Fig 2 - uploaded by Marc Benoît
Content may be subject to copyright.
Graphical illustration of the two-level sampling method of the Teruti land-cover survey between 1992 and 2003. (a) The entire territory is segmented into 4700 grids. (b) The position of aerial photos taken in each grid. (c) The distribution of 36 sampling points within an aerial photo. One Teruti sampling point covers roughly 100 ha.
Source publication
Assessing the environmental impacts of agricultural production systems requires spatially explicit information regarding cropping systems. Projecting changes in agricultural land use that are caused by changes in land management practices for analyzing the performance of land activity-related policies, such as agricultural policies, also requires t...
Contexts in source publication
Context 1
... sequential land-cover data that were used in this study were derived from Teruti databases. Teruti is a two-level sampling survey of land-cover, which was conducted by the French Ministry of Agriculture ( Ledoux and Thomas, 1992). Fig. 2 illustrates the sampling method that was performed in this survey. At the first sampling level, the entire territory was segmented into 4700 grids, with an area of 12 Â 12 km per grid (Fig. 2a). In most regions, four aerial photos among eight at the positions numbered in 1, 2, 3, 4 ( Fig. 2b) were taken within each grid. In total, ...
Context 2
... Teruti databases. Teruti is a two-level sampling survey of land-cover, which was conducted by the French Ministry of Agriculture ( Ledoux and Thomas, 1992). Fig. 2 illustrates the sampling method that was performed in this survey. At the first sampling level, the entire territory was segmented into 4700 grids, with an area of 12 Â 12 km per grid (Fig. 2a). In most regions, four aerial photos among eight at the positions numbered in 1, 2, 3, 4 ( Fig. 2b) were taken within each grid. In total, 15,579 aerial photos were taken every June during the survey period. One aerial photo covers approximately 3.24 square kilometers. At the second sampling level, 36 evenly spaced sampling points ...
Context 3
... Ministry of Agriculture ( Ledoux and Thomas, 1992). Fig. 2 illustrates the sampling method that was performed in this survey. At the first sampling level, the entire territory was segmented into 4700 grids, with an area of 12 Â 12 km per grid (Fig. 2a). In most regions, four aerial photos among eight at the positions numbered in 1, 2, 3, 4 ( Fig. 2b) were taken within each grid. In total, 15,579 aerial photos were taken every June during the survey period. One aerial photo covers approximately 3.24 square kilometers. At the second sampling level, 36 evenly spaced sampling points (approximately 300 m apart) were systematically distributed within the area of one aerial photo (Fig. ...
Context 4
... 2, 3, 4 ( Fig. 2b) were taken within each grid. In total, 15,579 aerial photos were taken every June during the survey period. One aerial photo covers approximately 3.24 square kilometers. At the second sampling level, 36 evenly spaced sampling points (approximately 300 m apart) were systematically distributed within the area of one aerial photo (Fig. 2c). The land-covers of the entire territory were recorded in a matrix, in which the sampling points are in a row and the annual records of land-cover are in a column. A corpus of 555,382 sampling points that were labeled with their land-cover during the period from 1992 to 2003 was used in this study. This corpus has detailed information ...
Citations
... spatial and temporal resolution or nomenclature) between these datasets makes them difficult to use in a cross-analysis. This situation has resulted in the production of regional to national studies of crop sequences (Levavasseur et al., 2016;Peltonen-Sainio and Jauhiainen, 2019;Stein and Steinmann, 2018;Xiao et al., 2014) but has hampered any analysis at the scale of the EU. Such an analysis would be useful, especially in the context of the EU's Farm to Fork strategy, in which the adoption of more diverse crop rotations is encouraged (European Commission, 2022). ...
