protein content.
In accordance with still another aspect of the present invention, a computer- implemented method for assisting a user with selecting at least one seed variety from a plurality of seed varieties for use in a growing operation comprises the step of identifying at least one and preferably several of the plurality of seed varieties as being optimal seed varieties based at least in part upon an objective function. Preferably, the method also comprises the step of providing an interface through which the user can input at least one and preferably several parameters related to the
growing operation, and the optimal seed varieties are identified using the several input parameters. In one preferred embodiment, the objective function is to maximize a variable such as average profit, minimum profit, or average yield. Preferably, the method includes the step of predicting, for each of the plurality of seed varieties, the profit to result from using that seed variety in the growing operation, and the optimal seed varieties are identified and ranked using these profit predictions. The optimal varieties may also be determined as a function of the user's tolerance for risk.
In accordance with still another aspect of the present invention, a computer system for assisting a user with selecting at least one seed variety from a plurality of seed varieties for use in a growing operation is configured to identify at least one of the plurality of seed varieties as being an optimal seed variety based at least in part upon an objective function. This system preferably includes an interface through which the user can input at least one and preferably several parameters related to the growing operation, and the system is configured to identify at least one of the plurality of seed varieties as being an optimal seed variety for the growing operation based at least in part upon the objective function and the parameter(s) input by the user.
While some of the principal advantages and features of the present invention have been described above, a greater and more thorough understanding of the invention may be attained by referring to the drawings and the detailed description of the preferred embodiments set forth below.
Brief Description of the Drawings
Figure 1 is a flow chart of a computer system constructed according to one preferred embodiment of the present invention; Figure 2 is a screen display for inputting field characteristic parameters into a computer system constructed according to a second preferred embodiment of the present invention;
Figure 3 is a screen display illustrating several seed varieties that are compatible with certain preferences input by a user; Figure 4 is a screen display illustrating a reduced set of compatible seed varieties which resulted from a change in the user preferences shown in Fig. 3; Figure 5 is a screen display for inputting cost parameters for a particular growing operation;
Figure 6 is a screen display which includes outcome predictions for a particular seed variety selected by the user;
Figure 7 illustrates a data array used by the computer system to produce the outcome predictions shown in Fig. 6; Figure 8 is a screen display which includes profit predictions for several optimal seed varieties;
Figure 9 is a screen display for inputting farm data into a computer system constructed according to a third preferred embodiment of the present invention;
Figure 10 is a screen display for inputting field data; Figure 11 is a screen display for inputting plot data;
Figures 12-14 illustrate a screen display for inputting scenario data;
Figures 15 and 16 illustrate a screen display which includes average yield, average profit, and minimum profit predictions for several optimal seed varieties;
Figure 17 illustrates the display of several predicted values upon request by the user;
Figure 18 is a screen display summarizing the results of two scenarios for a particular plot;
Figure 19 is a screen display summarizing the predicted average profit for multiple plots; Figure 20 is a screen display summarizing predicted average profits for multiple fields; and
Figure 21 is a screen display summarizing predicted average profits for multiple farms.
Detailed Description of the Preferred Embodiments A flowchart for a computer system constructed according to one preferred embodiment of the present invention is shown in Figure 1. Beginning at block 100, a user of the system inputs, via a user interface, at least one and preferably numerous parameters related to a particular growing operation for which at least one seed variety is to be selected for planting. These parameters may include, for example, characteristics of the field in which the seed will be planted, various costs related to the growing operation, whether and when specific processes will take place over the growing season, the anticipated end use of the crop to be grown, desired crop traits, the user's tolerance for risk, etc. Using one or more of these parameters, the system
then generates, at block 102, a list of seed varieties that are compatible with the growing operation of interest. Thus, block 102 can be characterized as an expert system that reduces what may be a relatively large list of commercially available seed varieties into a smaller, more manageable list of seed varieties that are compatible with the growing operation of interest to the user, as defined by the input parameters. Preferably, the list of commercially available seed varieties includes brands of "seed varieties" (which, as used herein, refers to hybrids, pure strains, chemically treated seeds, genetically engineered and enhanced seeds, etc.) from multiple seed companies (and preferably unrelated seed companies), and the computer system is configured to produce the list of compatible seed varieties in an unbiased fashion using objective data (e.g., field or laboratory data) and/or subjective data (e.g., consensus or interpretive data). Preferably, these data are normalized and/or standardized to take into account any differences in data collection methodologies, parameter definitions of multiple companies (e.g., differences in relative maturity group definitions, etc.),and so on. In this manner, the preferred system can provide growers (or any other users of the system) with an unbiased tool for identifying, from amongst several seeds offered by several seed producers, the particular seed varieties that are compatible with their growing operations.
