#DigitAg Day: Satellite imaging, spatial analysis & precision agriculture, November 27 2018

On Tuesday 27 November in Montpellier, #DigitAg is organising a “Satellite imagery, spatial analysis and precision agriculture” day on the occasion of the institute’s first PhD defence! In the morning, Ruth Kerry, member of her thesis jury, will give a lecture on “spatial analysis for natural resource management”. In the afeternoon, Corentin Leroux, doctoral student at Cifre SMAG-Montpellier SupAgro (UMR ITAP) will defend his thesis on “Processing and valorization of spatialized information in Precision Agriculture: Application to intra-parcel yield data”. In the morning, Ruth Kerry, member of her thesis jury, will give a lecture on “spatial analysis for natural resource management”. The conference is open to all on registration, mandatory, before November 20, 2018. The PhD defense is public, subject also to registration. Please note: the presentations and exchanges of the day will be in English.





Invited lecture registration

Campus Lavalette, AgroParisTech, room: Amazone (map – 648 Rue Jean François Breton à Montpellier)


10h: Presentation of the day – Introduction to #DigitAg Convergence Lab, by Véronique Bellon-Maurel

10h30-11h30:  Invited lecture “Using Spatial Analysis to Advance Natural Resource Management”
by Dr Ruth Kerry, Brigham Young University’s (Utah, USA)

Dr Ruth Kerry completed her undergraduate at Oxford University in Physical Geography. Her MSc was in Soil Spatial Analysis and Land Evaluation from Reading University (UK) and in 2004 she obtained her PhD from Reading supervised by Professor Margaret Oliver. Her topic was investigating geostatistical methods for soil mapping in precision agriculture. She joined Brigham Young University’s (Utah, USA) Geography Department in 2004 choosing to teach and research part-time whilst caring for her five children. Her primary research focuses have been involved with applying geostatistical methods to soil mapping; however, being based in a Geography Department, her range of applications has broadened from soils, agricultural and environmental problems to include studies in Crime and Health Geography. Dr Kerry’s research has benefitted from sharing techniques between human and physical geography. Indeed, she has edited a double special issue of Geographical Analysis on the application of geostatistical methods in 1) Human and 2) Physical Geography. She is currently working as editor on a Special Issue of Precision Agriculture as she also did in 2008 and is compiling a co-edited book for Springer on Sensing Approaches for Precision Agriculture.n. More information: https://www.researchgate.net/profile/Ruth_Kerry

Abstract:  Creating an accurate map of any natural phenomenon to manage a particular resource requires collecting and analyzing samples of the phenomenon of interest which is expensive. There are now many sensors and remotely sensed data that provide dense spatial data that are related to a range of environmental attributes but these always need ground truth or calibrating before they can be used to infer a given phenomenon. This ground-truthing and calibration requires academic research to determine the most appropriate sensing methods and data analysis techniques. Soil is a key natural resource that needs managing appropriately so that its quality is not degraded and so it can sustain future agricultural production. At the 2017 European Conference for Precision Agriculture, one of the plenary speakers, a farmer who has been practicing precision agriculture for 20 years, said that the problem of affordable yet accurate mapping of soil properties for precision farming had not been resolved. In this seminar a number of case studies showing more cost effective methods of soil mapping that utilize sensed data will be presented. The case studies cover a range of spatial scales and applications for mapping. They will show that many of the problems with accurate soil mapping have been resolved, but there is a great need for intelligent decision support software to bridge the gap between academic research and farmers/agronomists. Such software should be able to automatically but intelligently analyze data for the farmer. A flow diagram that could be used as a starting point for automatic processing of soil data for precision farming will be discussed and some new approaches from other disciplines which could solve other problems in precision agriculture will be presented.



13h30: Corentin Leroux’ PhD defense

> registration

Campus La Gaillarde, Montpellier SupAgro, amphi 208 (map – 2 Place Viala)



Corentin Leroux will defend his doctoral thesis on:
“Processing and value-adding to spatial information in Precision Agriculture: Application to within-field yield monitor data”



Keywords: Filtering, Geostatistics, Mapping, Precision Agriculture, Within-field spatial and temporal variability, Yield monitor data, Zoning

Abstract: Precision Agriculture (PA) aims at combining georeferenced information and communication technologies to improve agrosystems’ management. Within-field yield data arising from sensors mounted on combine harvesters are one of the primary and symbolic information in precision agriculture. Thanks to global navigation satellite systems (GNSS), yield sensors enable to generate highly resolute within-field yield maps. These latter maps can be used to help to characterize the within-field crop production variability and make decisions for the management of upcoming crops. Nowadays, yield maps are however under-exploited and sometimes not even used by the agriculture professionals for a couple of reasons: non-reliable processing chain, non-simplified maps, very few yield-based operational services….

Although several works have already been dedicated to yield data processing when yield monitors were first put into place, it is stressed here that digital technologies have significantly evolved since the raise of these yield monitors, e.g. improvement of communication networks, farm management information systems on the cloud, more and more connected farms. Those new factors require processing methodologies to meet several demands, e.g. automation (lower manual interaction), robustness (given the diversity of data to process), non-parametricity. This digital disruption in the agricultural world supports the need to revisit existing yield processing methods to make sure that they come along with the new context that agriculture is facing. The objective being to (i) evaluate their effectiveness in this new digital context, and (ii) identify the possible paths for improvement and the scientific obstacles that might be faced.

During this PhD program, several yield processing methods and algorithms have been proposed algorithms so that raw yield data could be turned into information and decision layers. Three major axes have been investigated (1) data quality, (2) data spatial representation, and (3) data interpretation. More specifically, a filtering approach has first been implemented to improve the reliability of within-field yield monitor data (1). The proposed methodology is original because it accounts for the data acquisition process of the combine harvester and because the filtering approach can be applied on irregularly-spaced within-field data. Two zoning methods have then been proposed to delineate homogeneous yield units from annual or multi-temporal yield mapping datasets (2). The zoning procedure is interesting in the sense that it can be applied to irregularly-spaced spatial data and that it incorporates operational considerations (farmer’s expertise, size of agriculture machinery). Finally, a decision tree has been put into place to help users to select an appropriate descriptor of within-field variability when working with spatial data (3). Careful attention was paid to the development of automated, robust, and unsupervised methods to cope with the new digital context previously mentioned.