Hybrid model for optimizing fungicide application schedules

Abstract

The invention relates to fungal disease management. In order to improve fungal disease management, a computer-implemented method (200) for determining disease progression for a fungicide spraying schedule on a farmland is provided. The method comprises a step of receiving (210) data including crop variety data, environmental data, crop management data and location data of the field. The crop variety data is related to a crop growing or to grow on a farmland. The environmental data indicates environmental conditions of the farmland. The crop management data indicates a fungicide spray history of the farmland. The method further comprises a step of applying (220) a machine learning model to the received data to determine disease progression time series data of the fungal disease, the machine learning model has been trained to learn disease progression under conditions defined by crop variety data, environmental data, crop management data, and location data based on historical data collected from one or more farmland. The method further comprises a step of determining (230) a disease onset date of the fungal disease based on the determined disease progression time series data.

Description

Technical Field

[0001] The present invention relates to fungal disease management. In particular, the present invention relates to a computer-implemented method and apparatus for determining disease progression for fungicide spray scheduling on agricultural fields, and to a computer program element. Background Art

[0002] Fungicides are used in agriculture. Fungal diseases in plants are numerous and affect all parts of the plant, i.e., the head (e.g., common root rot, crown rot, wilt), stem or sheath (e.g., eye spot, stem rust), leaves (e.g., leaf blight, rust, powdery mildew), ears (e.g., ergot) and seeds (e.g., black spot, carnal bunt). It is generally recommended that the timing of fungicide application is before the start of inoculation (primary infection). However, the start date of primary infection with crop diseases may vary from season to season, depending on the availability of inoculum and weather conditions. Therefore, fungal disease management becomes more complicated. Summary of the invention

[0003] Improved fungal disease management may be needed.

[0004] The objects of the invention are achieved by the subject-matter of the independent claims:

[0005] Further embodiments are included in the dependent claims. It should be noted that

[0006] The aspects of the invention described below also apply to computer-implemented data processing methods and devices, computer program elements and computer-readable media.

[0007] According to a first aspect of the invention, there is provided a computer-implemented method for determining disease progression for fungicide spray scheduling on agricultural fields, the method comprising:

[0008] a) receiving data, the data including:

[0009] - Crop variety data relating to crops growing or to be grown on the farmland;

[0010] - Environmental data indicating the environmental conditions of the agricultural land;

[0011] - Crop management data indicating the fungicide spraying history of the field; and

[0012] - Location data of the farmland;

[0013] b) applying a machine learning model to the received data to determine disease progression time series data of a fungal disease, wherein the machine learning model has been trained to learn the disease progression under conditions defined by crop variety data, environmental data, crop management data, and location data based on historical data collected from one or more agricultural fields; and

[0014] c) determining a disease onset date of the fungal disease based on the determined disease progression time series data.

[0015] Computer-implemented methods and apparatus as described herein apply machine learning models to identify first infection—that is, the start date of first infection with a crop disease.

[0016] The machine learning model may be any suitable model capable of determining disease progression time series data for fungal diseases. Examples of machine learning models may include, but are not limited to, XGB regression models, artificial neural networks, recursive neural networks, and support vector regression.

[0017] The start date of the primary infection of a crop disease (also called disease onset) can be used to determine the disease progression curve. For example, a process-based model can be used to simulate the interaction between the disease onset date and variety susceptibility to determine secondary infection, which will be explained in detail below.

[0018] Determining disease onset and disease progression curves to characterize the progression of disease over time is essential for understanding how plant diseases develop and how disease control measures should be taken to achieve the highest efficacy of fungicides. Based on this information, fungicide spraying arrangements can be determined. In this way, spraying timing is set by taking into account disease dynamics, thereby improving treatment. For example, spraying timing can be set based on predicted disease risk, allowing fungicides to be applied when fungicides are most effective during the growing season. In particular, the determined fungicide spraying arrangements can better align the optimal application period to prevent disease progression while reducing the chance of missing the risk period. Optimized spraying timing can have a considerable impact on the overall reduction in fungicide use, which in turn can reduce damage to agriculture, thereby reducing losses in yield, quality and profits.

[0019] According to an embodiment of the present invention, the disease onset date of a fungal disease is determined using a change point detection algorithm.

[0020] According to an embodiment of the present invention, multiple machine learning models are provided for two or more fungal diseases, and each machine learning model has been trained for a single disease.

[0021] In this way, each machine learning model is specifically trained to determine disease onset and disease progression curves to characterize the progression of a specific fungal disease over time.

[0022] According to an embodiment of the present invention, the machine learning model includes an extreme gradient boosting (XGB) regression model.

[0023] In some examples, the model selection can be an XGB regression model tuned by random search cross validation. XGB is a scalable tree boosting system and can be viewed as gradient boosting with regularization. It can implement parallel tree construction, cache-aware access, sparsity awareness, and weighted quantile sketches as part of its system optimization and algorithm enhancement.

[0024] According to an embodiment of the present invention, the computer-implemented method further comprises d) applying a process-based model to determine the infection rate of the fungal disease after the disease onset date under conditions defined by the crop variety data, the environmental data and the location data.

[0025] In other words, a process-based model is proposed to simulate secondary infection through a process including susceptibility, exposure, infection and recovery. The infection rate takes into account environmental factors including temperature and humidity. Variety susceptibility, fungicide and crop growth stage are also considered in the infection rate calculation. The process-based model can be used to simulate the interaction between disease onset date, variety susceptibility and fungicide application. The infection rate of fungal diseases can be provided in the form of a disease progression curve, which can preferably be used to optimize the scheduling of fungicides.

[0026] According to an embodiment of the present invention, the infection rate of the fungal disease is determined by further including conditions defined by the cultivar disease resistance level of the crop.

[0027] According to an embodiment of the present invention, the infection rate of the fungal disease is determined by further including conditions defined by fungicide application data including fungicide data of a fungicide product to be used and at least one planned application timing.

[0028] In other words, process-based models can be used to simulate the interactions between disease onset date, cultivar susceptibility, and fungicide application. Fig.25 The influence of the curve shown in Figure 2 can determine the fungicide spraying schedule.

[0029] According to an embodiment of the present invention, the process-based model includes a Susceptible-Exposed-Infected-Recovered (SEIR) model.

[0030] According to an embodiment of the present invention, the crop variety data includes one or more of the growth stage of the crop, the number of days after planting, and the variety disease resistance level of the crop.

[0031] According to an embodiment of the present invention, the environmental data includes one or more of the following: air temperature, cloud cover, shortwave radiation, longwave radiation, ice accumulation period, liquid accumulation period, relative humidity, adjusted precipitation accumulation period, snow accumulation period and wind speed.

[0032] According to an embodiment of the present invention, the location data includes latitude and longitude data.

[0033] According to an embodiment of the present invention, the computer-implemented method further comprises e) determining a fungicide spraying schedule based on said determined disease progression.

[0034] In some examples, the fungicide spray schedule may include spray application timing data.

[0035] In some examples, a fungicide spraying schedule can be provided to a farmer, who then controls a sprayer to apply the fungicide based on the spray application timing data.

[0036] In some examples, a configuration file is generated based on the fungicide spraying schedule. The configuration file can be loaded into the sprayer to configure the sprayer to apply the fungicide according to the spray application timing data.

[0037] According to an embodiment of the present invention, the computer-implemented method further comprises f) generating a configuration file based on the fungicide spraying schedule, wherein the configuration file is preferably usable for configuring a sprayer for fungicide spray application.

[0038] According to a second aspect of the invention, there is provided an apparatus for generating disease progression for use in fungicide spraying arrangements on agricultural fields, the apparatus comprising one or more processing units for generating application schemes, wherein the one or more processing units comprise instructions which, when executed on the one or more processing units, perform the method steps as claimed in any one of the preceding claims.

[0039] Generally speaking, the functions of any one or more components of the apparatus may be implemented by any suitable computing environment, such as a personal computing environment, a time-sharing computing environment, a distributed computing environment, a cloud computing environment, and a cluster computing environment.

[0040] In some examples, the apparatus is implemented as or in a device or apparatus, such as a server, workstation, or mobile device.

[0041] In some examples, the apparatus may be implemented with or without employing a processor, and may also be implemented as a combination of dedicated hardware to perform some functions and a processor (eg, one or more programmed microprocessors and associated circuitry) to perform other functions.

[0042] In some examples, the device may also be implemented in a distributed manner. For example, some or all units of the device may be arranged as separate modules in a distributed architecture and connected in a suitable communication network.

[0043] According to a third aspect of the present invention, there is provided a computer program element comprising instructions for causing an apparatus according to the second aspect to perform the steps of a method according to the first aspect and any related examples.

[0044] As used herein, the term "farmland" is understood to be any area where organisms (particularly crop plants) are produced, grown, sown and / or planned to be produced, grown or sown. The term "farmland" also includes horticultural fields, afforestation fields and fields used for the production and / or growth of aquatic organisms.