At block 104, the system determines whether the user has selected a "run" option or an "optimize" option. If the run option is selected, indicating the user's desire to obtain a prediction of one or more outcomes to result from using a particular seed variety in the growing operation of interest, processing branches to block 110 where the system generates such predictions. Although the preferred system predicts economic outcomes (e.g., crop selling price, crop revenue, crop profit, etc.) and crop trait outcomes (e.g., yield, protein content, etc.), it could be configured to predict virtually any type of outcome of interest to the user. Preferably, the predicted outcomes are generated using a quantitative model (such as a statistical model, or a mechanistic model based on crop growth processes) and one or more parameters relating to the growing operation of interest, which are input by the user. In this preferred embodiment, the predictions are generated using statistical quantitative models, where certain outcomes are modeled as random variables, and numerous Monte Carlo realizations are generated from the probability distributions for these random variables using historical crop data (e.g., historical mean and standard
deviation data) for the selected seed variety as well as one or more input parameters related to the growing operation of interest. The predicted outcomes are displayed to the user at block 112 before processing continues to block 114, where the user can save the predictions for future use and reference. Processing then reverts back to block 100, where the user can change some or all of the previously input parameters to determine the effect of these changes, if any, on the prior prediction(s), and/or select the optimize option for the same or a different set of input parameters.
If the optimize option is selected by the user at block 104, processing branches to block 116 where one or more outcomes are predicted for each of the compatible seed varieties identified in block 102 using historical crop (or other) data, and preferably using one or more of the parameters input by the user. At block 118, the system uses the predicted outcomes generated in block 116 to select at least one and preferably several optimal seed varieties according to an objective function that is either predefined in the system programming or selected by the user. In this particular embodiment, the objective function, which is predefined, is to maximize profit (in contrast to, for example, maximizing yield), and the system is configured to select the seed varieties that are predicted to produce the greatest profits. Preferably, this selection of the optimal seed varieties is also based upon the user's tolerance for risk, as further explained below. At block 120, the predicted outcomes for the optimal seed varieties are displayed to the user before processing continues at block 114 in the manner described above.
Preferably, the computer system is specifically developed and configured for implementation over a publicly distributed computer network (in this case, the Internet). It could instead be implemented in a private network, a stand-alone computer, or any other type of computer arrangement, as apparent to those skilled in the art.
As should already be apparent, this preferred computer system includes three especially useful subsystems, namely, an expert system by which a list of available seed varieties can be reduced to a more manageable list of seed varieties compatible with the growing operation of interest to the user, a system for predicting one or more outcomes to result from using a particular seed variety in the growing operation of interest, and a system for automatically selecting optimal seed varieties for such growing operation. Although each of these subsystems (and/or the methods
performed thereby) could be implemented independent from one another, they are preferably used in combination as described herein.
A screen display which forms part of a user interface for a computer system constructed according to a second preferred embodiment of the present invention is shown in Fig. 2. In this embodiment, the computer system is configured to assist a user with selecting a particular variety of wheat seed for use in a growing operation located within the United Kingdom.
The screen display shown in Figure 2 is preferably the initial screen displayed to the user after the user has completed a registration process and indicated a desire to run the wheat seed selector system. The user may subsequently access the screen or folder shown in Figure 2 by selecting a folder tab 130, which is labeled "field characteristics." In this folder, the user can input a variety of parameters related to a particular growing operation (which may be a particular farm, field or plot, multiple farms, or even multiple firms). For example, using a drop-list 132, the user can select a particular seed variety of interest from a list of commercially available varieties.
Preferably, the user can initially access a complete list of all seed varieties included in the system by selecting the down-arrow for the drop-list 132, and then moving a scroll bar, if necessary, to view the entire list. If no selection of a seed variety is made by the user from the drop-list 132, the first listed seed variety (i.e., the seed variety at the top of the list, which is preferably arranged alphabetically) is the default selection. In text box 134, the user can enter the size of the field for which a seed variety is to be selected for planting. Note that, in text box 134, the user could input, for example, the size of a plot or subfield included within a larger field if the user is interested in choosing a seed variety for that particular plot. The user could also enter the combined size of several fields, especially if these fields have similar characteristics, to choose a particular seed variety for use in the several fields. Alternatively, the system could be configured to allow the user to specify whether the input data is for a particular plot in a field, a particular field which may comprise several plots, a particular farm that may comprise several fields, or even several, perhaps commonly owned farms, as further explained below.