Crop diversification is considered a key element of agroecological transition, whereas current dominant cropping systems are known to rely on only a few crop species – like cereals in Europe. To assess the benefits of crop diversification at a large scale, an accurate description of current crop sequences is required as a baseline. However, such a description is lacking at the scale of Europe. Here, we developed the first map of dominant crop sequences in Europe for the period 2012–2018. We used the Land Use Cover Area frame statistical Survey (LUCAS) dataset that provides temporally incomplete (2012, 2015 and 2018) land cover information from a stable grid of points covering Europe. Eight crop sequence types were identified using hierarchical clustering implemented on LUCAS data and mapped over Europe. We show, in France, that the relative importance of these eight crop sequence types (as estimated from LUCAS data) was highly consistent with those derived from an almost spatially exhaustive temporally complete national dataset (the French Land Parcel Identification System) for the same period, thus validating the method and typology for this country. Land use (i.e. crop production area) derived from our map of dominant crop sequences was also highly consistent with land use reported by official statistics at both national and European levels, validating the approach at a European scale. This first map of dominant crop sequences in Europe should be useful for future studies dealing with agricultural issues that are sensitive to crop rotations. The map of dominant crop sequence types in Europe derived from our work is available at https://doi.org/10.5281/zenodo.7016986 (Ballot et al., 2022).
... As a result, expert knowledge-based models have limitations in terms of accuracy and applicability over large areas and long periods. Alternative approaches, such as estimation of crop sequence probabilities using survey data and hidden Markov models have been demonstrated in FR (Xiao et al., 2014), but these methods are not always feasible at large scale due to the extended size of the required sample. ...
Accurate in-season crop type classification is crucial for the crop production estimation and monitoring of agricultural parcels. However, the complexity of the plant growth patterns and their spatio-temporal variability present significant challenges. While current deep learning-based methods show promise in crop type classification from single- and multi-modal time series, most existing methods rely on a single modality, such as satellite optical remote sensing data or crop rotation patterns. We propose a novel approach to fuse multimodal information into a model for improved accuracy and robustness across multiple years and countries. The approach relies on three modalities used: remote sensing time series from Sentinel-2 and Landsat 8 observations, parcel crop rotation and local crop distribution. To evaluate our approach, we release a new annotated dataset of 7.4 million agricultural parcels in France and Netherlands. We associate each parcel with time-series of surface reflectance (Red and NIR) and biophysical variables (LAI, FAPAR). Additionally, we propose a new approach to automatically aggregate crop types into a hierarchical class structure for meaningful model evaluation and a novel data-augmentation technique for early-season classification. Performance of the multimodal approach was assessed at different aggregation level in the semantic domain spanning from 151 to 8 crop types or groups. It resulted in accuracy ranging from 91\% to 95\% for NL dataset and from 85\% to 89\% for FR dataset. Pre-training on a dataset improves domain adaptation between countries, allowing for cross-domain zero-shot learning, and robustness of the performances in a few-shot setting from France to Netherlands. Our proposed approach outperforms comparable methods by enabling learning methods to use the often overlooked spatio-temporal context of parcels, resulting in increased preci...
... However, within the six years of data available, we look for crop rotationsevidence of 1-5-year sequences cycling within the six yearsrather than for crop sequences regardless of whether there is evidence of rotation. Similarly to previous research, we chose to classify rotations using length (Osman et al., 2015), cropping composition (Xiao et al., 2014), and diversity (Conrad et al., 2017;Merlos and Hijmans, 2020;Scheiner and Martin, 2020;Socolar et al., 2021;Stein and Steinmann, 2018;Tiemann et al., 2015). We evaluated all three classification systems in the same investigation, so we could evaluate the findings together. ...
... The framework could be augmented to investigate rotational complexity, such as crop or rotation flexibility (Castellazzi et al., 2008) or transitions between rotations. Rotation mapping and predictions could also be refined by introducing crop classification certainty (Serra and Pons, 2016), agronomic rules or expertise (Bachinger and Zander, 2007;Detlefsen and Jensen, 2007;Dogliotti et al., 2003;Schönhart et al., 2011;Sharp et al., 2021;Xiao et al., 2014), or a metric for agricultural land capability and other biophysical covariates (e.g. Goodwin et al., n.d.;Socolar et al., 2021). ...
