The present disclosure relates to systems and/or methods for agricultural optimization. Particular embodiments of systems and methods of the present disclosure provide environmental optimization and methods, particularly the area of soil carbon enhancement or increase and/or biomass accumulation in previously unused biomass accumulation regions within farm parcels.
Agriculture is intrinsic to global health and prosperity while simultaneously putting an enormous strain on the planet's water, soil, and human resources, and contributing a significant fraction of global greenhouse gas emissions. However, agricultural lands also have the capacity to reverse or mitigate adverse environmental impacts when subjected to interventions that purposefully reduce an associated pressure on compromised environmental resources. These interventions can shift the environmental impacts of an agricultural system from extractive to regenerative.
Cover crops have been used to enhance the productivity of commodity crops. Cover cropping is an intervention in which complementary, unharvested vegetation is planted in an agricultural system for the purpose of enhancing its agro-economy through any combination of lowering required inputs, mitigating agronomic pressures, or regenerating environmental attributes such as soil health. In specialty crop systems where the commodity crop is planted in rows, cover crops can be planted in the alleyways between rows and sometimes underneath the trees or vines of the cash crop. Cover crops can be annual or perennial and multiple species can be planted as a mix in the same space. As farmers weigh costs with benefits and consider the barriers to trying unfamiliar practices, adoption rates for cover cropping remain low in specialty crop systems. Perceived competition for water and soil resources are significant drawbacks, as is the cost of clearing and re-seeding annual varieties each year.
Despite low adoption rates, the upside of cover cropping is very favorable to certain types of farming operations, such as those practicing no-till. In no-till systems, soil is left undisturbed by tillage, providing a variety of agronomic benefits including erosion control. A cover crop can raise soil carbon content by absorbing the greenhouse gas carbon-dioxide through photosynthesis and depositing the resulting organic carbon byproducts into the soil through its root systems and by interaction with other biological processes. This soil carbon can be temporarily removed from the atmosphere when combined with the practice of no-till. A need exists for a trustworthy cover cropping system that does not impede agricultural productivity and complements existing biological and agronomic farm schedules while providing an incentive for implementation by farmers.
Systems, methods, and/or processes for agricultural parameter determination are provided. One or more can include or use processing circuitry configured to: manage one or more land parcels for agricultural parameter optimization (such as soil organic carbon optimization); and determine one or more agricultural parameters of at least a portion of one land parcel (such as an alleyway between commodity crops) during a first time period without sampling the portion of the one land parcel.
The determining can include: collecting target agricultural data from collection sites of at least one portion of one land parcel, the collecting comprising compiling target agricultural data and candidate predictor parcel data associated with the collection sites; associating a subset of the candidate predictor parcel data with points throughout the portions of the land parcels; and processing both the compiled target agricultural data and the subset of predictive parcel data to generate target agricultural data for unsampled parcel portions of the one land parcel or of another, unsampled, land parcel.
The processing can include: building a target agricultural data model from collected target agricultural data and the subsets of predictive data; and applying the target agricultural data model to determine one or more agricultural parameters of a portion of the at least one land parcel.
Systems, methods, and/or processes for increasing soil organic carbon are also provided. They can include: managing a parcel having commodity crops planted in rows, the commodity crops having a dormant season; defining alleyways between the rows of commodity crops; planting a cover crop within the alleyways, the cover crops having a growing season within the dormant season of the commodity crop, and a dormant season within the growing season of the commodity crop; and increasing the soil organic carbon content of the parcel during the dormant season of the commodity crop with the cover crop.
Systems, methods, and/or processes for increasing soil organic carbon within land parcels having commodity crops separated by alleyways are also provided. They can include processing circuitry configured to: manage one or more land parcels for agricultural parameter optimization, the land parcels having commodity crops in rows with alleyways therebetween; and determine carbon per acre of at least one land parcel during a first time period without sampling the one land parcel.