[0045] As used herein, the term "crop" is a plant that can be grown and harvested extensively for profit or livelihood. Examples of crops may include, but are not limited to, onions (Allium cepa), pineapples (Ananas comosus), peanuts (Arachis hypogaea), asparagus (Asparagus officinalis), oats (Avena sativa), sugar beets (Betavulgaris spec.altissima), sugar beets (Beta vulgaris spec.rapa), rapeseed (Brassicanapus var.napus), rutabaga (Brassica napus var.napobrassica), turnip (Brassica rapavar.silvestris), kale (Brassica oleracea), black mustard (Brassica nigra), big leaf tea (Camellia sinensis), safflower (Carthamus tinctorius), American pecan (Caryaillinoinensis), lemon (Citrus limon), sweet orange (Citrus sinensis), small fruit coffee (Coffea arabica) (Coffea canephora, Coffea liberica), cucumber (Cucumissativus), dog tooth root (Cynodon dactylon), carrot (Daucus carota), oil palm (Elaeisguineensis), European strawberry (Fragaria vesca), soybean (Glycine max), upland cotton (Gossypiumhirsutum), (tree cotton (Gossypium arboreum), grass cotton (Gossypium herbaceum), Gossypiumvitifolium), sunflower (Helianthus annuus), rubber tree (Hevea brasiliensis), barley (Hordeum vulgare), hops (Humulus lupulus), sweet potato (Ipomoea batatas), walnut (Juglans regia), lentil (Lens culinaris), flax (Linum usitatissimum), tomato (Lycopersicon lycopersicum), apple (Malus) species, cassava (Manihot esculenta), alfalfa (Medicago sativa), Musa (Musa) species, tobacco (Nicotiana tabacum) (N.rustica), olive (Oleaeuropaea), rice (Oryza sativa), golden bean (Phaseolus lunatus), bean (Phaseolus vulgaris), European spruce (Picea abies), pine (Pinus) species, pistachio (Pistacia vera), pea (Pisum sativum), European sweet cherry (Prunus avium), peach (Prunus persica), pear (Pyrus communis), apricot (Prunus armeniaca), European sour cherry (Prunus cerasus), almond (Prunus dulcis) and European plum (Prunus domestica), currant (Ribes sylvestre), castor (Ricinus communis), sugarcane (Saccharum officinarum), rye (Secale cereale), white mustard (Sinapis alba), potato (Solanum tuberosum), sorghum (Sorghum bicolor) (sorghum (s. vulgare)), cocoa (Theobroma cacao), red clover (Trifolium pratense), common wheat (Triticum aestivum), triticale (Triticale), durum wheat (Triticum durum), broad bean (Vicia faba), grape (Vitis vinifera) and maize (Zea may). The most preferred crops are: groundnut, sugar beet (Beta vulgaris spec.altissima), rapeseed, kale, lemon, orange, small fruit coffee (Coffee cane, coffee macro), bermudagrass, soybean, upland cotton, (cotton tree, cotton, Gossypium vitifolium), sunflower, barley, walnut, lentil, flax, tomato, apple species, alfalfa, tobacco (Nicotiana vulgaris), olive, rice, golden bean, bean, pistachio, pea, almond, sugarcane, rye, potato, bicolor sorghum (sorghum), triticale, common wheat, durum wheat, broad bean, grape and maize. Preferred crops are cereal crops, maize, soybean, rice, rapeseed, cotton, potato, peanut or perennial crops. .

[0046] In a preferred embodiment of the present invention, the disease or plant disease may be caused by, for example, one or more of the following factors:

[0047] Albugo species (white rust) on ornamentals, vegetables (e.g., A. candida) and sunflower (e.g., A. tragopogonis); vegetables (e.g., A. dauci or A. porri), rapeseed (A. brassicola or A. brassicae), sugar beet (A. tenuis), fruit (e.g., A. grandis), rice, soybean, potato and tomato (e.g., A. solani, A. rapa), Alternaria species (Alternaria leaf spot) on citrus or A. alternata), tomatoes (e.g., A. solani or A. alternata), and wheat (e.g., A. triticina); Aphano-myces species on sugar beets and vegetables; Ascochyta species on cereals and vegetables, e.g., A. tritici (anthracnose) on wheat and A. hordei on barley; Aureobasidium zeae on corn (e.g., A. tritici) on corn (anthracnose); and A. hordei on barley (e.g., A. triticina). zeae) (synonym Kapatiella zeae); Bipolaris species and Drechslera species (telemorph: Cochliobolus species), for example small spot (D. maydis) or large spot (B. zeicola) on maize, for example leaf spot (B. sorokiniana) on cereals and B. oryzae, for example on rice and grass; Blumeria graminis (formerly known as Erysiphe graminis) (powdery mildew) on cereals, for example on wheat or barley; Botrytis cinerea on fruits and berries, for example strawberries, vegetables, for example lettuce, carrots, celery and cabbage. cinerea (teeromorph: Botryotinia fuckeliana: gray mold); B. squamosa or B. allii on onion family, rapeseed, ornamentals (e.g. Be eliptica), vines, forest plants and wheat; Bremialactucae (downy mildew) on lettuce; Ceratocystis (synonym Ophiostoma) species (rot or wilt) on broadleaf and evergreen trees, e.g. C.ulmi) (Dutch elm disease); Cercospora species on corn (e.g. gray leaf spot: C. zeae-maydis), rice, sugar beet (e.g. C. beticola), sugar cane, vegetables, coffee, soybean (e.g. C. sojina or C. kikuchii), and rice (Cercospora leaf spot); Cladobotryum (synonym Dactylium) species on mushrooms (e.g. C. mycophilum (formerly Dactylium dendroides, teleomorph: Nectria albertinii, Nectria rosella synonym Hypomyces rosellus; Cladosporium species on tomatoes (e.g., C. fulvum: leaf mold) and cereals, such as C. herbarum (blackear) on wheat; Claviceps on cereals purpurea) (ergot); Cochliobolus (anamorph: Helminthosporium of Bipolaris) species on corn (C. carbonum), cereals (e.g. C. sativus, anamorph: B. sorokiniana), and rice (e.g. C. miyabeanus, anamorph: H. oryzae) (leaf spot); cotton (e.g. C. gossypii), corn (e.g. C. gram inicola: anthracnose stem rot), soft fruit, potato (e.g. C. coccodes: black spot), bean (e.g. C. lindemuthi-anum), soybean (e.g. C. truncatum or C. gloeosporioides), vegetable (e.g. C. lagenarium or C. capsici), fruit (e.g. C. acutatum), coffee (e.g. C. coffeanum or C. kahawae), and various crops (C.gloeosporioides) on rice (anthracnose); Corticium species, e.g. C. sasakii on rice (sheath blight); Coryne-sporacassiicola (leaf spot) on soybeans, cotton and ornamentals; Cycloconium species, e.g. C. oleaginum on olive trees; Fruit trees, vines (e.g. C. lirio-dendri, teleomorph: Neonectria liriodendri: Black Foot Disease) and Cylindrocarpon species on ornamentals (e.g. fruit tree canker or vine blight, teleomorphs: Nectria or Neonectria species); Dematophora necatrix (teleomorph: Rosellini necatrix) on soybeans (root and stem rot); Diaporthe species, e.g. D. phaseolorum on soybeans (damping-off); corn, cereals such as barley (e.g. D. teres, net blotch) and wheat (e.g. D. tritici-repentis: tan blotch) on wheat and sedge. spot), Drechslera (synonym Helminthosporium, teleomorph: Pyrenophora) species on rice and turfgrass; Esca (dieback, apoplexy) on vines, caused by Formitiporia punctata (synonym Phellinus punctata), F. mediterranea, Phaeomoniella chlamydospora (formerly Phaeoacremonium chlamydosporum), Phaeoacremonium aleophilum and / or Botryosphaeria obtusa; pome fruit (E. pyri), soft fruit (E. cyprinis), and soft fruit (E.veneta: anthracnose) and Elsinoe species on vines (E. ampelina: anthracnose); Entyloma oryzae (leaf smut) on rice; Epicoccum species (black mold) on wheat; Erysiphe species (powdery mildew) on sugar beets (E. betae), vegetables (e.g. E. pisi) such as cucurbits (e.g. E. cichoracearum), cabbage, rapeseed (e.g. E. cruciferarum); Eutypa lata (Eutypa canker or dieback, anamorph: Cytosporina spp.) on fruit trees, vines and ornamental trees lata, synonym Liber-tella blepharis; Exserohilum (synonym Helminthosporium) species on corn (e.g. E. turcicum); Fusarium (temomorph: Gibberella) species on various plants (wilt, root or stem rot), such as F. graminearum or F. culmorum on cereals (e.g. wheat or barley) (root rot, scab or head blight). blight), F. oxysporum on tomatoes, F. solani (F. sp. glycines, new synonym F. virguliforme) and F. tucumaniae and F. brasiliense (all causing sudden death syndrome) on soybeans, and F. verticillioides on maize; Gaeumanno-myces graminis (take-all) on cereals (e.g. wheat or barley) and maize; Gibberella species on cereals (e.g. G. zeae) and rice (e.g. G. fujikuroi: Bakanae disease); Glomerella cingulata on vines, pome and other plants, and G.gossypii; Grain-staining complex on rice; Guignardia bidwellii (black rot) on vines; Gymnosporangium species on rosaceae and junipers, such as G. sabinae (rust) on pears; Helmintho-sporium species (synonym Drechslera, teleomorph: Cochliobolus) on corn, cereals, potatoes and rice; Hemileia species, such as H. vaportrix (coffee leaf rust) on coffee; Isariopsis clavispora (synonym Cladosporium vitis) on vines; Macrophomina phaseolina (synonym Macrophomina ) on soybeans and cotton phaseoli (root and stem rot); Microdochium nivale (synonym Fusarium nivale) (pink snow mold) on cereals (e.g. wheat or barley); Microsphaera diffusa (powdery mildew) on soybeans; Monilinia species, such as M. laxa, M. fructicola and M. fructigena (synonyms from Monilia species: blossom end and branch dieback, brown rot) on stone fruits and other Rosaceae plants; Mycosphaerella species, such as M. graminicola (anamorph: Zymoseptoria tritici, formerly Septoria tritici) on cereals, bananas, soft fruits and peanuts tritici: Septoria leaf spot), or M.fijiensis (synonym Pseudocercospora fijiensis: black Sigatoka disease) and M.musicola on banana, M.arachidicola (synonym M.arachidis or Cercospora arachidis), M.berkeleyi on peanuts, M.pisi and M. brassiciola on Brassica; Peronospora species (downy mildew) on cabbage (e.g. P. brassicae), rapeseed (e.g. P. parasitica), onion (e.g. P. destructor), tobacco (P. tabacina), and soybean (e.g. P. manshurica); Phakopsora pachyrhizi and P. meibomiae (soybean rust) on soybean; Phialophora species, for example on vines (e.g. P. tracheiphila and P. tetraspora) and soybean (e.g. P. gregata: stem rot); Phoma nigromaculata on rapeseed and cabbage (e.g. P. gregata: stem rot); Phoma nigromaculata on rapeseed and cabbage (e.g. P. lingam (synonyms Leptosphaeria biglobosa and L. maculans in Cruciferae: root and stem rot) and P. betae on sugar beet (root rot, leaf spot and damping-off), and P. zeae-maydis (synonym Phyllostica zeae) on maize; Phomopsis species on sunflower, vines (e.g. P. viticola: vine and leaf spot) and soybean (e.g. stem rot: P. phaseoli, teleomorph: Diaporthe phaseolorum); Physoderma maydis on maize. maydis (brown spot); Phytophthora species (wilt, root, leaf, fruit and stem rot) on a variety of plants, such as red pepper and cucurbit (e.g., P. capsici), soybean (e.g., P. megasperma, synonym P. sojae), potato and tomato (e.g., P. infestans: late blight), and broadleaf trees (e.g., P. ramorum: sudden oak death); Plasmodiophora brassicae (club root) on cabbage, rapeseed, radish and other plants; Plasmopara species, such as P. viticola (grape downy mildew) on vines and P. hallii on sunflower.halstedii); Podosphaera species (powdery mildew) on rosaceae, hops, pome and soft fruits, for example P. leucotricha on apples and P. xanthii on cucurbits; Polymyxa species on cereals such as barley and wheat (P. graminis) and sugar beets (P. betae), and viral diseases transmitted thereby; Pseudocercosporella herpotrichoides on cereals such as wheat or barley (synonyms Oculimacula yallundae, O. acuformis: eyespot, teleomorph: Tapesia yallundae; Pseudoperonospora (downy mildew) on various plants, such as P.cubensis on cucurbits or P.humili on hops; Pseudopezicula tracheiphila (red fire disease or rotbrenner, asexual form: Phialophora); Puccinia species (rust diseases) on various plants, for example P. triticina (brown or leaf rust), P. strii-formis (stripe or yellow rust), P. hordei (dwarf rust), P. graminis (stem or black rust) or P. recondita (brown or leaf rust) on cereals such as wheat, barley or rye, P. kuehnii (citrus rust) on sugarcane and P. asparagi on asparagus; Pyrenopeziza species, for example P. brassicae on rapeseed; Pyrenophora spp. on wheat. tritici-repentis) (asexual form: Drechslera tritici-repentis) (brown spot) on wheatgrass or P. teres (net spot) on barley; Pyricularia species, such as P. oryzae (sexual form: Magnaporthe grisea: rice blast or leaf blast) on rice, and P.grisea) and Magnaporthe oryzae (ear neck blast) on rice; Pythium species (damping-off) (e.g. P. ultimum or P. aphani-dermatum) on turfgrass, rice, corn, wheat, cotton, rapeseed, sunflower, soybean, sugar beet, vegetables and a variety of other plants, and P. oligandrum on mushrooms; Ramularia species, e.g. R. Collo-cygni (R. collo-cygni leaf spot, physiological leaf spot) on barley, R. areola (teleological form: Mycosphaerella sphaeroides) on cotton areola) and R. beticola on sugar beet; Rhizoctonia species on cotton, rice, potato, turf, corn, rapeseed, potato, sugar beet, vegetables and a variety of other plants, such as R. solani (root and stem rot) on soybeans, R. solani (shear blight) on rice, or R. cerealis (rhizoctonia spring leaf blight) on wheat or barley; Rhizopus stolonifer (black mold, soft rot) on strawberries, carrots, cabbage, vines and tomatoes; Rhynchosporium secalis and R. commune (scald) on barley, rye and triticale; Sarocladium oryzae on rice. oryzae and S. attenuatum (sheath rot); Sclerotinia species on vegetables (S. minor and S. sclerotiorum) and field crops, such as rapeseed, sunflower (e.g. S. sclerotiorum) and soybean (stem rot or white mold), S. rolfsii (synonym Athelia rolfsii) on soybean, peanuts, vegetables, corn, cereals and ornamentals; Septoria species on various plants, such as S. glycines (brown spot) on soybean, S. tritici (synonym Zymoseptoria tritici, Septoria leaf spot) on wheat and S. glumes on cereals (S.nodorum (synonym Stagonospora nodorum) (Stagonospora leaf blight); Uncinula necator (synonym Erysiphe necator) (powdery mildew, anamorph: Oidium tuckeri) on vines; Setosphaeria species (leaf blight) on corn (e.g., S. turcicum, syn. Helminthosporium turcicum) and turfgrass; Sphacelotheca species (smut) on corn (e.g., S. reiliana, syn. Ustilago reiliana: head smut), sorghum, and sugarcane; Sphaerotheca fuliginea (syn. Podosphaera fuliginea) on cucurbits (syn. xanthii: powdery mildew); Spongospora subterranea on potato (powdery scab) and viral diseases transmitted thereby; Stagonospora species on cereals, e.g. S. nodorum on wheat (Stagonospora leaf spot, teleomorph: Leptosphaeria nodorum [syn. Phaeosphaeria nodorum], syn. Septoria nodorum); Synchytrium endobioticum on potato (potato cancer); Taphrina species, e.g. T. deformans on peach (leaf curl) and T. pruni on plum (plum fruit blight). pocket); Thielaviopsis species (black root rot), such as T. basicola (syn. Chalara elegans) on tobacco, pome, vegetables, soybean and cotton; Tilletia species (common bunt or stinking smut) on cereals, such as T. tritici (syn. T. caries, wheat bunt) and T.controversa (dwarf bunt); Trichoderma harzianum on mushrooms; Typhula incarnate (grey snow mold) on barley or wheat; Urocystis species, such as U. occulta (stem smut) on rye; smut); Uromyces species (rust) on vegetables such as beans (e.g., U. appendiculatus, synonym U. phaseoli), sugar beets (e.g., U. betae or U. beticola), and dry beans (e.g., U. vignae, U. pisi, U. viciae-fabae, and U. fabae); Ustilago species (loose smut) on cereals (e.g., U. nuda and U. avaenae), corn (e.g., U. maydis: corn smut), and sugar cane; Ustilaginoidea pyrifera on rice virens (false smut); Venturia species (scab) on apples (e.g. V. inaequalis) and pears; and Verticillium species (wilt) on a variety of plants, such as fruit and ornamental plants, vines, soft fruit, vegetables and field crops, such as V. longisporum on rapeseed, V. dahliae on strawberry, rapeseed, potato and tomato, and V. fungicola on mushrooms; Xanthomonas oryzae (bacterial leaf blight) on rice; Septoria tadella on cereals;

[0048] In a preferred embodiment of the present invention, the disease is a plant disease caused by:

[0049] Sheath blight (CORTSS), leaf blast (also known as rice blast; PYRIOR), ear neck blast (PYRPRO), bacterial leaf blight (XANTOR), and false smut (USTNVI). BRIEF DESCRIPTION OF THE DRAWINGS

[0050] These and other aspects of the invention will become apparent and be further elucidated with reference to the embodiments described by way of example in the following description and with reference to the accompanying drawings, in which:

[0051] Figure 1 A block diagram of an exemplary apparatus for determining disease progression for fungicide spray scheduling on agricultural fields is presented.

[0052] Figure 2 Diagram showing a disease model simulating primary and secondary infection.

[0053] Figure 3A and Figure 3B Shown are the distributions of observations for sheath blight (CORTSS), leaf blast (PYRIOR), panicle blast (PYRPRO), bacterial leaf blight (XANTOR), and false smut (USTNVI) after data cleaning and processing (top left).

[0054] Figure 4 An example of disease severity interpolation using exponential growth curves is shown.

[0055] Figure 5 The distribution of the number of observations for each disease in each country is shown.

[0056] Figure 6 A comparative analysis of disease severity observed in experimental and untreated sites is shown.

[0057] Figure 7 The original prediction and the smoothed original prediction are shown.

[0058] Figure 8 Shown is the distribution of the difference in days between the observed and predicted first infection dates for leaf blight (PYRIOR), ear blast (PYRPRO), and bacterial leaf blight (XANTOR) for all untreated locations in Japan.

[0059] Fig. 9 The distribution of true positives, false positives, true negatives, and false negatives is shown.

[0060] Fig.10 Shown are the geospatial distribution of true positives, false positives, false negatives, and true negatives for sheath blight occurrence predictions for the test locations in Japan.

[0061] Fig.11 Shown are the geospatial distribution of true positives, false positives, false negatives, and true negatives for leaf blight occurrence predictions for the trial locations in Japan.

[0062] Fig.12Shown are the geospatial distribution of true positives, false positives, false negatives, and true negatives for panicle blast occurrence predictions for the trial locations in Japan.

[0063] Fig.13 Shown are the geospatial distributions of true positives, false positives, false negatives, and true negatives for false smut occurrence predictions for the test locations in Japan.

[0064] Fig.14 Shown are the geospatial distribution of true positives, false positives, false negatives, and true negatives for bacterial leaf blight occurrence predictions for the test locations in Japan.

[0065] Fig.15 Accuracy, precision, and recall are shown.

[0066] Fig.16 The feature importance for each model is shown.