As shown in Figure 2, the field characteristics folder includes a section 136 labeled "factors" in which the user can specify several field and operational parameters for the growing operation of interest. For example, using a drop-list 138,
the user can select the geographic region within the United Kingdom of the field for which the size was entered in box 134. In this particular embodiment, the region can be designated as Northeast, Northwest, Central, Southeast or Southwest. Using drop- lists 140 and 142, the user can indicate whether the seed variety to be selected will be used in either the first or second planting of a growing season, and whether the seed variety will be sown early or late in the season, respectively. Via drop-list 144, the user can identify the soil type of the field in which the selected seed variety will be used as being silty clay, sandy clay, or sandy. Whether the field will be treated with a fungicide is also input via drop-list 146. The field characteristics folder also includes a section 148 labeled
"preferences" in which the user can specify, among other things, the anticipated end use of the crop to be grown from the selected seed variety via a drop-list 150. In this particular embodiment, which is focused on selecting wheat seeds, the anticipated end use for the crop can be designated as either feed, biscuit (i.e., cookies) or bread. Also provided within the preferences section 148 are a number of dials 152, 154, 156, 158, 160, 162 whereby the user can input a minimum desired level of certain wheat crop traits (including standing power, straw length, maturation date, resistance to shedding and resistance to sprouting) as well as the user's tolerance for risk. Preferably, the user specifies the minimum desired crop trait levels by moving a slider bar indicator to an appropriate location along the dial. For example, if a slider bar 164 for standing power is placed at the far left end of the dial 152, as shown in Fig. 2, this indicates that the user is not concerned with the level of standing power that may exist in the wheat crop to be grown. Conversely, by placing the slider bar 164 at the far right end of the dial 152, the user can indicate that a high level of standing power is desired. Placing the slider bar 164 between the left and right ends of the dial 152 would indicate the user's desire to grow a wheat crop having an intermediate level of standing power, where the midpoint of the dial 152 represents an average level. With respect to risk tolerance, the user can move a slider bar 166 to the far left of the dial 162, as shown in Fig. 2, to indicate that the user is completely risk averse, or to the far right end of the dial to indicate that the user is completely risk tolerant, or to a position between the left and right ends of the dial 162 to indicate some intermediate level of risk tolerance. In this particular embodiment of the invention, the level of
risk tolerance input by the user is processed by the system when selecting one or more optimal seed varieties, as explained below.
Although the system user can initially view a list of all seed varieties that are processed by the system, as noted above, the system preferably includes an expert system corresponding to block 102 in Fig. 1 which removes seed varieties from the drop-list 132 shown in Fig. 2 upon determining that these seed varieties are incompatible with one or more parameters input by the user. In this particular embodiment, the input parameters that are processed by the expert system are restricted to the anticipated crop end use that is selected using the drop-list 150, as well as the minimum crop trait levels that are entered using the dials 152-160, although other parameters, including, for example, those input in the factors section 136, could likewise be processed by the expert system, if desired, assuming sufficient data is available for each seed variety/expert system parameter combination.
As shown in Fig. 3, a number of seed varieties are included in the drop-list 132. Due to how the system is configured, each of these varieties is compatible with the crop end use and the desired levels of straw length and maturation date input by the user in the preferences section 148. Otherwise, the listed varieties would have been removed by the expert system from the drop-list 132. Note that because the slider bars for standing power, resistance to shedding, and resistance to sprouting are all set at their lowest possible levels in Fig. 3, these values are not used by the expert system to eliminate certain seed varieties, since every seed variety is expected to produce some level of these traits, and will necessarily meet or exceed the minimum desired levels specified by the user.
An example of how incompatible seed varieties are removed from the drop-list 132 can be seen by comparing Figs. 3 and 4. The notable difference between these figures is that the anticipated crop end use selected from the drop-list 150 is changed from "feed" in Fig. 3 to "bread" in Fig. 4. When this occurs, the expert system removes from the drop-list 132 those seed varieties which are not suited for growing wheat that will be used in bread. In this particular embodiment, the expert system determined that only four seed varieties (i.e., Hereward, Hussar, Rialto and Spark) are compatible with the crop end use and the desired levels of straw length and maturation date input by the user in the preferences section 148 of Fig. 4. As shown in Fig. 3, the seed variety "Reaper" was already selected by the user before the crop
end use was changed to "bread" in Fig. 4. In the course of removing the "Reaper" seed variety from the drop-list 132 for being incompatible with the selected end use of "bread," the expert system also discarded the user's selection of the "Reaper" seed variety as the seed variety of interest, and chose the seed variety at the top of the new list ("Hereward," in this example) as the default selection.
Because this particular embodiment of the invention focuses on the selection of wheat seed varieties for growing within the United Kingdom, the expert system (and other subsystems in this embodiment) is preferably implemented using historical crop data from one or more handbooks published by the National Institute of Agricultural Botany ("NIAB"), Cambridge, U.K. As known in the art, these handbooks provide data and rankings on what are deemed the best wheat seed varieties for use within the United Kingdom. The expert system preferably uses these data and rankings to determine whether particular seed varieties are compatible with the anticipated end use and minimum desired trait levels input by the user in the preferences section 148. For example, if the user moves the slider bar 164 on the standing power dial 152 to the far right, then the expert system will require each seed variety to have the highest ranking (e.g., 9 on a scale of 1 to 9) for standing power in the NIAB handbook in order for that seed variety to be deemed compatible with the minimum desired level of standing power specified by the user. With respect to the crop end use, the data provided by NIAB is used to classify each seed variety as either "acceptable" or "not acceptable" for each anticipated end use.