Cropping decisions affect the nature, timing and intensity of agricultural management strategies. Specific crop rotations are associated with different environmental impacts, which can be beneficial or detrimental. The ability to map, characterise and accurately predict rotations enables targeting of mitigation measures where most needed and forecasting of potential environmental risks. Using six years of the national UKCEH Land Cover® plus: Crops maps (2015–2020), we extracted crop sequences for every agricultural field parcel in Great Britain (GB). Our aims were to first characterise spatial patterns in rotation properties over a national scale based on their length, type and intra- and inter-rotation diversity values, second, to test an approach to predicting the next crop in a rotation, using transition probability matrices, and third, to test these predictions at a range of spatial scales. Strict cyclical rotations only occupy 16 % of all agricultural land, whereas long-term grassland and complex-rotational agriculture each occupy over 40 %. Our rotation classifications display a variety of distinctive spatial patterns among rotation lengths, types and diversity values. Rotations are mostly 5 years in length, short mixed crops are the most abundant rotation type, and high structural diversity is concentrated in east Scotland. Predictions were most accurate when using the most local spatial approach (spatial scaling), 5-year rotations, and including long-term grassland. The prediction framework we built demonstrates that our crop predictions have an accuracy of 36–89 %, equivalent to classification accuracy of national crop and land cover mapping using earth observation, and we suggest this could be improved with additional contextual data. Our results emphasise that rotation complexity is multi-faceted, yet it can be mapped in different ways and forms the basis for further exploration in and beyond agronomy, ecology, and other disciplines.
... These models would be extremely useful for the evaluation of agricultural and environmental policies that will shape the farms of tomorrow. Indeed, "In the context of the establishment of new economic, agronomic and governmental policies, farmers will be paid for re-establishing and increasing ecosystem services on agricultural land" [40]. ...
... For Klöcking et al. [41] for instance, the economic factor is often more important than the agronomic factor in the final decision of the producer. Yet agronomic principles are at the heart of many models [40]. ...
To meet global food requirements while responding to the environmental challenges of the 21st century, an agri-environmental transition towards sustainable agricultural practices is necessary. Crop rotation is an ancestral practice and is a pillar of sustainable agriculture. However, this practice requires more organization on the part of producers for the management of crop inputs. That is why the development of a methodology for forecasting crop rotations in the medium term and at the field level is necessary. However, to date, only a methodology based on the Seq2Seq-LSTM has been theorized without being tested on a concrete case of application. The objective of this article is therefore to evaluate the performance of a Seq2Seq-LSTM methodology to predict crop rotations on a real case. The methodology was applied to a problem of crop rotation prediction for field crop farms in Québec, Canada. Using the Recall(N) metric and a historical sequence of length 6, the next 3 crops grown in a field can be predicted with over 81% success when considering 10 selected options. In addition, the methodology was augmented with contextual information such as economic and meteorological data to refine the forecasts. This augmentation systematically improves the performance of the model. This observation provides a relevant line of research for identifying other factors that influence producers’ decision-making on crop rotation.
... spatial and temporal resolution) between these datasets makes them difficult to use in a cross-analysis. This situation has resulted in the production of regional to national studies of crop sequences (Levavasseur et al., 2016;Stein and Steinmann, 2018;Xiao et al., 2014) but has hampered any analysis at the scale of the European Union (EU). 45 ...