The determining can include: collecting soil organic carbon data from collection sites of alleyways of another land parcel, the collecting can include: determining collection sites and compiling soil organic carbon data from the collection sites; associating a subset of the candidate predictor parcel data with points throughout the alleyways of the land parcels; and processing both the compiled soil organic data and the subset of predictive parcel data to generate soil organic data for unsampled alleyways. The processing can include: building a soil organic carbon data model from collected soil organic data and the subsets of predictive data; and applying the soil organic data model to determine soil organic carbon of the alleyway of the at least one land parcel.
Embodiments of the disclosure are described below with reference to the following accompanying drawings.
This disclosure is submitted in furtherance of the constitutional purposes of the U.S. Patent Laws “to promote the progress of science and useful arts” (Article 1, Section 8).
The present disclosure will be described with reference to
The method includes an intervention 14 into the area. This intervention can be the preparation of any kind of agricultural treatment, but in particular embodiments, environmentally positive agricultural treatments, such as but not limited to, the use of cover crops as will be detailed later by example.
The method can continue by defining initial condition 16. This can include collecting target agricultural data for at least one other land parcel, the collecting comprising determining sampling sites and compiling target agricultural data associated with the sampled sites. As an example, using soil data from publicly available, soil databases (e.g., gNATSGO), vegetation indices from drone imagery, and elevation data from public datasets can be processed by aggregating data and deriving the number of statistically different units or zones present within the parcel or parcel portion and then deriving where to sample within these units in order to best approximate the unit's statistical distribution. At least one example of this is depicted and described with reference to
The method continues with data collection of target variables 18. This can include the collection of target agricultural data for at least one portion of another land parcel or portion. This can be a collection of Soil Organic Carbon from an alleyway, for example. The collecting can include determining the collection sites and compiling target agricultural data associated with the samples acquired from the collection sites. This target agricultural data can include Total Carbon, Total Organic Carbon, Soil Organic Carbon, Calcium, Magnesium, and/or Nitrogen.
The method continues with collecting predictor variables based on locations of target variables 20. For example, candidate predictor parcel data can be compiled and associated with the collected target agricultural data samples acquired from the collection sites.
Predictor variables can include but are not limited to predictors derived from drone imagery and summarized over space across statistical metrics including but not limited to minimum, mean, maximum, standard deviation of various spectral indices including but not limited to CI, MCARI, NDWI, VARI, kNDVI, NDVI, SAVI, GNDVI, ENDVI, LCI, EVI, NIRv, GLI, CVI, CI RedEdge, and NDRE; predictors derived from various satellite imagery (e.g., ESA Copernicus Sentinel-1, ESA Copernicus Sentinel-2, PlanetScope, MODIS, etc) and summarized over space and time across statistical metrics including but not limited to minimum, mean, maximum, standard deviation of various spectral indices including but not limited to SCI, CI, NDWI, VARI, kNDVI, SAVI, ENDVI, LCI: B5, B6, B7; NIRv; GLI, CVI, CI RedEdge: B5, B6, B7; NDRE: B5, B6, B7; 3BSI, mND, MCARI, IRECI, NDVI, S2REP, SR, GNDVI, MDVI, MSI, EVI; predictors derived from sunlight hours at the location between the start of the growing season identified from the first time of rain and the drone flight date: average sunlight hours, total sunlight hours; predictors derived from daily meteorological variables and summarized over time across statistical metrics including but not limited to minimum, mean, maximum, standard deviation) of five variables (precipitation, shortwave radiation, maximum temperature, minimum temperature, vapor pressure); predictors derived from location information: system (e.g., “vineyard”, “orchard”), age, fertilizer rate, pure live seed goal, year, predictors derived from gNATSGO: taxorder, taxsuborder, taxgrtgroup, taxsubgrp, taxpartsize, ksat_r, awc_r, dbovendry_r, wthirdbar_r, wfifteenbar_r, kwfact, kffact, claytotal_r, om_r, musym; predictors derived from 12 SoilGrids250m: wrb, ocs, bdod, cec, cfvo, clay, nitrogen, phh2o, sand, silt, soc, ocd across four quantiles, Q0.05, Q0.5, mean and Q0.95, and 6 various depths: 0-5, 5-15, 15-30, 30-60, 60-100, and 100-200 cm (ocs is only distributed at depth 0-30 cm); and/or predictors derived from 13 Polaris (30-m resolution) variables: ph, om, clay, sand, silt, bd, hb, n, alpha, ksat, lambda, theta_r, theta_s across 5 quantiles (mean, mode, p50, p5, p95) at 3 depths (0-5, 5-15, 15-30 cm). Statistical metrics for summarizing drone imagery, satellite imagery and daily weather data can be automatically determined by the processing circuitry.