Preferably, the computer system of this embodiment is configured to predict the profit a grower could expect to realize by using a particular seed variety in the growing operation of interest, and, if desired by the user, to automatically identify those seed varieties that will produce maximum profits for such growing operation. To accomplish these functions, the system must subtract operating costs from expected revenues, as explained in greater detail below. Thus, in addition to inputting various data in the field characteristics folder as described above, the user can also enter cost parameters related to the growing operation of interest by selecting the cost parameters tab 170, which calls up the cost parameters folder shown in Figure 5. In a land and equipment costs section 172, the user can input the cost of the land (per unit area) for the growing operation of interest using text box 174, and can also input the associated equipment cost and fuel/repairs cost (per unit area) using text boxes 176
and 178, respectively. In a contract work section 180, a plurality of check boxes 182, 184, 186, 188 and 190 are provided for the user to indicate whether any contract labor will be used for the operations of spraying, drilling, combining, carting and drying, respectively. In a labor section 192, the user can input in text box 194 the hourly wage rate paid for contract workers in the growing operation of interest. Labor section 192 also includes a contract labor dial 196 via which the user can specify, by moving a slider bar 198, the level of contract labor that will be used to perform the operations corresponding to each box 182-190 that is checked by the user in the contract work section 180. For example, if the slider bar 198 is moved to the far right end of the dial 196, this indicates that the operations for which boxes 182-190 are checked will be performed by contract labor exclusively, whereas moving the slider bar 198 to the center of the dial 196 would indicate that only 50% of such operations would be performed by contract labor. Note that, in this particular embodiment, the number of hours required to perform the operations listed in the contract work section 180 is preprogrammed into the system (or accessed from a data file), and does not have to be specified by the user.
After completing the field characteristics and cost parameters folders, the user can select either "run" or "optimize" from the menu bar 200. In this particular embodiment, selecting "run" will cause the system to display a user selected results folder, shown in Fig. 6, where predictions are provided of the profit and the crop traits which will result from using, in the growing operation of interest, the seed variety selected from the drop-list 132 in the field characteristics folder. In this particular embodiment of the invention, which pertains to wheat seeds, the system is preferably configured to predict, in addition to profit, several crop traits believed to be of greatest interest to wheat growers, namely, yield, Hagberg falling number, specific weight, and protein content. However, the system could also be configured to predict different or additional economic and crop trait outcomes, if desired. As shown in Figure 6, the predictions are preferably displayed to the user in the form of probability distribution functions and, specifically, in a form that is equal to one minus the cumulative distribution function. In this manner, the user can readily ascertain, for example, the probability of obtaining a certain level of profit if the selected seed variety is used in the growing operation of interest. For example, the profit distribution provided in Figure 6 reveals that if a particular seed variety (in this case,
Abbot) is used in the growing operation of interest, there is a 50% chance that a profit of 81,339 British pounds or higher will be realized. The means and standard deviations for the profit and crop trait distributions are also preferably displayed immediately adjacent to the distribution functions, as shown in Fig. 6. As apparent to those skilled in the art, the predictions could be displayed to the user in a wide variety of other formats, such as in the form of cumulative distribution functions, bar charts, box plots, etc. without departing from the scope of the present invention.
The manner in which the predicted outcomes shown in Fig. 6 are produced by the system in this particular embodiment will now be described. Preferably, the expected profit is modeled as the expected revenue minus the expected cost. The expected cost is computed directly from the parameters input by the user in the cost parameters folder shown in Fig. 5 (and the field size entered in box 134 of Fig. 2), as well as from the predefined number of hours required to perform the operations listed in the contract work section 180, which is taken directly from The Farm Management Pocketbook: 1998 produced by John Nix. The expected cost is also computed using the current price of the selected seed variety. Preferably, this price (and the price of the other seed varieties included in the drop-list 132) is preprogrammed into the system (or accessed from a data file), rather than being input by the user. In this manner, the seed prices are "transparent" to the user, and the user will be more likely to make seed selections based on overall predictions of, for example, profitability, rather than by comparing prices of several seed varieties that may produce markedly different results in any given growing operation.
The expected revenue is modeled as the expected yield (per unit area) times the field size times the expected selling price. In the case of wheat, the expected selling price depends upon the anticipated crop end use (e.g., feed, biscuit or bread) which, in turn, depends upon the expected values of Hagberg falling number, protein content, and specific weight. In other words, a premium is paid for wheat having certain minimum trait levels, since such wheat can be used in a greater number of applications. Thus, in this embodiment, the expected revenue is a function of the expected yield and the expected values of Hagberg falling number, protein content and specific weight.