Crop diversification is considered as an important linchpin of the agroecological transition, whereas current dominant cropping systems are known to rely only on a few crops species – like cereals in the European Union (EU). To assess the benefits of crop diversification at large scale, an accurate description of current crop sequences is required as a baseline. However, such a description is lacking at the scale of the EU. Here, we developed the first map of dominant crop sequences in the EU for the period 2012–2018. We used the LUCAS dataset that provides temporally-incomplete land cover information from a stable grid of points covering the whole EU. Eight crop sequence types were identified using hierarchical clustering implemented on LUCAS data, and mapped over EU. We show, in France, that the relative importance of these eight crop sequence types (as estimated from LUCAS data) was highly consistent with those derived from an almost spatially-exhaustive temporally-complete national dataset (the French Land Parcel Identification System) for the same period, thus validating the method and the typology for this country. Land use (i.e. crop production area) derived from our map of dominant crop sequences was also highly consistent with land use reported by official statistics, both at national and EU levels, validating the approach at the EU-scale. This first map of dominant crop sequences in the EU should be useful for future studies dealing with agricultural issues that are sensitive to crop rotations. The map of dominant crop sequences types in the EU derived from our work is available at https://doi.org/10.5281/zenodo.7016986 (Ballot et al., 2022).
... The second goal, enabled by the fixed sample structure within each type of survey (i.e., time series), examines the annual succession of land-cover types of the areas of sampling points classified as agricultural land to establish their crop sequences. Agronomic studies were performed using these surveys to analyze spatial differentiation of crop sequences in France and its dynamics (Mignolet et al. 2004(Mignolet et al. , 2007Leenhardt et al. 2012b;Xiao et al. 2014Xiao et al. , 2015. These studies also resulted in development of data-mining software that automates analysis of spatiotemporal patterns in land-use and land-cover data (Le Ber et al. 2006). ...
This chapter addresses collection and integration of large amounts of data from a variety of sources when developing agronomic approaches at the landscape and territory levels. Data commonly used in agro-environmental studies are climate, soils, land cover and land use (including cropping practices). They are used as inputs for indicator- or model-based assessment methods. Additional information is now required for agroecological studies, such as spatial distribution of weeds, pests, and diseases; biodiversity; or landscape features. Gathering and using this scattered and heterogeneous information for integrated studies at watershed, regional, or national levels still requires further methodological efforts. Remote sensing is a source of data that continuously progresses in spatial and temporal resolution, and accessibility. Such integrative approaches are illustrated for a case study in France. The next challenges and opportunities for data collection, integration, and governance are discussed, with a focus on mainland France.KeywordsClimatic dataSoil dataLand coverCropping practicesRemote sensingMapsDatabases
... The second goal, enabled by the fixed sample structure within each type of survey (i.e., time series), examines the annual succession of land-cover types of the areas of sampling points classified as agricultural land to establish their crop sequences. Agronomic studies were performed using these surveys to analyze spatial differentiation of crop sequences in France and its dynamics (Mignolet et al. 2004(Mignolet et al. , 2007Leenhardt et al. 2012b;Xiao et al. 2014Xiao et al. , 2015. These studies also resulted in development of data-mining software that automates analysis of spatiotemporal patterns in land-use and land-cover data (Le Ber et al. 2006). ...
... However, there are many challenges that remain. Determining the spatial distribution of crop cultivation is one of the most important studies needed to understand the implementation directions to achieve food security (Xiao et al., 2014;Yusianto and Hardjomidjojo, 2020). Several countries are preparing up-to-date land crop/cover maps for their agricultural areas, but in many parts of the world, it is not possible to access up-to-date land crop/cover data and maps. ...
