The final product will be a set of smart farming products and services, such as dynamic phenology estimation, yield prediction, crop growth indices, sowing date recommendation, crop classification, yield damage assessment. The final set of products and services that will be offered will depend on available data and user interactions. The aforementioned services could be customized and fine-tuned to target different stakeholders, i.e. farmer, agri consultants, insurance companies and CAP stakeholders. The baseline methodologies are general, thus, user specific customizations can produce targeted services to the various users.

The first working prototype for the dynamic phenology estimation, crop verification and crop growth indices was delivered at the end of Sprint 1. Based on this, user feedback was used to further fine-tune the product. A yield estimation preliminary methodology concluded at  the end of Sprint 1, as well. Further research and sophistication on the yield prediction and phenology estimation was realised in Sprint 2, with a fully working alpha version being tested in the cotton cultivation period of 2021. The levels of specificity on both the phenology estimation and yield prediction methods change depending on the problem required to solve. For example, the preliminary methodology for yield estimation might be more applicable for the agri insurance stakeholders, where they need to roughly assess the damage effect of the natural disaster on the production of the field. However, this might not be enough for the farmer, where a greater level of detail is necessitated. The idea is to rapidly prototype the methods, as fast as we have useful results targeting different user needs of different specificities. This way we have a continuous co-design method that produces multiple rapid prototypes that can be tested, while at the same time we evolve the models and methods to greater levels of complexity and sophistication.

The final outcome of the pilot will be the beta version of agrowth platform. Agrowth platform combines Earth Observation data from Sentinel-1 and Sentinel-2 missions, along with pertinent vegetation indices, soil and weather data. It makes use of state of the art in Artificial Intelligence in order to support smart farming using EO data. Specifically, it provides a crop monitoring service which offers information and/or actionable advice at the parcel level.

In the current alpha version, the available services per parcel are:

Phenology estimation: It estimates the current phenology stage of crop and the fuzzy transition of them in the course of time. Specifically, it provides the percentage of completion of each stage, and it indicates the major and minor phenological stage in order to highlight this fuzziness of transition between them.

Weather Forecast: It visualizes the max, min ambient temperatures and growing degree days per parcel and it interplays with risk.resagri.eu. The user can be transferred at the click of a button to their parcel on the Resagri app, where they are provided with detailed weather information.

Yield prediction: It predicts the yield in kg/ha, months earlier from the harvest.

Vegetation Indices: It provides charts for the evolution of the vegetation indices NDVI, NDWI, PSRI and some crop specific indices (BORI and BARI in the case of cotton). See Appendix below.

Furthermore, there are some services for a quick overall view of the state of all available parcels on the map of the user.

Overlay Satellite Images: It allows to overlay satellite images (RGB, NDVI, NDMI) over the users’ parcels for a chosen acquisition date and area of interest.

Phenology Stage: It colors the outline of each parcel with a color equivalent to its major phenology stage.

Crop Classification: It colors the outline of each parcel with a color equivalent to its crop.

Agrowth platform served and will be served by NOA’s infrastructure.

Web based platform Agrowth
http://agrowth.beyond-eocenter.eu/

We have come in contact with different types of end-users, and we have deployed several parallel co-design processes (of different co-design types) with each of them in order to upgrade and enrich our services.

In the course of e-shape we have co-designed, developed and provided the aforementioned services, and we aim to optimize them more till the end of the project based on the feedback of end-users.

Different markets/industries

Agricultural chemicals

We are in initial discussions with two of the biggest cotton seed companies in Greece. Specifically, we are in extended dialogue with Corteva Agriscience Hellas in order to co-design existing services in their needs. We wish to consolidate a formal collaboration in terms of applied research and maybe commercialization.

Farmers’ Cooperatives

In this market we have the most extended collaborations till now. We have already signed MoUs with the agricultural cooperative of Orchomenos and producers’ organisation ‘CottonFarsala’. They have provided us with their parcel data, historical yields and in-situ measurements. We freely opened and evaluated with them the alpha version of Agrowth(http://agrowth.beyond-eocenter.eu/) platform (Phenology estimation, Yield prediction, Vegetation Indices, Weather Forecast) for the cultivation period of 2021. Moreover, we  provided the service of optimal seeding to more cooperatives of above-mentioned (i.e., ThesGi, SynAgros etc.) and we are in the process of formalizing our collaboration.

Agri-consultants

Our main co-designer and project partner is Neuropublic. We have shared data for cotton and industrial tomato, and they were the third main user of the alpha version of Agrowth in cotton parcels for the cultivation period of 2021.

Ref. Expected outcome of the pilot and timeline of the pilot.

One can easily refer to the sections of the “Expected outcome of the pilot”, “Timeline of the pilot” or “Nature of the outcome” if one wishes to delve into our accomplished work. Not only this, but also any future perspective of our projects.

The algorithms and models built will be as general and transferable, as possible. Exploring pseudo-labeling techniques, ground truth independent methodologies, to the extent possible.

Specifically, we have already

1. developed a semi-supervised and an unsupervised method for the within-season or near real-time phenology estimation for cotton that can be used in operational agriculture management scenarios,
2. acquired a dataset of cotton growth ground observations, and we intend to release it in open access.
3. designed and used a new protocol for ground observation collection.

Datasets acquired by NOA’s field campaigns in the context of this pilot will be opened to the community. GitHub support for code will be offered for certain pipelines – remains to be defined which. Best practices documents on EO based phenology extraction and yield estimation.

Provided that the pilot is of interest for the GEO/GEOSS, registration will be performed towards the end of Sprint 3.

SDG 2 | ZERO HUNGER

Pilot S1P2 addresses the zero hunger SDG (Goal 2) via supporting farmers to comply with agricultural policies and strengthen the performance of their farm. The pilot offers the agrowth platform, which is a suite of smart farming services enabled by artificial intelligence, earth observations and numerical weather predictions. Crop phenology and crop yield estimation, the two core services of agrowth, constitute key information for agricultural management and thereby actionable knowledge for the farmer, the agricultural consultant, the insurance company and the policy maker. Using agrowth we can i) reduce the cultivation costs, ii) increase productivity and iii) protect the yield from from adverse weather events. All this work is of great significance towards achieving zero hunger, as we help to make more with less.

 SDG 13 | CLIMATE ACTION

Pilot S1P2 addresses the climate action SDG (Goal 13) via ensuring sustainable practises are met (according to the CAP requirements) and via assisting in the educated farm intervention through its services of phenology estimation, yield estimation and weather event predictions. The farmer is enabled to use less input, while achieving greater yield. This is particularly important towards achieving greater agricultural sustainability while making sure the farmers have a fair and stable income.

Podcast

Vassilis Sitokonstantinou SC 1, pilot 2