As apparent to those skilled in the art, the expert system described above allows the user to exert some control over the selection of a seed variety using
variables familiar to the user that are not explicitly incorporated into the quantitative models. By including such variables in the quantitative models, the expert system may be eliminated without affecting system performance.
Preferably, the crop traits of yield, Hagberg falling number, protein content and specific weight are each modeled as normal random variables. For all four of these traits, standard deviations were deduced for the selected seed variety using the 95% confidence intervals plotted in the NIAB handbook. For Hagberg falling number, protein content and specific weight, the means for the selected seed variety were taken directly from NIAB. Thus, in this particular embodiment of the invention, these crop trait outcomes are predicted without regard to the parameters input by the user using statistical quantitative models and historical crop data only. Alternatively, these outcomes could be modeled as a function of one or more parameters input by the user, much like the model for mean yield discussed immediately below, assuming the availability of appropriate data. The mean yield is modeled as a function of five factors for which NIAB provides mean yield data. These factors are region, rotation, soil type, sowing date, and fungicide treatment. The mean yields provided by NIAB for each region are used to compute a scaling factor by which a base yield (i.e., the overall mean yield provided by NIAB for the selected seed variety) is adjusted to account for specific regional effects. Scaling factors are similarly computed for the rotation, soil type, sowing date and fungicide treatment factors and are similarly used to adjust the base yield value. More specifically, the mean yield is computed using the following multiplicative model. Let seeds be indexed by i and let model variables be indexed by j. For the i'th seed, let β be the effect of the j'th variable on yield. Let Xj be the input level of the j'th variable. Let the model contain J total variables and I different seeds. Then yield for the i'th seed is given by:
Y, = Π^ PB XJ i = w
For this particular embodiment, J = 15, where the 15 variables are as follow: (1) Base yield; (2) Northeast (region); (3) Northwest (region); (4) Central (region); (5) Southeast (region); (6) Southwest (region); (7) First planting (rotation); (8) Second
planting (rotation); (9) Silty clay (soil); (10) Sandy clay (soil); (11) Sandy (soil); (12) Early sown (date); (13) Late sown (date); (14) Treated (fungicide); and (15) Untreated (fungicide). In this model, all the X s are dummy variables. That is, X, takes on the value 0 or 1 depending upon whether characteristic j is absent or present, respectively. Xj always equals one.
Although in the particular embodiment under discussion, the crop traits of yield, Hagberg falling number, protein content and specific weight are modeled as described above, it should be understood that many other types of statistical and non- statistical models for these and other crop traits could be used without departing from the scope of the present invention.
To construct the distributions for yield, Hagberg falling number, protein content and specific weight shown in Figure 6 in this second preferred embodiment, which uses a simulation-based analysis of statistical models, preferably 20,000 realizations are generated from the probability distributions for these traits. The results of these realizations are then stored in a data array of the type shown in Fig. 7, and are also plotted for the user in the form of probability distribution functions, as shown in Fig. 6, adjacent to which the means and variances for each distribution function are displayed.
The selling price is modeled as a base price plus a premium. In this embodiment, the base price is the expected selling price for wheat to be used as feed (the "feed price"), and the premium is the price increase that can be expected for wheat suitable for the end use of bread (the "bread price premium"). The feed price is modeled as a normal random variable with its mean and standard deviation estimated from the Ministry of Agricultural Fisheries and Food ("MAFF") data. The bread price premium is modeled as a gamma random variable with four degrees of freedom, with its mean also estimated from the MAFF data. For feed price and bread price premium, 20,000 realizations of these random variables were generated from their respective probability distributions and stored in the data array shown in Fig. 7. For each quantitative realization of the random variables Hagberg falling number, specific weight and protein content, a selling price for wheat having those characteristics is generated (and stored in the data array). This price is expressed as a feed price plus a fraction of a maximum bread price premium. The fraction of the bread price premium is based upon a linear interpolation of the values of Hagberg falling number, specific
weight and protein content between the minimum and maximum obtainable values, which are preferably obtained from NIAB data. If the level of any one of these three traits falls below the minimum value, a zero bread price premium is earned.
Total revenue is then computed for each selling price and yield realization (i.e., by multiplying, for each realization, the yield times the field size times the selling price). 20,000 realizations of profit are then computed by subtracting expected cost from each realization of revenue. A mean and standard deviation of profit are then computed using the 20,000 profit realizations. These profit realizations are then plotted for the user in the form of a probability distribution function, as shown in Figure 6, adjacent to which the computed mean and standard deviation for the predicted profit are displayed.
When the user is finished reviewing the profit and crop trait predictions shown in Figure 6, the user can change one or more of the previously input parameters (by selecting the field characteristics tab or the cost parameters tab, as appropriate) to determine the effect of these changes, if any, on the profit and crop trait predictions. In the field characteristics folder, the user can also select a different seed variety using the same or different input parameters to obtain profit and crop trait predictions for that variety.