Preparing up-to-date land crop/cover maps is important to study in order to achieve food security. Therefore, the aim of this study was to evaluate the impact of surface biophysical features in the land crop/cover classification accuracy and introduce a new fusion-based method with more accurate results for land crop/cover classification. For this purpose, multi-temporal images from Sentinel 1 and 2, and an actual land crop map prepared by Agriculture and Agri-Food Canada (AAFC) in 2019 were used for 3 test sites in Ontario, Canada. Firstly, surface biophysical features maps were prepared based on spectral indices from Sentinel 2 including Normalized Difference Vegetation Index (NDVI), Index-based Built-up Index (IBI), Wetness, Albedo, and Brightness and co-polarization (VV) and cross-polarization (VH) from Sentinel 1 for different dates. Then, different scenarios were generated; these included single surface biophysical features as well as a combination of several surface biophysical features. Secondly, land crop/cover maps were prepared for each scenario based on the Random Forest (RF). In the third step, based on the voting strategy, classification maps from different scenarios were combined. Finally, the accuracy of the land crop/cover maps obtained from each of the scenario was evaluated. The results showed that the average overall accuracy of land crop/cover maps obtained from individual scenario (one feature) including NDVI, IBI, Wetness, Albedo, Brightness, VV and VH were 66%, 68%, 63%, 60%, 57%, 62% and 58%, respectively, which by the surface biophysical features fusion, the overall accuracy of land crop/cover maps increased to 83%. Also, by combining the classification results obtained from different scenarios based on voting strategy, the overall accuracy increased to 89%. The results of this study indicate that the feature level-based fusion of surface biophysical features and decision level based fusion of land crop/cover maps obtained from various scenarios increases the accuracy of land crop/cover classification.
... This knowledge is the basis for the prevention of soil and water pollution (Beaudoin et al., 2005). Although many models are based on agronomic knowledge to define crop rotations (Xiao et al., 2014), according to Klöcking et al. (2003), crop rotations are more impacted by economic rather than agronomic factors. As mentioned by Xiao et al. (2014), "In the context of the establishment of new economic, agronomic and governmental policies, farmers will be paid for re-establishing and increasing ecosystem services on agricultural land". ...
... Although many models are based on agronomic knowledge to define crop rotations (Xiao et al., 2014), according to Klöcking et al. (2003), crop rotations are more impacted by economic rather than agronomic factors. As mentioned by Xiao et al. (2014), "In the context of the establishment of new economic, agronomic and governmental policies, farmers will be paid for re-establishing and increasing ecosystem services on agricultural land". Ex ante evaluation of agricultural and environmental policies would benefit from the use of farm models (Edwards-Jones and McGregor, 1994). ...
Meeting an increasing demand for food while preserving the environment is one of the most important challenges of the 21st century. To meet this challenge, conservation agriculture can rely on the age-old practice of crop rotation. The objective of this article is to develop a methodology for predicting and visualizing crop rotations, supporting discussions between agronomists and producers. Based on crop history data, the 6-phase methodology, uses Markov chains for the prediction of the N most likely crops grown in year n + 1. Process mining and Directly-Follows Graphs (DFG) enables modelling and visualization of the results. Generalisation and filtering operations highlight the frequent behaviors of producers. Applied to analyse the crop history of 10,376 fields from 409 field crop farms in Quebec, Canada, the methodology is competitive with the performance of various recurrent neural networks (LSTM, RNN, GRU) with a successful prediction rate that exceeds 90%, while allowing for an intelligibility of results and a relative computational simplicity.
... However, within the six years of data available, we look for crop rotationsevidence of 1-5-year sequences cycling within the six yearsrather than for crop sequences regardless of whether there is evidence of rotation. Similarly to previous research, we chose to classify rotations using length (Osman et al., 2015), cropping composition (Xiao et al., 2014), and diversity (Conrad et al., 2017;Merlos and Hijmans, 2020;Scheiner and Martin, 2020;Socolar et al., 2021;Stein and Steinmann, 2018;Tiemann et al., 2015). We evaluated all three classification systems in the same investigation, so we could evaluate the findings together. ...
... The framework could be augmented to investigate rotational complexity, such as crop or rotation flexibility (Castellazzi et al., 2008) or transitions between rotations. Rotation mapping and predictions could also be refined by introducing crop classification certainty (Serra and Pons, 2016), agronomic rules or expertise (Bachinger and Zander, 2007;Detlefsen and Jensen, 2007;Dogliotti et al., 2003;Schönhart et al., 2011;Sharp et al., 2021;Xiao et al., 2014), or a metric for agricultural land capability and other biophysical covariates (e.g. Goodwin et al., n.d.;Socolar et al., 2021). ...