[0067] Fig.17 The geospatial distribution of predicted sheath blight occurrence is shown.

[0068] Fig.18 The geospatial distribution of leaf blight occurrence predictions is shown.

[0069] Fig.19 The predicted geospatial distribution of panicle blast occurrence is shown.

[0070] Fig. 20 The predicted geospatial distribution of false smut occurrence is shown.

[0071] Fig.21 The predicted geospatial distribution of bacterial leaf blight occurrence is shown.

[0072] Fig. 22 Shown are the geospatial distribution of the maximum severity predicted by models trained for sheath blight (CORTSS), leaf blast (PYRIOR), panicle blast (PYRPRO), bacterial leaf blight (XANTOR), and false smut (USTNVI) for all disease occurrence locations.

[0073] Fig.23 The effect of disease onset date on the disease progression curve is shown.

[0074] Fig.24 The effect of cultivar disease resistance level on the disease progression curve is shown.

[0075] Fig.25 The effect of fungicide application on the disease progression curve is shown.

[0076] Fig.26 A fungal disease management system is shown.

[0077] Fig. 27 A flow chart is presented illustrating a computer-implemented method for determining disease progression for fungicide spray scheduling on agricultural fields.

[0078] It should be noted that these drawings are schematic only and are not drawn to scale. In the drawings, elements corresponding to elements already described may have the same reference numerals. Examples, embodiments or optional features, whether or not indicated as non-limiting, should not be construed as limiting the invention as claimed. DETAILED DESCRIPTION

[0079] Figure 1 1 is a block diagram of an exemplary apparatus 10 for determining disease progression for fungicide spraying scheduling on agricultural fields. The exemplary apparatus 10 includes an input unit 12, a primary infection determination unit 14, and an output unit 18. Optionally, as Figure 1 As shown, the apparatus 10 may further include a secondary infection determination unit 16 and a fungicide spraying schedule determination unit 20 .

[0080] In general, the apparatus 10 may include various physical and / or logical components for communicating and manipulating information, which may be implemented as hardware components (e.g., computing devices, processors, logic devices), executable computer program instructions (e.g., firmware, software) to be executed by various hardware components, or any combination thereof, as desired for a given set of design parameters or performance constraints. Figure 1 A limited number of components may be shown as examples, but it is understood that a greater or lesser number of components may be employed for a given implementation. In addition, the functionality provided by one or more components of the device 10 may be combined or separated. In addition, the functionality of any one or more components of the device 10 may be implemented by any suitable computing environment, such as a personal computing environment, a time-sharing computing environment, a distributed computing environment, a cloud computing environment, and a cluster computing environment. Fig.26 An exemplary distributed computing environment is shown in FIG.

[0081] In some embodiments, the device 10 can be implemented as a device or apparatus or in a device or apparatus, such as a server, a workstation, or a mobile device. The device 10 may include one or more microprocessors or computer processors that execute appropriate software. The primary infection determination unit 14, the secondary infection determination unit 16, and the fungicide spraying schedule determination unit 20 of the device 10 may be implemented by one or more of these processors. The software may have been downloaded and / or stored in a corresponding memory, for example, a volatile memory such as RAM or a non-volatile memory such as flash memory. The software may include instructions that configure one or more processors to perform the functions described herein.

[0082] Note that the apparatus 10 may be implemented with or without a processor, and may also be implemented as a combination of dedicated hardware that performs certain functions and a processor (e.g., one or more programmed microprocessors and associated circuitry) that performs other functions. For example, the functional units of the apparatus 10 (e.g., the input unit 12, the primary infection determination unit 14, the secondary infection determination unit 16, the fungicide spraying schedule determination unit 20, and the output unit 18) may be implemented in a device or apparatus in the form of programmable logic (e.g., as a field programmable gate array (FPGA)). In general, each functional unit of the apparatus may be implemented in the form of a circuit.

[0083] In some embodiments, the device 10 may also be implemented in a distributed manner. For example, some or all units of the device 10 may be arranged as separate modules in a distributed architecture and connected in a suitable communication network, such as a third generation partnership project (3GPP) network, a long term evolution (LTE) network, the Internet, a LAN (local area network), a wireless LAN (local area network), a WAN (wide area network), etc.

[0084] Below, by way of example, we describe a hybrid model for simulating crop disease epidemics and the effects of weather, varieties, and fungicides on disease progression. The model can be used to optimize the timing of fungicide applications. In the hybrid model, a machine learning (ML) model is trained to simulate primary infection—that is, the onset of the disease. A process-based model was developed to simulate secondary infection through a process that includes susceptibility, exposure, infection, and recovery. The infection rate takes into account environmental factors including temperature and humidity. Variety susceptibility, fungicides, and crop growth stage are also considered in the infection rate calculation. We found that for 75% of the locations, the ML model can simulate the disease onset date with an accuracy of less than 5 days. The ML model can simulate disease occurrence with an accuracy of up to 85%. Process-based models can also be used to simulate the interaction between disease onset date, variety susceptibility, and fungicide application.

[0085] 1. Overall model architecture

[0086] Multi-cycle disease progression can be described by the following logistic equation:

[0087]

[0088] where y is the disease severity (which ranges from 0 to 1), and r is the rate parameter (infection rate).

[0089] Integrating the logistic equation yields the following terms:

[0090]

[0091] In equation 2, y 0 Describes the level of disease at the onset of illness and is closely related to the life cycle of the pathogen.

[0092] Many pathogens (for example, leaf blight) can produce spores associated with sexual reproduction and asexual reproduction. Sexual spores usually span unfavorable phase and / or non-host phase, and are the reason for the first infection (primary infection) of the season. During the growing season of crops, the asexual spores produced on the primary lesion can produce repeated (secondary) infection. These two kinds of spores usually have different environmental conditions requirements and different epidemiological characteristics. The occurrence and progress of the epidemic caused by these fungi are due to the coexistence of two types of infection cycles (primary cycle and secondary cycle) during the growth season of host plants.

[0093] The progression of these epidemics can be written as follows:

[0094]

[0095] Among them, y′ 0 = daily increase in disease intensity due to primary inoculum, t 0 = time to first seasonal illness onset, t p = time when the primary inoculum is no longer present, p = incubation period of the first infection cycle, y′ t = the daily increase in disease intensity due to the secondary inoculum, and t e =The time when the epidemic ends.

[0096] Other pathogens (such as the primary spores of Rhizoctonia solani of sheath blight) usually originate from viable forms such as stromata, sclerotia or hyphae. The progression of epidemics caused by these pathogens can be written as follows:

[0097]

[0098] Due to the complexity and uncertainty of primary inocula for airborne diseases such as leaf blight, there are currently no reliable mechanistic models to simulate disease onset and initial severity.

[0099] To this end, an apparatus and computer-implemented method are provided that use a trained machine learning model to simulate a disease onset date or a whole season without infection. Optionally, after predicting the onset date, the apparatus and computer-implemented method can use the onset date as a starting point and use a mechanistic model to simulate secondary infection and disease cycles.

[0100] For example, Figure 2A diagram of a disease model simulating primary and secondary infections is shown. State variables (shown as rectangular boxes) are linked by ratio variables. Primary infections can be simulated by machine learning-based models, for example, via Figure 1 The primary infection determination unit 14 of the device 10 shown in FIG. Secondary infection can be simulated by a process-based model, for example, by Figure 1 The secondary infection determination unit 16 of the device 10 shown in FIG.

[0101] 2. Methods

[0102] 2.1 Experimental data and newly collected data from public sources

[0103] The data can be derived from public sources. For example, the fields shown in the public data are geographically or historically high-risk fields in Japan. These fields have high disease pressure almost every year.

[0104] The public data for untreated locations in Japan consists of information on the transplant date of the crop, the rooting growth stage of the crop, the evaluation or observation, the disease evaluated, the observed parameters (such as growth stage, infection percentage, first symptom observation date, number of lesions, etc.), latitude and longitude, number of hills evaluated, and the evaluation score (i.e., growth stage value or disease infection). The data also contains other metadata such as field name, field address, province, number of replicates, and crop variety.

[0105] There are 96 locations where disease data is captured. The date of the historical disease data collected is from 2004 to 2019. Data is available for all rice diseases, namely, false smut, sheath blight, leaf blast, panicle blast, and bacterial leaf blight, except for brown spot. The data consists of 21 columns and 4304 rows. Of the 4304 rows of observations, 243 rows are about GS observations.

[0106] 2.2 Preparation of historical untreated data

[0107] The public dataset of untreated plots needed to be cleaned and transformed for retraining and re-evaluating the model performance for predicting disease severity. For data cleaning, we removed rows with incorrect “Evaluation Date” date values ​​from the data. These values ​​contained values ​​like “early July”. We also cleaned the “Evaluation Date” column by converting the Chinese values ​​to their numerical equivalents. The values ​​that were cleaned were the numerical equivalents of “middle ten days of a month”, “first ten days of a month”, and “last ten days of a month”. We removed rows that did not have a transplant date. Next, we also removed rows with incorrect evaluation dates for the transplant year. We compared the evaluation year with the transplant year and removed rows where the values ​​did not match. There were 76 cases where the evaluation year did not match the transplant year. Later, we corrected the rows where the evaluation score contained non-numeric values ​​such as “1 leaf stage”.

[0108] After data cleaning, we converted all disease information into disease severity. Since each disease type had many target values, we considered multiple values ​​for 'Assessment Item' in different diseases. For PYRIOR (Leaf Blight), we used 'Severity (Total Hills)', and if this value was missing, we used 'Number of Lesions'. For PYRPRO (Neck Blight), we considered 'Severity (Total Hills)'. For USTNVI (False Smut), we used 'Number of Infected Grains', and if this value was missing, we considered 'Number of Infected Ears'. For CORTSS (Sheath Blight), we considered 'Severity (Total Hills)', and for XANTOR (Bacterial Leaf Blight), we considered 'Severity (Total Hills)'. We used any value of the 'Assessment Item' column, since some locations had only one value for one assessment item.