Instead of selecting "run" from the menu bar 200 to obtain profit and crop trait predictions for a particular seed variety, the user could instead select "optimize." When this occurs, the computer system of this particular embodiment performs an optimization routine to select the best seed varieties based upon the mean and variance of profitability evaluated via a constant relative risk aversion utility function in accordance with the degree of risk tolerance input by the user. In other words, the system produces a profit distribution for each compatible seed variety in the same fashion that a profit prediction was generated for the selected seed variety when the user selected the "run" option. The system then evaluates the mean and standard deviation of the predicted profit distribution for each compatible seed variety together with the risk tolerance value input by the user according to the constant relative risk aversion utility function, and produces a utility value for each compatible seed variety. The system then ranks the compatible seed varieties according to their utility values, and identifies the top five varieties (i.e., the five varieties having the maximal utility values) as being the optimal varieties. For these optimal varieties, the predicted
profit distributions are plotted, and the corresponding means and standard deviations are displayed, as shown in Fig. 8.
As apparent to those skilled in the art, if the user sets the risk tolerance slider bar 166 at the maximum level (i.e., at the far right end of the dial 162 shown in Fig. 3), thus indicating that the user is completely risk tolerant, the profit means — rather than the variances of profit — will dominate the risk aversion utility function, and the system will identify as the optimal seed varieties the five varieties having the highest profit means. In this manner, the user can control the system, if desired, so as to effectively identify the optimal seed varieties without taking into account the user's tolerance for risk.
Upon reviewing and perhaps saving the profit predictions shown in Fig. 8 for the five optimal seed varieties, the user can change one or more of the previously input parameters and then again select "optimize" to determine the effect of such a change on the identity of the optimal seed varieties (i.e., to see whether the change in input parameters affects which varieties are identified as optimal) and their profitability distributions.
As should be apparent to those skilled in the art, it is envisioned that the data- driven computer system of this preferred embodiment will grow over time as data for future growing seasons, additional seed varieties, and additional growing factors become available.
A third embodiment of the present invention will now be described with reference to Figs. 9-21. In this embodiment, which focuses on the selection of corn and soybean seed varieties for growing operations within the United States, the user is allowed to define field and operational parameters on a plot-by-plot basis. In other words, the system allows for the possibility that a user may own multiple farms, where each such farm may include multiple fields and each such field may include multiple plots, and that the user may want to develop an optimal planting strategy for each plot so as to maximize overall profitability at the field, farm or firm (i.e., an owner of multiple farms) level. The user interface for this embodiment includes a screen display, shown in
Fig. 9, which is configured for the user to input, via a text box 300, the name of a farm for which the user would like to develop a planting strategy. This screen also includes a drop-list 302 by which the user can select the particular state where the
farm of interest is located. Once this data has been input, the user can select a "save farm" button 304 for saving this input data in the system. The user can then proceed to define one or more fields within the farm of interest. A preferred screen display through which the user can input parameters related to a particular field for the farm identified in Fig. 9 is shown in Fig. 10. As shown therein, this screen includes a text box 306 in which the user can input the name of the field in question, as well as a plurality of check boxes 308, 310, 312, 314, 316 which the user can check, as appropriate, to indicate whether the crop to be grown in this particular field is soybean or corn, and whether soybean, corn or some other crop were last grown in this field. Text boxes 318 and 320 are also provided for the user to input the minimum crop selling price and the maximum crop selling price, respectively, that the user expects to realize for the crop to be grown. Preferably, the system produces default values for these prices from historical data, and the user can either use the default values or enter new values. By selecting the "save field" button 322, the field-related parameters input by the user can be saved.
A preferred screen display through which the user can input parameters related to a particular plot for the field identified in Fig. 10 is shown in Fig. 11. This screen includes a text box 326 in which the user can input the name of the plot in question, as well as text boxes 328 and 330 via which the user can input the size and soil type of the plot, respectively. Check boxes 332, 334, 336 are preferably provided for the user to define the drainage of the plot as being either well, moderate, or poor. Text boxes 338 and 340 are provided for the user to input the soil organic matter and soil pH levels, respectively. By selecting a "save plot" button 342, the plot-related input parameters can be saved. Although the inputting of information for only one farm, one field and one plot has been described above, it should be understood that, in this particular embodiment of the invention, the system is configured for the user to input, if desired, data for several farms, data for several fields within each farm, and data for several plots within each field, as will be further apparent from the description below. After defining a plot within a field within a farm, the user can proceed to develop planting strategies for that plot by entering scenario data into the system via the screen display shown in Figs. 12-14. As shown in Fig. 12, the scenario data input screen includes a text box 350 via which the name of this particular scenario can be
specified. Positioned just below this text box 350 is a display of the crop choice and minimum and maximum crop selling prices input by the user in (or the default values from) Fig. 9. Also shown in Fig. 12 are several text boxes 352-368 via which various fertilizer-related parameters can be input. A drop-list 370 is provided for inputting tillage information, and check boxes 372, 374 are provided for the user to specify whether the plot in question will be irrigated. In text box 376, the user can input the plant population. Referring to Figs. 12 and 13, drop-lists 378, 380 and 384, list box 382, and text boxes 386, 388 are provided for the user to input several desired seed characteristics. Specifically, drop-list 384 is provided for the user to select, if desired, a "key seed" for comparison purposes, as further explained below. In text box 386, the user can input a "bar chart probability" (i.e., a number between 1 and 99), which affects the manner in which the results of the scenario will be produced and displayed, as explained below. In text box 388, the user can input a calibration or scaling factor which will be used to adjust the mean yield (and thus profitability) predicted by the system for this particular scenario. As an example, if the user knows that the particular plot for which this scenario will be run consistently produces a yield 10% greater than similar plots, the user may enter the value 1.1 in text box 388, which will cause the system to increase its projected yield for that plot by 10%. Referring to Figs. 13 and 14, text boxes 400-414 are provided for inputting cash production costs, and text boxes 416-424 are provided for inputting economic production costs. The computer system of this preferred embodiment also includes an expert system to inform the user, for example, when there is no seed variety that is compatible with one or more of the parameters input by the user.
