B037-0003
A Novel Approach on Smart Farming: Combining Past, Present and Future EO, Climatic and NWP datasets Towards an Integrated Crop Monitoring, Risk-Assessment and Yield Forecasting Suite

Wednesday, 9 December 2020
Poster
Nikolaos S Bartsotas1, Vasileios Sitokonstantinou1, Charalambos Kontoes1, Alkiviadis Koukos1, Alexia Tsouni1, Savvas Rogotis2, Dimitrios Sykas2 and Nikolaos Marianos2, (1)National Observatory of Athens, Institute for Astronomy and Astrophysics, Space Applications and Remote Sensing - BEYOND Center of Earth Observation Research and Satellite Remote Sensing, Athens, Greece, (2)NEUROPUBLIC S.A., Piraeus, Greece
Abstract:
The increasing demand for detailed information on crop monitoring and reliable yield assessment, necessitates the procurement of smart farming services and sophisticated implementations that exploit every available source of information towards tailored solutions on specific crop types and regions. Farmers, as well as field professionals (e.g. insurance companies, farmers’ associations, milling agencies), nowadays require a reliable level of information, both in terms of historical risk and vulnerability, as well as an accurate and quick assessment of imminent and current damages. In addition, early warning systems that are capable of predicting conditions that require extreme measures are essential to mitigate the risk.

Using Rodopi area in northeastern Greece as a testbed, an area dominated by cotton fields, we hereby present an integrated approach that has been under development in the framework of the e-shape H2020 EU project (GA 820852), which combines:

i/ 40-year meteorological reanalysis datasets (climatology),

ii/ A dense network of IoT agro-climatic sensing stations, leveraging on gaiasense smart farming system (in-situ observations),

iii/ Sentinel imagery (earth observation/remote sensing), and

iv/ High-resolution atmospheric modeling (NWP).

The main weather perils for the specific area and per crop type are analysed and areas of high risk are identified so as to cater for insurance underwriting needs. A quick damage assessment on parcel level is feasible through the analysis of Sentinel imagery, a feature that greatly simplifies the claim and payout procedure after a damage. Dynamic phenology classifications, as well as the expected yield are estimated through machine learning procedures. Finally, an early warning system that flags the proximity to the upcoming phenological stage or extreme conditions that require immediate actions from the farmers (optimization of irrigation and fertilization) is supported operationally. The system is currently at an advanced development stage and, upon completion, is expected to be transferable and scalable to other regions as well as crop types.