[0109] To perform data transformation, we aggregated all replicates of disease data for each location (i.e. latitude and longitude) and each season. After this aggregation, we ended up with 450 rows of disease severity for 85 unique locations.

[0110] exist Figure 3A The distribution of the estimated counts for all 450 estimates for each disease can be seen in (left). The observed counts for leaf blight are higher than the observed counts for the remaining diseases.

[0111] Figure 3A and Figure 3B The distribution of observations for sheath blight (CORTSS), leaf blight (PYRIOR), ear blast (PYRPRO), bacterial leaf blight (XANTOR), and false smut (USTNVI) after data cleaning and processing is shown (top left). The position count distribution of the first symptom observation date or disease onset date in the cleaned and transformed untreated disease data for Japan (see Figure 3AThe position count distribution of the total number of observations at each position (see Figure 3B ). For example, locations where bacterial leaf blight was observed at least three times across all seasons had a count of only 3. Similarly, locations where leaf blast was observed a total of 6 times had a count of 2.

[0112] The ultimate goal of our data preprocessing is to convert discrete disease estimates into continuous disease severity values ​​for each day. To do this, we need the starting disease date of the disease as well as the starting severity. After estimating the starting disease date information, we will fit an exponential growth curve from the disease start date to the last evaluation day. Unfortunately, we do not have the starting disease date for all locations. This adds complexity to the process of converting discrete disease severity into a daily continuous form. Figure 2 (Bottom) shows the number of locations for each disease and its overall location count. For example, from this inset we can notice that locations where bacterial leaf blight was observed at least three times across all seasons had a count of only 1. Similarly, leaf blast was found in only 2 locations with a minimum of 4 observations.

[0113] exist Figure 3A (Right) The differences in locations with a first symptom observation date, or disease start date, are shown. We found leaf blight at 79 locations, panicle blast at 9 locations, and bacterial leaf blight at 8 locations. For the 2020 trial data, we ideally use 5 to 10 disease observations with a maximum interval of 14 days. Currently, we only considered those locations with a start disease date, and we ended up with only 57 observations out of a total of 278 cleaned and transformed observations. For those observed disease severity, we fit an "exponential growth function."

[0114] We designed an algorithm to convert discrete disease severity to its continuous form. First, we considered all disease estimation points and their index starting from the starting disease date. Next, the disease severity values ​​were normalized to preserve the upward trend of the disease data. For example, if for six observations, our disease values ​​were [1, 19, 25, 6, 32], then we changed these values ​​to [1, 19, 25, 25, 32]. The smoothing of the downward curve preserves the property of exponential growth. We then fit the disease severity values ​​to the following equation:

[0115] y=a*e -b*x +c (Equation 5)

[0116] By applying the above algorithm, we were able to convert discrete disease severity into a daily continuous disease curve.

[0117] Figure 4An example of disease severity interpolation using an exponential growth curve is shown. The black dots represent the observed disease severity, the vertical line represents the observed disease onset date, and the curve is interpolated based on the algorithm.

[0118] 3. Initial infection determination unit

[0119] The input unit 12 is configured to receive data including crop variety data, environmental data, crop management data, and location data of a farmland.

[0120] Crop variety data is related to crops growing or to be grown on farmland. Exemplary crop variety data may include, but are not limited to, growth stage at a particular point in time, crop density (i.e., the number of crops present in a field per unit area), and days after planting. Crop variety data may be obtained from field data for farmland.

[0121] The environmental data indicates environmental conditions of the farmland. Exemplary environmental data may include, but are not limited to, air temperature, cloud cover, dew point, shortwave radiation, longwave radiation, ice accumulation period, liquid accumulation period, relative humidity, adjusted precipitation accumulation period, snow accumulation period, wind speed. In some examples, the environmental data may be collected by sensors deployed on the farmland. In some examples, the environmental data may be received from a weather forecast service.

[0122] The crop management data indicates a fungicide spraying history of the farmland. The crop management data may be obtained from a data management system storing a fungicide spraying history of the farmland.

[0123] The location data of the farmland may include latitude and longitude data of the farmland (eg, decimal degrees, with negative values ​​for south or west), which may be obtained from field data of the farmland.

[0124] The primary infection determination unit 14 is configured to apply a machine learning model to the received data to determine disease progression time series data of a fungal disease. The machine learning model has been trained to learn disease progression under conditions defined by crop variety data, environmental data, and location data based on historical data collected from one or more farm fields.

[0125] The machine learning model can be any suitable model capable of determining disease progression time series data for fungal diseases. Examples of machine learning models can include, but are not limited to, XGBoost, artificial neural networks, recurrent neural networks, and support vector regression.

[0126] The first infection determination unit 14 may be further configured to determine the disease onset date of the fungal disease based on the determined disease progression time series data. For example, the disease onset date of the fungal disease may be determined using a change point detection algorithm.

[0127] Machine learning modeling is explained in detail below.

[0128] 3.1 Feature Engineering

[0129] Historical interpolated untreated disease data from Japan were merged with data from Xarvio trials in India, Japan, and China. Figure 5 represents the distribution of the number of observations for each disease in each country. We can notice that the distribution of the number of observations for each disease in each country is uniform. This is because if no infection is observed on a given observation date, the severity of all diseases is recorded as zero.

[0130] Figure 6 A comparative analysis of disease severity observed in trial locations and untreated locations is shown. We observe that the distribution of disease severity for all diseases is higher in the historical untreated field public data compared to disease severity in trial locations. The public untreated data has disease severity for all diseases, however the distribution captured is not. Figure 6 (Right) Only for leaf blast, panicle blast, and bacterial leaf blight. These diseases were only considered for modeling because their observations had disease start dates.

[0131] The data for each disease were preprocessed and feature engineered. No scaling was done on the numerical data because we had planned to train the machine learning model using a nonlinear tree-based algorithm. Tree-based models are not sensitive to scale. The features used for training, testing, and validating the models can be divided into three subtypes and they are represented in the table. 1. The growth stage data were simulated based on the growth stage model developed in the Growth Stage Simulation Model for Rice. The weather data for each location was collected from ITERIS (ClearAG, nd). The target of the model is disease severity / infection percentage.

[0132]

[0133] Table 1

[0134] The latitude and longitude data were divided by 10 and the upper limit was taken for modeling. This avoids the possibility of the model being too sensitive or biased to these two features. Of the 500 locations in the dataset, we separated 40 fields for validation of each model. The remaining 460 fields were split into 80% training and 20% testing.

[0135] 3.2 Machine Learning Model Training

[0136] We trained separate models for each disease, rather than training one model for all diseases combined. This allows each individual model to gain a deeper understanding of the abstract relationships between features.

[0137] Any suitable machine learning model capable of determining the disease progression time series data of fungal diseases may be employed. Examples of machine learning models may include, but are not limited to, extreme gradient boosting (XGB) regression models, artificial neural networks, and support vector regression. An exemplary machine learning model, i.e., the XGB regression model, will be described in detail below.

[0138] The model chosen is the XGB regression model tuned via random search cross validation. XGB is a scalable tree boosting system and can be viewed as gradient boosting with regularization. It implements parallel tree construction, cache-aware access, sparsity awareness, and weighted quantile sketches as part of its system optimization and algorithm enhancements.

[0139] The hyperparameter ranges used to tune XGB are explained in Table 2. The most important hyperparameters are the maximum depth and the number of estimators. Tuning the hyperparameters gives the model the ability to search for the best fit. Each hyperparameter affects the model in some way. By tuning the learning rate, we prevent overfitting. Also, the lower the learning rate, the more robust the model will be against overfitting. The maximum depth refers to the depth of the trees or estimators within the XGB model. It sets the maximum number of nodes that can exist between the root and the furthest leaf. The lower the maximum depth, the less overfitting there is. The minimum child weight performs regularization at the split step. It is the minimum Hessian weight required to create a new node. The Hessian is the second-order derivative of the XGB model equation. The maximum incremental step size sets the maximum absolute value that the weight can have. It is very useful when dealing with unbalanced classes or data. The number of estimators represents the total number of trees built for the entire XGB model. Generally, the fewer the number of estimators, the better the overall fit. However, this may not be true for data with high bias.

[0140]

[0141]

[0142] Table 2

[0143] 3.3 Change detection of disease onset

[0144] To calculate the disease onset date, we cleaned the original disease time curve predicted by the model and ran a change point detection algorithm to obtain the onset index of the curve.

[0145] To clean up the predicted disease curve, we first stabilize the noise from the model by converting any prediction below 0.3 to 0. We then decompose the curve using a seasonal decomposition process. The result is obtained by first estimating the trend by applying a convolution filter to the data. The trend is then saturated out of the curve and passed on. This reduces the Gaussian noise from the original prediction. The decomposed curve is smoothed using convolution-based smoothing using the Python wrapper tsmoothie. The smoothed curve is then passed to a slope-based onset detection algorithm. In this algorithm, we apply a 7-day rolling slope. The first point on the rolling slope curve with a value greater than or equal to 0.1 is selected as the onset. The following figure represents disease onset detection in action.

[0146] Figure 7 The original predictions are shown, which are smoothed and cleaned into curves. The black dashed line is the actual severity. The vertical dashed line is the generated disease onset date. The vertical dashed line represents the predicted disease onset date, and the black vertical dashed line represents the observed onset of disease severity.