Once the scenario data has been input, the user can proceed to select the "run scenario" button 426, which causes the system to display the scenario summary screen shown in Figs. 15 and 16. In addition to displaying the various parameters related to the growing operation which were input by the user in prior screens, the scenario summary screen includes a predicted average yield chart 450, a predicted average profit chart 460, and a predicted minimum profit chart 470. Preferably, these predictions are displayed in the form of bar charts containing probability intervals, as shown in Figs. 15 and 16. For each chart, several seed varieties identified by the system as being optimal, based upon the parameters input by the user, are listed in the left-most column. In this embodiment, each chart will list the top four seed varieties
(in terms of predicted average yield, predicted average profit, or predicted minimum profit) as well as the key seed for comparison purposes. It may be noted that maximizing minimum profit is a risk avoidance strategy, where greater BCP values reflect greater risk aversion. Preferably, if the user does not select a key seed from the screen display shown in Fig. 13, the system will display the top five seed varieties in charts 450, 460, 470.
The yield chart 450 includes a separate bar 452, 454, 456, 458 for each of the four seed varieties that are projected to produce the highest average yields, as determined by the system, as well as a bar 459 for the "key seed" selected by the user. As shown in Fig. 15, the rankings of the optimal seed varieties and the key seed are preferably listed in the bars 454-459. Note that for the parameters input by the user for this particular scenario, the key seed was ranked #24 in terms of its expected average yield. Preferably, the bar 459 for this key seed is displayed in a different color than bars 452-458 so it can be readily distinguished from the other bars. Each bar represents a probability interval, and is centered about the average yield projected for the corresponding seed variety. For example, bar 459 is centered about 92, which is the average yield expected to result from using the key seed variety (i.e., RX490) in the particular scenario defined by the user (see Fig. 16). The width of each bar or probability interval is such that there is a Bar Chart Probability ("BCP") that the specific yield to be realized for this scenario will fall within the bar. For example, because the BCP in this scenario was set at 90, there is a 90% chance that the actual yield to be realized from using the key seed variety in the defined scenario conditions will fall within the bar 459, and a 10% chance that the actual yield will fall outside the bar 459. Thus, it should be noted that as the BCP increases, the width of the bars in the yield chart 450 will likewise increase. The BCP is used in this same way to construct the profit bar-charts 460, 470.
When the user selects (e.g., using a touch screen, mouse, etc.) one of the bars displayed in a chart such as, for example, bar 459 in the average yield chart 450, the system preferably displays the minimum, average and maximum values 480 for that bar, as shown in Figure 16.
The average profit chart 460 includes a separate bar 462, 464, 466, 468 for each of the four seed varieties that are projected to produce the highest average profits, as determined by the system, as well as a bar 469 for the "key seed" selected
by the user. Similarly, the minimum profit chart 470 includes a separate bar 472, 474, 476, 478 for each of the four seed varieties that are projected to produce the highest minimum profits, as determined by the system, as well as a bar 479 for the "key seed." Although it is possible for the same four (or five) seed varieties to have both the highest projected average profits and the highest projected minimum profits, as in the example shown in Figs. 15 and 16, this will not always be the case. In fact, Fig. 15 shows how the key seed variety selected by the user was ranked #24 in terms of projected average profit, but #23 in terms of projected minimum profit.