The method continues with processing and modeling these variables including machine learning 22. Accordingly, both the compiled target agricultural data and the predictive parcel data are processed to generate target agricultural data for unsampled parcel portions. The processing can include selecting a predictive parcel data subset from the predictive parcel data. The selection of the subset favors a parsimonious model that accomplishes the desired level of explanation or prediction with as few predictor variables as possible. The goodness of fit of a statistical model describes how well it fits a set of observations. The processing can include building a target agricultural data model using the sample data acquired at the collection sites and the subset of predictive data, and applying the target agricultural data model to determine one or more agricultural parameters of the unsampled land parcel.
The processing can include determining additional sample sites for the one land parcel and additional sample sites for the other land parcel. This can include selecting unsampled sites for target variable prediction.
The predictive parcel data subset can be associated with portions of the sampled and/or unsampled parcels. For example sampled alleyways and/or unsampled alleyways. Accordingly, the subset of predictive parcel data can be associated with a myriad of points throughout the parcel or parcel portion. The model can be derived using the subset of predictive data and the sampled data or the entirety of the predictive data and the sampled data.
The model can then be applied to the target agricultural data associated with the collection sites and/or unsampled land parcels or portions to determine the one or more agricultural parameters of the at least one land parcel.
As shown the method can continue with additional predictions including but not limited to annual predictions of target variables 24, and then a calculation of annual change 26, and finally quantifying the environmental benefits 28.
Referring next to
Referring next to
P6
13
1
11
34
P7
25
8
0.1
0
P8
29
10
0.01
2
In accordance with the below reference, a second statistical algorithm is run for each zone that identifies where within the sampling zone to best represent the statistical distribution of the sampling zones. Processes for determining sample sites can also include those disclosed in Minasny, B., & McBratney, A. B. (2006). A conditioned Latin hypercube method for sampling in the presence of ancillary information. Computers & geosciences, 32(9), 1378-1388 the entirety of which is incorporated by reference herein.
In italics of Table 1 are group three data corresponding to group 3(P7-P9) of
The processing circuitry of the system of the present disclosure is not limited to that depicted in
The computing system may also include a web server that can be of any type of computing device adapted to distribute data and process data requests. The web server can be configured to execute system application software such as the reminder schedule software, databases, electronic mail, and the like. The memory of the web server can include system application interfaces for interacting with users and one or more third party applications. Computer systems of the present disclosure can be standalone or work in combination with other servers and other computer systems that can be utilized, for example, with larger corporate systems such as financial institutions, insurance providers, and/or software support providers. The system is not limited to a specific operating system but may be adapted to run on multiple operating systems such as, for example, Linux and/or Microsoft Windows. The computing system can be coupled to a server and this server can be located on the same site as the computer system or at a remote location, for example.
In accordance with example implementations, these processes may be utilized in connection with the processing circuitry described. The processes may use software and/or hardware of the following combinations or types. For example, with respect to server-side languages, the circuitry may use Java, Python, PHP, .NET, Ruby, JavaScript, Golang, R, or Dart, for example. Some other types of servers that the systems may use include Apache/PHP, .NET, Ruby, NodeJS, Java, R, Golang, and/or Python. Databases that may be utilized are Oracle, MySQL, SQL, NoSQL, or SQLite (for Mobile) as well as any type of relational, key-value, in memory, document, wide column, graph, time series, or ledger databases. Client-side languages that may be used, this would be the user side languages, for example, are ASM, C, C++, C#, Java, Objective-C, Swift, ActionScript/Adobe AIR, or JavaScript/HTML5. Communications between the server and client may be utilized using TCP/UDP Socket based connections, for example, as Third-Party data network services that may be used include GSM, LTE, HSPA, UMTS, CDMA, WiMAX, WIFI, Cable, and DSL. The hardware platforms that may be utilized within processing circuitry include embedded systems such as (Raspberry PI/Arduino), (Android, iOS, Windows Mobile), phones and/or tablets, or any embedded system using these operating systems, i.e., cars, watches, glasses, headphones, augmented reality wear, etc., or desktops/laptops/hybrids (Mac, Windows, Linux). The architectures that may be utilized for software and hardware interfaces include x86 (including x86-64), or ARM.