[0147] 3.4 Machine Learning Model Validation

[0148] Based on observed disease severity at 2020 Xarvio trial locations, we calculated and identified locations in a given crop season where correct and incorrect predictions were made. Disease or no disease scenarios were calculated based on the presence of disease onsets. Disease curves where no disease onsets were observed were labeled as no disease locations and vice versa. Crop seasons were labeled as true positives (i.e., locations where disease was observed and the model also predicted disease), false positives (i.e., locations where disease was not observed but the model predicted disease), false negatives (i.e., locations where disease occurred but the model did not predict disease), and true negatives (i.e., locations where disease was not observed and the model did not predict any disease).

[0149] After the machine learning model training was completed, we used the untreated field data from Japan to determine the model effectiveness. In addition to considering the mean absolute error (MAE) and root mean square error (RMSE) of the test data for model validation, we also considered the difference in days between the start disease date in the disease curve predicted by the model and the start disease date in the observed disease severity curve for each season, each disease, and each location. We generated the difference in days between the start disease date in the disease curve predicted by the model and the start disease date in the observed disease severity curve for each season, each disease, and each location, starting with the prediction, then cleaning and smoothing the predicted disease curve, and finally generating change points, i.e., disease start points / dates. Later, we compared the results for all untreated fields. Of the 74 disease severity time series of untreated fields, 20 fields were from the validation data (i.e., they were not used for training the model), and the remaining 54 disease severity time series were used for training.

[0150] Based on the prediction data of the trial and untreated locations, we calculated the overall accuracy, precision, and recall for all diseases according to the formulas in the following equations.

[0151]

[0152] We also simulated all rice growing areas in Japan in 2019. There are a total of 2721 locations across Japan. We predicted the disease severity for all locations, and we later cleaned and smoothed each predicted disease severity curve to pass it to the disease onset detection algorithm to detect the onset of the disease. After we predicted the disease onset date, we calculated the growth stage and post-planting days for the disease onset day.

[0153] 4. Secondary infection determination unit

[0154] The optional secondary infection determination unit 16 is configured to apply a process-based model to determine the infection rate of the fungal disease after the disease onset date under conditions defined by the crop variety data, environmental data, crop management data and location data.

[0155] In some examples, the infection rate of a fungal disease can be determined by further including conditions defined by the cultivar disease resistance level of the crop.

[0156] In some examples, the infection rate of the fungal disease is determined by further including conditions defined by fungicide application data including fungicide data for a fungicide product to be used and at least one planned application timing.

[0157] The process-based disease model is explained in detail below.

[0158] 4.1 Structure of process-based disease models

[0159] Process-based disease models can include any suitable model that can predict how infectious diseases progress, to show the possible outcomes of epidemics in plants, and to help inform plant health interventions. An example of a process-based disease model is a compartment model expressed as a Markov chain. The classic compartment model in epidemiology is the SIR model, which can be used as a simple model for epidemic modeling. A variety of other types of compartment models are also used. An exemplary process-based disease model, the SEIR model, will be explained in detail below. However, it will be appreciated that any model capable of modeling the prevalence of infectious diseases of plants can be implemented as a process-based disease model.

[0160] The structure is a SEIR model (Susceptible-Exposed-Infected-Recovered), which has been widely used to simulate epidemics of infectious diseases in plants. The process-based model structure is based on the concepts of the following reference: Van der Plank, JE, 2013. Plant diseases: epidemics and control. Elsevier., using the system representation of the following reference: González-Domínguez, E., Fedele, G., Salinari, F., Rossi, V., 2020. A general model of the effects of crop management on plant disease epidemics at different levels of complexity. [Agronomy] 10., translated into plant epidemiology by the following reference: Zadoks, J., 1971. Systematic analysis and dynamics of epidemics. [Plant Pathology] 61, 600-610.

[0161] The system considered is 1m 2 The rice crop population is simulated according to the growth stage model results to simulate the epidemic during the crop growth period. It involves four state variables of the crop population: health (H), latent (L), infection (I) and post-infection site (P).

[0162] The core element of this model is the daily infection rate y′ t , which is written as:

[0163] y′ t =RC i *INF*COFR Agg (Equation 9)

[0164] Among them, y′ t is the daily infection rate, RC i is the favorable degree of the environment, INF is the number of infectious sites, and COFR is the correction factor for disease sites.

[0165] RC i =RCOpt*RCGS*RCT*RCW*RCV (Equation 10)

[0166] Among them, RCOpt is the infection rate under favorable conditions, RCGS is the correction factor for crop growth stage, RCT is the correction factor for temperature, RCW is the correction factor for leaf wetness, and RCV is the correction factor for variety resistance.

[0167]

[0168] Where RRG is the relative rate of tissue growth, HTA is i is the healthy tissue on day i, TA max is the maximum tissue area.

[0169] HTA i =∑RTG i -DST i -Sene i (Equation 12)

[0170] Among them, DST i is the disease tissue on day i, Sene i is the senescence on day i.

[0171] 5. Results

[0172] 5.1 Initial infection determination unit

[0173] 5.1.1 Machine Learning for Predicting Disease Onset Date

[0174] exist Figure 8 Figure 2 shows the distribution of the difference in days between the predicted and observed start dates for all 74 disease severity time series, including the distribution of the difference in days between the observed and predicted first infection dates for leaf blight (PYRIOR), panicle blast (PYRPRO), and bacterial leaf blight (XANTOR) at all untreated locations in Japan.

[0175] 5.1.2 Model Accuracy of Simulating Disease Occurrence

[0176] Fig. 9 The distribution of true positives, false positives, true negatives and false negatives is shown in , which shows the distribution of true positives, false positives, true negatives and false negatives for sheath blight (CORTSS), leaf blast (PYRIOR), ear neck blast (PYRPRO), bacterial leaf blight (XANTOR) and false smut (USTNVI) at all test locations in Japan.

[0177] The geospatial representation of true positives, false positives, true negatives, and false negatives for each disease can be found in Figures 10 to 14 The left side of the figure confirms where we made correct predictions by providing labels of true positives and true negatives to the geospatial locations, and the right side panel indicates where incorrect predictions were made, i.e. where we predicted false positives and false negatives.

[0178] In particular, Fig.10 The geospatial distribution of true positives, false positives, false negatives, and true negatives for sheath blight occurrence predictions for the trial locations in Japan is shown. Crosses represent locations used for validation, while circles represent locations used during training.

[0179] Fig.11 The geospatial distribution of true positives, false positives, false negatives, and true negatives for leaf blight occurrence predictions for the trial locations in Japan is shown. The crosses represent locations used for validation, while the circles represent locations used during training.

[0180] Fig.12 The geospatial distribution of true positives, false positives, false negatives, and true negatives for spikelet occurrence predictions for the test locations in Japan is shown. The crosses represent locations used for validation, while the circles represent locations used during training.

[0181] Fig.13 The geospatial distribution of true positives, false positives, false negatives, and true negatives for false smut occurrence predictions for the test locations in Japan is shown. The crosses represent the locations used for validation, while the circles represent the locations used during training.

[0182] Fig.14 The geospatial distribution of true positives, false positives, false negatives, and true negatives for bacterial leaf blight occurrence predictions for the test locations in Japan is shown. The crosses represent locations used for validation, while the circles represent locations used during training.

[0183] exist Fig.15 The accuracy, precision, and recall are presented in Figure 2 for sheath blight (CORTSS), leaf blight (PYRIOR), ear blast (PYRPRO), bacterial leaf blight (XANTOR), and false smut (USTNVI) for all Xarvio test locations in Japan. Disease or no disease was calculated based on the disease start point in the predicted disease curve for each location. Values ​​were calculated for all test locations in Japan. A total of 140 locations were considered for calculating performance, 80 for training, 20 for testing the model, and 40 for validation.

[0184] 5.1.3 Importance of Features

[0185] We calculated the importance of each feature used to train each model. The features used to train the models are listed in Table 1. Two plant-related features, twelve weather-related features, and two location-related features were used to train each model.

[0186] exist Fig.16 The feature importance of each model is represented in the heat map in , which shows the feature importance of trained machine learning models for sheath blight (CORTSS), leaf blight (PYRIOR), ear blast (PYRPRO), bacterial leaf blight (XANTOR), and false smut (USTNVI). Each column indicates the trained model for the disease, and the row represents each feature used to train the model. The number in each cell indicates the importance ratio of the feature in the model. The color also indicates the relative importance. The number in each cell indicates their importance ratio in the model.

[0187] Each XGB model gives importance to the number of days after planting, location indicators, growth stage and weather-related features in different orders. The number in each cell represents the importance ratio of the feature in the model. The color also indicates the relative importance. For the model trained for sheath blight, the relatively important features may be shortwave radiation and cloud cover. For leaf blight, the number of days after planting and dew point seem to have significant predictive power for disease severity. For ear neck blast, wind speed and cloud cover are the most important features. However, the model of false smut gives higher importance to shortwave radiation and longitude. This phenomenon of giving higher importance to location indicators (i.e., longitude) is the result of the disease severity used to train the model. Since the fields in Japan were properly treated with fungicides, the overall disease infection or disease severity observed was less than that in the fields in China and India. The model compensates for this distribution of disease severity by giving more priority to location indicators. Since the model is likely to be more sensitive to location information (i.e., latitude and longitude), we reduce bias by dividing the latitude and longitude values ​​by 10 and taking the upper limit. For example, if the latitude value is 79.884, we convert the value to 8 instead of considering the actual floating point value. This reduces the degree of sensitivity that the model may assign to location information. For bacterial leaf blight, the most important features are cloud cover, wind speed, and adjusted precipitation accumulation period. Important weather features related to overall leaf wetness are dew point, air temperature, and wind speed.