Note that for the average yield chart 450 and the average profit chart 460, the optimal seed varieties are identified by the system using the projected values for average yield and average profit, respectively, for each seed variety. Because the BCP value input by the user does not affect the projections of these average values, the BCP value does not affect which seed varieties are identified as optimal in terms of projected average yield or profit. As explained above, however, the BCP value does affect the probability intervals that are represented by the bars displayed in the bar charts 450, 460, 470, and specifically affects the minimum and maximum values for each interval. Accordingly, because the optimal seed varieties listed in the minimum profit chart 470 are identified by ranking each variety with respect to its projected minimum profit, the BCP value input by the user may affect which varieties are deemed optimal in terms of projected minimum profit. By using the BCP as described above and by ranking the seed varieties according to highest average yields, highest average profits, and highest minimum profits, the system provides users having various risk tolerances with a number of ways for identifying the best seed variety and overall planting strategy for a particular scenario. In this particular embodiment of the invention, which uses an analytical-based analysis of a statistical model, the yield is preferably modeled with both additive and multiplicative terms. Let seeds by indexed by i, let multiplicative model variables be indexed by j, and let additive model variables be indexed by k. For the i'th seed, let β be the effect of the j'th variable on yield and let αlk be the effect of the k'th variable on yield. Let Xj and Xk be the input levels of the j'th and k'th variables. Let the model contain K total additive variables, J total multiplicative variables, and I different seeds. Then yield for the i'th seed is given by:
The coefficients βy and αl and the distribution of e are estimated by regression analysis of raw data, inferred from consensus data, or are input directly by the user. As apparent from Figs. 9-14, the specific variables considered by the model include state, maturity, soil type, organic matter, pH, fertilizer, diseases, seeding rates, etc.
Using this yield model, the yield means and variances are computed for each variety. The average yields are then used by the system to identify and rank the top four (or five) varieties in terms of projected average yields, and are also used with the yield variances for each variety, along with the BCP input by the user, to produce the average yield chart 450.
The mean and variance of the crop selling price are computed using the minimum and maximum prices input by the user (or by using the default values, if no values were input). The means and variances of profit for each seed variety are computed using the price and yield means and variances and expected cost (which is obtained from the cost parameters input by the user and the predefined seed prices). Profit is assumed to be logistically distributed. The average profits are then used by the system to identify and rank the top four (or five) varieties in terms of projected average profits, and are also used with the profit variances for each variety, along with the BCP input by the user, to produce the average profit chart 460. Similarly, the minimum profits are used by the system to identify and rank the top four (or five) varieties in terms of projected minimum profits, and are also used with the average and maximum profit values for each variety, along with the BCP input by the user, to produce the minimum profit chart 470.
After reviewing the scenario results shown in Figs. 15 and 16, the user can produce additional scenarios and review their results with a view towards ascertaining the optimal planting strategy for a particular plot. If the user desires to change only a limited number of parameters in an existing scenario to see the effect of these changes on the projected results, the user can conveniently select a "copy" button 484 which is provided at the top of the scenario results screen, as shown in Fig. 14. This causes the system to make a duplicate scenario which the user can save under a new name and
modify as desired before running the new scenario. In this preferred embodiment, the user can create and save up to five scenarios per plot.
The computer system of this particular embodiment also allows the user to aggregate the scenario results from multiple plots, fields and even farms to determine an optimal overall projected profitability. Preferably, where the user has run multiple scenarios for a particular plot, the user is prompted to designate one of the scenarios as the "default" scenario which will be used in the profit aggregation process. For example, in the screen display shown in Fig. 18, the first of two scenarios is designated as the default scenario, which means the average and minimum profits projected for this scenario will be aggregated with the projected average and minimum profits for other plots in the same field. If desired, the user can designate the second of the two scenarios as the default scenario by, in this particular embodiment, selecting "No" in the "Default" column of the table shown in Fig. 18. For each scenario, the optimal seed variety for maximizing average profit and the optimal seed variety for maximizing minimum profit are also provided, as shown in Fig. 18. Selecting the "plot" link 492 in Fig. 18 causes the system to display the screen shown in Fig. 19, which includes a table listing the projected average and minimum profits from the default scenarios for two plots in a particular field. Selecting the "field" link 494 in Fig. 19 causes the system to display the screen shown in Fig. 20, which includes a table listing the projected average and minimum profits for each of two fields in a particular farm. Note that the projected average profit for "Field 1" in Fig. 20 is the aggregate of the projected average profits for the two plots listed in Fig. 19. Selecting the "farm" link 496 in Fig. 20 causes the system to display the screen shown in Fig. 21, which includes a table listing the aggregate average and minimum profits for each of two farms, as well as the sum of those profits (which is referred to as the "firm's total" in Fig. 21 based on the assumption that the two farms are owned by a single firm).
Although the aggregation process in this preferred embodiment is performed for the projected average and minimum profits, the system could instead be configured to perform this aggregation using projected yields, projected maximum profits, projected revenues, etc., if desired.
There are various changes and modifications which may be made to the invention and the exemplary embodiments described above, as apparent to those
skilled in the art. However, these changes and modifications are including within the teachings of the disclosure, and the invention should therefore be limited only by the scope of the claims set forth below, and their legal equivalents.