The systems and/or processing circuitry of the present disclosure can include a server or cluster of servers, one or more devices, additional computing devices, several network connections linking devices to server(s) including the network connections, one or more databases, and a network connection between the server and the additional computing devices, such as those devices that may be linked to an adjuster.
Devices and/or processing circuitry and/or plurality of devices and the additional computing device can be any type of communication devices that support network communication, including a telephone, a mobile phone, a smart phone, a personal computer, a laptop computer, a smart watch, a personal digital assistant (PDA), a wearable or embedded digital device(s), a network-connected vehicle, etc. In some embodiments, the devices and the computing device can support multiple types of networks. For example, the devices and the computing device may have wired or wireless network connectivity using IP (Internet Protocol) or may have mobile network connectivity allowing over cellular and data networks.
The various networks may take the form of multiple network topologies. For example, networks can include wireless and/or wired networks. Networks can link the server and the devices. Networks can include infrastructure that support the links necessary for data communication between at least one device and a server. Networks may include a cell tower, base station, and switching network as well as cloud-based networks.
Referring next to
At t3, domain 90 has target agricultural data 120 collected to expand modeling domain. This data a is processed with predictive parcel data to determine modeled data 122. Additionally, at t3, sample for performance is taken at 124. At each iteration the target agricultural model is updated. Referring next to
Referring to
Referring next to
As shown in
In accordance with example implementations, to tables 4-9, the processing circuitry utilizing publicly available machine learning programs such as any ensemble or combination of a time series analysis, a spectral analysis, a network analysis, a correlation analysis, a generalized linear model, a generalized additive model, a nearest-neighbor model, a decision tree model, a support vector machine model, a Bayesian model, a Gaussian processes model, an artificial neural network, a recurrent neural network, a convolutional neural network, a generative adversarial neural network, a transformer model, and with operations such as a clustering operation, a regression operation, a classification operation, a feature selection operation, a feature engineering operation, a spatio-temporal resampling operation, a benchmarking operation, a training operation, a testing operation, a validation operation, a tuning operation, a genetic algorithm operation, a Monte-Carlo operation, a bagging operation, an ensembling operation, a gradient boosting operation, a regularization operation, a supervised learning operation, an unsupervised learning operation, a self-supervised learning operation, a semi-supervised learning operation, a reinforcement learning operation, and a sequence-to-sequence learning operation
As an example and with reference to Table 4 multiple samples were taken at collection sites of parcels 190 and 200.
In accordance with example implementations, the target data of Table 4 samples were associated with a myriad of candidate predictor data associated with the collection site at 183, for example. The sample data is shown associated with candidate predictor data in Table 5.
These 68 predictor values as shown in Table 5 are summarized over space across four summary metrics. There are also 128 predictors derived from Sentinel-2 imagery summarized over time and then two predictors derived from sunlight hours, 20 predictors derived from daily meteorological summarized over time, five predictors derived from location information and 15 predictors derived from gNATSGO as well as 245 predictors derived from soil grids 250 variables. Predictors can also include legacy predictors, two of which are derived from clay content as well.
Processing circuitry being supplied with this multitude of predictors and the analysis points given can utilize a threshold for predictors that are informative in determining or have a sufficient impact on determination and the predictors that are not utilized. This generates a subset of candidate predictor variables. A selection of points throughout the parcels are associated with this subset and shown in Table 6 and modeled data provided in Table 8. Table 6 represents exemplary 100 points distributed within the parcels. As many points as practical can utilizing informative predictor values. After determining the subset, data as shown in Table 6 can be provided for parcels or portions of parcels as shown in Step 2. For example, the V4's represent selected predictor values within parcel 200 and the X2's indicate selected predictor values within parcel 190. Additionally, the predictor values as shown in location are shown for V3 are provided for parcel 300 and these are the predictor values that can be used to determine what the soil organic carbon is for a plot or parcel that has not been sampled.