[0188] 5.1.4 Simulation results of all growth regions

[0189] exist Figures 17 to 21 The GS or growth status of the disease onset date and the number of days after planting are geographically represented in the .

[0190] Fig.17The geospatial distribution of the prediction of sheath blight occurrence is shown. The left figure of the figure represents the mapping of the attack GS (i.e., the growth stage recorded on the disease start date), and the left figure of the inset represents the number of days after planting (i.e., the difference in days between the crop rooting date and the disease start date). The intensity of the attack GS value and the number of days after planting are displayed by a color map. The inset histogram represents the distribution of each variable considered. The gray position represents the disease-free situation, and the position with a forest green cross represents the position where the disease is predicted to occur after harvest.

[0191] Fig.18 The geospatial distribution of leaf blight occurrence predictions is shown. The left panel of the figure represents the mapping of the attack GS (i.e., the growth stage recorded on the disease start date), and the left panel of the inset represents the number of days after planting (i.e., the difference in days between the crop rooting date and the disease start date). The intensity of the attack GS values ​​and the number of days after planting are shown by color maps. The inset histograms represent the distribution of each variable considered. Gray locations represent disease-free conditions, and locations with forest green crosses represent locations where disease is predicted to occur after harvest.

[0192] Fig.19 The geospatial distribution of the prediction of the occurrence of ear blast is shown. The left figure of the figure represents the mapping of the attack GS (i.e., the growth stage recorded on the disease start date), and the left figure of the inset represents the attack days after planting (i.e., the difference in days between the crop rooting date and the disease start date). The intensity of the attack GS value and the attack days after planting are shown by a color map. The inset histogram represents the distribution of each variable considered. The gray position represents the disease-free situation, and the position with forest green cross represents the position where the disease is predicted to occur after harvest.

[0193] Fig. 20 The geospatial distribution of the false smut occurrence prediction is shown. The left panel of the figure shows the mapping of the attack GS (i.e., the growth stage recorded on the disease onset date), and the left panel of the inset shows the attack days after planting (i.e., the difference in days between the crop rooting date and the disease onset date). The intensity of the attack GS values ​​and the attack days after planting are shown by color maps. The inset histograms show the distribution of each variable considered. Gray locations represent disease-free conditions, and locations with forest green crosses represent locations where the disease is predicted to occur after harvest.

[0194] Fig.21The geospatial distribution of the predicted occurrence of bacterial leaf blight is shown. The left panel of the figure shows the mapping of the attack GS (i.e., the growth stage recorded on the disease onset date), and the left panel of the inset shows the number of days after planting (i.e., the difference in days between the crop rooting date and the disease onset date). The intensity of the attack GS values ​​and the number of days after planting are shown by color maps. The inset histograms show the distribution of each variable considered. Gray locations represent disease-free conditions, and locations with forest green crosses represent locations where disease is predicted to occur after harvest.

[0195] exist Fig. 22 The maximum predicted disease severity for all rice growing locations in Japan is represented geographically, showing the geospatial distribution of the maximum severity predicted by models trained for sheath blight (CORTSS), leaf blast (PYRIOR), panicle blast (PYRPRO), bacterial leaf blight (XANTOR), and false smut (USTNVI) for all disease occurring locations.

[0196] For sheath blight, Fig.17 It can be observed that the onset growth stage (i.e. the growth stage recorded on the disease onset day) is usually close to 21. It can also be observed that the disease is usually predicted in central Japan. For leaf blight, the geospatial representation of the onset growth stage and the number of days of onset after planting is represented by 18. It can be observed that it is predicted in almost all parts of Japan and that it is usually predicted to onset in the early growth stage. For ear blast, Fig.19 It can be observed that the onset growth stage is usually close to growth stage 70 to 80. It is predicted to occur later in the crop cycle. False smut is predicted to be more common in southern Japan, and the onset of the disease is also predicted to be later in the crop cycle, i.e., at growth stage 70 to 80, as with panicle blast. However, we predict that there are no locations infected with bacterial leaf blight in rice growing locations in Japan. Fig. 22 The geospatial representation of the maximum disease severity indicated that the majority of disease records had a predicted maximum severity in the range of 20 to 40.

[0197] 5.2 Secondary infection determination module

[0198] 5.2.1 Process-based model for simulating secondary disease infection

[0199] Fig.23 The effect of disease onset date on the disease progression curve is shown.

[0200] Fig.24 The effect of the variety's disease resistance level on the disease progression curve is shown. Here, the resistance level is ranked between 1 and 9. 1 means susceptible and 9 means resistant.

[0201] Fig.25The effect of fungicide application on the disease progression curve is shown. cur indicates curative efficacy, cur_pd indicates curative protection days, era indicates eradication efficacy, and era_pd indicates eradication protection days.

[0202] Predicting disease onset and disease progression curves to characterize the progression of the disease over time are critical to understanding how plant diseases develop and how disease control measures should be applied to achieve the highest efficacy of fungicides. As discussed above, we proposed an ML-based model to simulate disease onset, and a process-based model was used to simulate the disease progression curve and the effects of weather conditions, growth stage, variety, and fungicide spraying. ML achieved high accuracy in simulating disease onset dates, and the process-based model could reflect the disease response to the most important factors.

[0203] Go to Figure 1 As shown, the determined disease progression is provided via output unit 18. The determined disease progression may preferably be used for fungicide spraying scheduling on the agricultural field.

[0204] Alternatively, if Figure 1 As shown, the apparatus 10 may further include a fungicide spraying schedule determination module 20, which is configured to determine a fungicide spraying schedule based on the determined disease progression. The fungicide spraying schedule may be based on the fungicide application versus disease progression curve (e.g. Fig.25 The invention relates to a method for spraying a fungicide, wherein the spraying time is set based on the predicted disease risk, thereby improving the treatment. In other words, the spraying time is based on the predicted disease risk, thereby allowing the fungicide to be applied when the fungicide is most effective during the growing season. In particular, the determined fungicide spraying schedule can better align the optimal application period to prevent disease progression while reducing the chance of missing the risk period. Therefore, the optimized spraying time can have a considerable impact on the overall reduction of fungicide use.

[0205] The fungicide spray schedule may include application timing data. Based on the fungicide spray schedule, a configuration file may be generated. The configuration file may be used to configure a sprayer to apply a fungicide spray to the farmland according to the application timing data.

[0206] Fig.26 A fungal disease management system 100 is shown, which may be a cloud environment. As shown, the fungal disease management system 100 includes one or more data sources 110, a data analysis server 120, an electronic communication device 130, a network 140, and a sprayer 150. In this example, Figure 1 The illustrated apparatus 10 is implemented in a data analysis server 120 , for example, resident as software in the data analysis server 120 .

[0207] The data source 110 of the illustrated example may include a database, an application, a local file, or any combination thereof. The data source 110 may include data obtained from one or more sources. For example, the data source 110 may include crop variety data obtained from field data of a farmland, environmental data obtained from sensors deployed in the field and / or from a weather forecast service, crop management data obtained from a data management system, and location data from field data of a farmland.

[0208] The data analysis server 120 of the illustrated example may be a server that provides web services to facilitate data management. The data analysis server 120 may include a data extraction module (not shown) configured to identify data to be extracted from the data source 110, retrieve data from the data source, and provide the retrieved data to the device 10, which processes the extracted data according to the method described herein. The processed data is then provided to a data analysis application resident in the electronic communication device 130 so that the user can display and manipulate the data. In some examples, the device 10 may provide a fungicide spraying schedule that may be provided to the electronic communication device 130 to allow the farmer to configure the sprayer 150 according to the fungicide spraying schedule. In some examples, the device 10 may provide a configuration profile that may be loaded into the sprayer 150 to configure the sprayer 150 to apply the fungicide according to the determined spraying timing.

[0209] The electronic communication device 130 of the illustrated example may be a desktop, a notebook computer, a laptop computer, a mobile phone, a smart phone, and / or a PDA. The electronic communication device 130 may include a data analysis application, which may be a software application that enables a user to manipulate the data extracted from the data source 110 by the data analysis server 120, and select and specify actions to be performed on each data. For example, the data analysis application may be a desktop application, a mobile application, or a web-based application. The data analysis application may include a user interface, such as an interactive interface, including but not limited to a GUI, a character user interface, and a touch screen interface. Via the software application, a user may access the data analysis server 120 to obtain information, such as disease onset data, infection rates of fungal diseases after disease onset days, fungicide spraying arrangements, and / or configuration files that can be used to configure the sprayer 150.

[0210] Sprayer 150 may be, for example, a ground robot with a variable rate applicator, an aerial sprayer, or other variable rate applicator for applying fungicides to agricultural fields. Fig.26In the example of , the sprayer 150 may be a smart agricultural machine. The smart agricultural machine may be a smart sprayer and include a connection system 152. The connection system 152 may be configured to communicatively couple the smart agricultural machine 150 to a computing environment.

[0211] The network 140 of the illustrated example can communicatively couple the data source 110, the data analysis server 120, the electronic communication device 130, and the sprayer 150. In some examples, the network 140 can be the Internet. Alternatively, the network 140 can be any other type and number of networks. For example, the network 140 can be implemented by a number of local area networks connected to a wide area network. For example, the data source 110 can be associated with a first local area network, the data analysis server 120 can be associated with a second local area network, and the electronic communication device 130 can be associated with a third local area network. The first local area network, the second local area network, and the third local area network can be connected to the wide area network. Of course, the network 140 can be implemented using any other configuration and topology, including any combination of wired networks, wireless networks, wide area networks, local area networks, etc.

[0212] Fig. 27 A flow chart is shown showing a computer-implemented method 200 for determining disease progression for fungicide spray scheduling on agricultural fields. The method may be understood to emphasize the operation of the apparatus 10 described above. However, it will also be understood that Fig. 27 The method steps explained in Figure 1 The described architecture of the apparatus 10 is related to the following. More particularly, the method described below may be understood as a teaching in itself.

[0213] At block 210, i.e. step a), for example, Figure 1 or Fig.26 The exemplary device 10 shown in FIG. 1 receives data. The received data includes crop variety data, environmental data, crop management data and location data of the farmland.

[0214] Crop variety data is related to crops growing or to be grown on farmland. Exemplary crop variety data may include, but are not limited to, growth stage at a particular point in time, crop density (i.e., the number of crops present in a field per unit area), and days after planting. Crop variety data may be obtained from field data for farmland.

[0215] The environmental data indicates environmental conditions of the farmland. Exemplary environmental data may include, but are not limited to, air temperature, cloud cover, dew point, shortwave radiation, longwave radiation, ice accumulation period, liquid accumulation period, relative humidity, adjusted precipitation accumulation period, snow accumulation period, wind speed. In some examples, the environmental data may be collected by sensors deployed on the farmland. In some examples, the environmental data may be received from a weather forecast service.

[0216] The crop management data indicates a fungicide spraying history of the farmland. The crop management data may be obtained from a data management system storing a fungicide spraying history of the farmland.

[0217] The location data of the farmland may include latitude and longitude data of the farmland (eg, decimal degrees, with negative values ​​for south or west), which may be obtained from field data of the farmland.

[0218] At box 220, i.e., step b), a machine learning model is applied to the received data to determine disease progression time series data of the fungal disease. The machine learning model has been trained to learn disease progression under conditions defined by crop variety data, environmental data, crop management data, and location data based on historical data collected from one or more agricultural fields.

[0219] The machine learning model may be any suitable model capable of determining disease progression time series data of a fungal disease. Examples of machine learning models may include, but are not limited to, XGB regression models, artificial neural networks, and support vector regression.

[0220] An exemplary machine learning model, namely the XGB regression model, is discussed in detail above and specifically in Section 2, “First Infection Determination Module”.

[0221] In some examples, multiple machine learning models are provided for two or more fungal diseases. Each machine learning model is trained for a single disease.

[0222] At block 230, i.e., step c), the disease onset date of the fungal disease is determined based on the determined disease progression time series data. For example, the disease onset date of the fungal disease is determined using a change point detection algorithm.

[0223] As an optional step, at box 240, i.e., step d), a process-based model can be applied to determine the infection rate of the fungal disease after the disease onset date under conditions defined by the crop variety data, environmental data, crop management data, and location data. The process-based model can include a SEIR model. In some examples, the infection rate of the fungal disease can be determined by further including conditions defined by the variety disease resistance level of the crop.

[0224] As another optional step, at block 250, step e), a fungicide spray schedule may be determined based on the determined disease progression. The fungicide spray schedule may be based on the fungicide application versus disease progression curve (e.g., Fig.25 In this way, the timing of spraying can be improved by taking into account the dynamic changes of the disease, thereby improving the treatment.

[0225] As yet another optional step, at box 260, step f), a configuration file is generated based on the fungicide spraying schedule, which configuration file is preferably usable for configuring a sprayer for fungicide spray application.

[0226] It should be understood that the above operations may be performed in any suitable order, for example, serially, simultaneously, or a combination thereof, subject to specific order necessitated by, for example, input / output relationships where applicable.

[0227] In another exemplary embodiment of the present invention, a computer program or a computer program element is provided, which is characterized by being adapted to execute the method steps of the method according to one of the preceding embodiments on a suitable system.

[0228] Therefore, the computer program elements can be stored on a computer unit, which can also be part of an embodiment of the present invention. The computing unit can be adapted to perform or cause the steps of the above method to be performed. In addition, the computing unit can be adapted to operate the components of the above device. The computing unit can be adapted to automatically operate and / or execute the user's commands. The computer program can be loaded into the working memory of a data processor. Therefore, the data processor can be equipped to perform the method of the present invention.

[0229] This exemplary embodiment of the invention covers both a computer program which right from the beginning uses the invention and a computer program which by means of an update turns an existing program into a program which uses the invention.

[0230] Furthermore, the computer program element may be capable of providing all necessary steps to complete the procedure of the exemplary embodiment of the method described above.

[0231] According to another exemplary embodiment of the invention, a computer readable medium, such as a CD-ROM, is presented, wherein the computer readable medium has stored thereon a computer program element, which has been described in the preceding section.

[0232] The computer program may be stored and / or distributed on a suitable medium, such as an optical storage medium or solid-state medium provided with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunications systems.

[0233] However, the computer program may also be presented via a network such as the World Wide Web and may be downloaded from such a network into a working memory of a data processor. According to another exemplary embodiment of the present invention, a medium for making a computer program element available for download is provided, the computer program element being arranged for performing a method according to one of the aforementioned embodiments of the present invention.

[0234] It must be noted that embodiments of the present invention are described with reference to different subject matters. In particular, some embodiments are described with reference to method type claims, while other embodiments are described with reference to device type claims. However, those skilled in the art will appreciate from the above and following descriptions that, unless otherwise notified, any combination of features relating to different subject matters, in addition to any combination of features belonging to one type of subject matter, is also considered to be disclosed with the present application. However, all features may be combined to provide synergistic effects, rather than simply the addition of these features.

[0235] Although the present invention has been shown and described in detail in the drawings and the foregoing description, such showing and description should be considered illustrative or exemplary rather than restrictive. The present invention is not limited to the disclosed embodiments. Other variations of the disclosed embodiments may be understood and implemented by those skilled in the art in practicing the claimed invention through study of the drawings, the present disclosure, and the appended claims.

[0236] All definitions, as defined and used herein, should be understood to control over dictionary definitions, definitions in documents incorporated by reference, and / or ordinary meanings of the defined terms.

[0237] As used herein in the specification and claims, the indefinite article "a" or "an" should be understood to mean "at least one", unless expressly indicated to the contrary. In other words, the indefinite article "a" or "an" does not exclude a plurality.

[0238] The phrase "and / or" as used herein in the specification and claims should be understood to mean "one or both" of the elements so connected (i.e., elements that are present in conjunction in some cases and separately in other cases). Multiple elements listed with "and / or" should be interpreted in the same manner, i.e., "one or more" of the elements so connected. In addition to the elements specifically identified by the "and / or" clause, other elements may optionally be present, whether or not related to those elements specifically identified.

[0239] A single processor or other unit may fulfill the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not mean that a combination of these measures cannot be used to advantage. Any reference signs in the claims should not be construed as limiting the scope.

Claims

1. A computer-implemented method (200) for determining disease progression for fungicide spray scheduling on agricultural fields, the method comprising: a) receiving (210) data, the data comprising: - Crop variety data relating to crops growing or to be grown on the farmland; - Environmental data indicating the environmental conditions of the agricultural land; - Crop management data indicating the fungicide spraying history of the field; and - Location data of the farmland; b) applying (220) a machine learning model to the received data to determine disease progression time series data for a fungal disease, wherein the machine learning model has been trained to learn disease progression under conditions defined by crop variety data, environmental data, crop management data, and location data based on historical data collected from one or more agricultural fields; and c) determining (230) a disease onset date of the fungal disease based on the determined disease progression time series data.

2. The computer-implemented method of claim 1, in, The disease onset date for this fungal disease was determined using a change point detection algorithm.

3. The computer-implemented method according to claim 1 or 2, in, Multiple machine learning models are provided for two or more fungal diseases, and each machine learning model has been trained for a single disease.

4. A computer-implemented method according to any one of the preceding claims, in, The machine learning model includes the extreme gradient boosting XGB regression model.

5. The computer-implemented method of any preceding claim, further comprising: d) applying (240) a process-based model to determine the infection rate of the fungal disease after the disease onset date under conditions defined by the crop variety data, the environmental data, the crop management data, and the location data.

6. The computer-implemented method of claim 5, in, The infection rate of the fungal disease is determined by further including conditions defined by the cultivar disease resistance level of the crop.

7. A computer-implemented method according to claim 5 or 6, in, The infection rate of the fungal disease is determined by further including conditions defined by fungicide application data including fungicide data for a fungicide product to be used and at least one planned application timing.

8. A computer-implemented method according to any one of claims 5 to 7, in, The process-based model includes the susceptible-exposed-infected-recovered SEIR model.

9. A computer-implemented method according to any one of the preceding claims, in, The crop variety data includes one or more of the growth stage of the crop, the number of days after planting, and the variety disease resistance level of the crop.

10. A computer-implemented method according to any one of the preceding claims, in, The environmental data includes one or more of the following: air temperature, cloud cover, shortwave radiation, longwave radiation, ice accumulation period, liquid accumulation period, relative humidity, adjusted precipitation accumulation period, snow accumulation period, and wind speed.

11. A computer-implemented method according to any one of the preceding claims, in, The location data includes latitude and longitude data.

12. The computer-implemented method of any preceding claim, further comprising: e) determining (250) a fungicide spray schedule based on the determined disease progression.

13. The computer-implemented method of any preceding claim, further comprising: f) generating (260) a configuration file based on the fungicide spraying schedule, the configuration file preferably being usable for configuring a sprayer for fungicide spray application.

14. An apparatus for generating disease progression for use in fungicide spraying schedules on agricultural fields, the apparatus comprising one or more processing units for generating application schemes, wherein: The one or more processing units comprise instructions which, when executed on the one or more processing units, perform the method steps of any one of the preceding claims.

15. A computer program element comprising instructions for causing an apparatus as claimed in claim 14 to perform the method steps as claimed in any one of claims 1 to 13.

Images