As shown in Table 7, the testing of the two machine learning programs (catboost and a featureless, mean, baseline model) are shown. This testing can include but is not limited to the use of the following equations.
Mean Squared Error (MSE)
The Mean Squared Error is defined as
Mean Absolute Error (MAE)
The Mean Absolute Error is defined as
Mean Absolute Percent Error (MAPE)
The Mean Absolute Percent Error is defined as
Bias
The Bias is defined as
Relative Squared Error (RSE)
The Relative Squared Error is defined as
The performance of the trained CatBoost model demonstrates learning from data in comparison with the performance of the featureless, mean, baseline model.
Referring next to Table 8, as shown the additional mean of the analysis points are the data for 280 is shown as the observed standard deviation and model mean and the model standard deviation. The same thing is shown for plot 283 as X2 and plot 282 as V3. As there is no observed mean at V3, the modeled mean is 16.06 and the modeled standard deviation is 0.07.
Referring next to
In accordance with example implementations, this cultivar can be planted in alleyways between commodity crops. In some parcels in some portions of alleyways, target agricultural data such as carbon content can be sampled and determined. Without actually sampling other portions of parcels or other portions of alleyways, the systems and methods of the present disclosure can be used to provide modeled data with the sample or different alleyways of the same or different parcels. This modeled data can be used to determine the amount of carbon being sequestered by the cover crop.
Accordingly, methods for sequestering carbon are provided. The methods can include: identifying a parcel having commodity crops planted in rows, the commodity crops having a dormant season; defining the alleyways between the rows of commodity crops; planting a cover crop within the alleyways, the cover crops having growing season within the dormant season of the commodity crop, and a dormant season within the growing season of the commodity crop; and increasing the carbon content of the parcel during the dormant season of the commodity crop with the cover crop. The commodity crop can be a vineyard or orchard. The cover crop can be Poa bulbosa or hybrid of same. The cover crop can be retained between commodity crop growing seasons.
Systems for agricultural carbon sequestration are also provided. They system processing circuitry can be configured to: identify one or more land parcels for agricultural parameter optimization, the land parcels having commodity crops in rows with alleyways therebetween; and determine carbon per acre of parcel of at least one land parcel during a first time period without sampling the one land parcel, the determining can include: collecting target agricultural data for at least one other land parcel, the target agricultural data including total carbon and the collecting comprising determining sampling sites and compiling target agricultural data associated with the sampled sites; collecting predictive parcel data associated with the target agricultural data; and processing both the compiled target agricultural data and the predictive parcel data to generate target agricultural data for unsampled parcel portions, the processing can include: selecting a predictive parcel data subset from the predictive parcel data, the selecting comprising identifying impactful predictive parcel data for selection; determining sample sites for the one land parcel and additional sample sites for the other land parcel; associating the predictive parcel data subset with both the sample sites for the one land parcel and the additional sample sites for the other land parcel to form a target agricultural data model; and applying the model to the carbon data associated with the sampled sites to determine the total carbon of the at least one land parcel.
In compliance with the statute, embodiments of the invention have been described in language more or less specific as to structural and methodical features. It is to be understood, however, that the entire invention is not limited to the specific features and/or embodiments shown and/or described, since the disclosed embodiments comprise forms of putting the invention into effect.
This application claims priority to and the benefit of U.S. Provisional Patent Application Ser. No. 63/274,668 filed Nov. 2, 2021, entitled “Bulbosa 7PB55 Cultivar Cover Crop Systems and Methods for Northern and Central California Wine Grape Crops” and U.S. Provisional Patent Application Ser. No. 63/274,625 filed Nov. 2, 2021, entitled “System and/or Methods for Cover Crop Application, Determining Cover Crop Performance and/or Carbon Credits Generation from Cover Crop Application”, the entirety of each of which is incorporated by reference herein.
Number | Date | Country | |
---|---|---|---|
63274625 | Nov 2021 | US | |
63274668 | Nov 2021 | US |