Several tools were developed during the past decades to support farmers in nutrient management and to meet legal requirements such as the farm specific excretion tool. This tool is used by dairy farmers to estimate the farm specific nitrogen (N) excretion of their animals, which is calculated from farm specific data and some normative values. Some variables, like intake of grazed grass or roughage, are hard to measure. A data driven approach could help finding structures in data, and identifying key factors determining N excretion. The aim of this study was to benchmark machine learning methods such as Bayesian Network (BN) and boosted regression trees (BRT) in predicting N excretion, and to assess how sensitive both approaches are on the absence of hard-to-measure input variables. Data were collected from 25 Dutch dairy farms. In the period 2006–2018, detailed recordings of N intake and output were made during 6–10 weeks distributed over each year. Variables included milk production, feed intake and their composition. Calculated N excretion was categorized as low, medium, and high, with limits of 300 and 450 g/day/animal. Accuracy of prediction of the farm specific N excretion, and distinguishing the low and high cases from the medium ones, was slightly better with BRT than with BN. Leaving out information on intake during grazing did not negatively influence validation performance of both models, which opens opportunities to diminish data collection efforts on this aspect. Further analyses are required to confirm these results, such as cross-validation.
Air pollution control and mitigation are important factors in wellbeing and sustainability. To this end, air pollution monitoring has a significant role. Today, air pollution monitoring is mainly done by standardized stations. The spread of those stations is sparse and their cost hinders the option of adding more. Thus, arises the need for cheaper and available means to assess air pollution. In this article, a method for assessing air pollution levels by means of multi angle imaging is presented. Specifically, the focus is on estimating images’ blur as an indication for PM (Particulate Matter) ambient levels. The suggested method applies back-projection Radon transform. By back projection methodology, particles’ concentration at each voxel in a 3D space is reconstructed from photos taken from a few different angles.
Tackling some of the grand global challenges, agro-environmental research has turned more and more into an international venture, where distributed research teams work together to solve complex research questions. Moreover, the interdisciplinary character of these challenges requires that a large diversity of different data sources and information is combined in new, innovative ways. There is a pressing need to support researchers with environments that allow them to efficiently work together and co-develop research. As research is often data-intensive, and big data becomes a common part of a lot of research, such environments should also offer the resources, tools and workflows that allow to process data at scale if needed. Virtual research environments (VRE), which combine working in the Cloud, with collaborative functions and state of the art data science tools, can be a potential solution. In the H2020 AGINFRA+ project, the usability of the VREs has been explored for use cases around agro-climatic modelling. The implemented pilot application for crop growth modelling has successfully shown that VREs can support distributed research teams in co-development, helps them to adopt open science and that the VRE’s cloud computing facilities allow large scale modelling applications.
Virtual Research Environments (VREs) bridge the gap between the compute and storage infrastructure becoming available as the ‘cloud’, and the needs of researchers for tools supporting open science and analytics on ever larger datasets. In the AGINFRA PLUS project such a VRE, based on the D4Science platform, was examined to improve and test its capabilities for running large numbers of crop simulations at field level, based on the WOFOST-WISS model and Dutch input datasets from the AgroDataCube. Using the gCube DataMiner component of the VRE, and based on the Web Processing Service standard, a system has been implemented that can run such workloads successfully on an available cluster, and with good performance, providing summarized results to agronomists for further analysis. The methods used and the resulting implementation are briefly described in this paper. Overall the approach seems viable and opening the door to many follow-up implementation opportunities and further research. Some of them are indicated in more detail in the conclusions.
There is growing interest in the application of the latest trends in computing and data science methods to improve environmental science. However we found the penetration of best practice from computing domains such as software engineering and cloud computing into supporting every day environmental science to be poor. We take from this work a real need to re-evaluate the complexity of software tools and bring these to the right level of abstraction for environmental scientists to be able to leverage the latest developments in computing. In the Models in the Cloud project, we look at the role of model driven engineering, software frameworks and cloud computing in achieving this abstraction. As a case study we deployed a complex weather model to the cloud and developed a collaborative notebook interface for orchestrating the deployment and analysis of results. We navigate relatively poor support for complex high performance computing in the cloud to develop abstractions from complexity in cloud deployment and model configuration. We found great potential in cloud computing to transform science by enabling models to leverage elastic, flexible computing infrastructure and support new ways to deliver collaborative and open science.
Effective grassland management practices require a good understanding of soil and vegetation properties, that can be quantified by farmers' knowledge and remote sensing techniques. Many systems have been proposed in the past for grassland monitoring, but open-source alternatives are increasingly being preferred. In this paper, a system is proposed to process data in an open-source and automated way. This system made use of Sentinel-2 data to support grassland management at Haus Riswick in the region around Kleve, Germany, retrieved with help of a platform called Sentinelsat that was developed by ESA. Consecutive processing steps consisted of atmospheric correction, cloud masking, clipping the raster data, and calculation of vegetation indices. First results from 2018 resembled the mowing regime of the area with four growing cycles, although outliers were detected due to a lack of data caused by cloud cover. Moreover, that year's extremely dry summer was visible in the time-series pattern as well. The proposed script is a primary version of a processing chain, which is suitable to be further expanded for more advanced data pre-processing and data analysis in the future.
Crisis Management (CM) is a challenging area when it comes to connecting solutions aiming to support the various tasks involved in handling of CM situations. DRIVER+ [1] an EU-funded project launched in 2014 was setting up a technical infrastructure (so called Test-bed) that allows to interconnect solutions, so they can interact and exchange all crisis relevant information that commanders need, to make their decisions and plan their actions related to a specific crisis. To verify the DRIVER+ Test-bed as well as the DRIVER+ Trial Guidance Methodology [3] and furthermore, to overcome identified CM gaps [2], a series of Trials was setup. Trial-Austria was the fourth one to be executed. This Trial was especially challenging as it was held as a field exercise in parallel to a huge European Civil Protection Exercise (called IRONORE2019). The scenario to be dealt with was an earthquake scenario. The developed methodology, Test-bed as well as various solutions taking part in DRIVER+ are a perfect base and platform to deal with whatever hazard (e.g. chemical, physical, etc.) is endangering our environment or wellbeing.
A simulation framework is presented (WISS, Wageningen Integrated Systems Simulator) which targets the agro-ecological modelling domain. Especially simulation for a large number of locations, such as in detailed regional and global simulation studies. The framework strengths are in modularization, control, speed, robustness and computational protection (multiple system checks during simulation). The WOFOST model is currently implemented in WISS, through which it is used in a number of Wageningen University and Research projects. WISS is written in Java and the framework code is freely available.
Mathematical programming has been the principal workhorse behind most diet models since the 1940s. As a predominantly hypothesis-driven modelling paradigm, its structure is mostly defined by a priori information, i.e. expert knowledge. In this paper we consider two machine learning paradigms, and three instances thereof that could help leverage the readily available data and derive valuable insights for modelling healthier, and acceptable human diets.
The growing relevance of Big Data and the Internet of Things (IoT) leads to a need for an efficient handling of this data. One key concept to achieve efficient data handling is their semantic description. In the environmental and energy domain, these issues become more relevant since there are measurement stations that produce large amounts of data that software systems have to deal with. In the context of cloud-based infrastructure and virtualisation via containers, microservice architectures and scalability become important aspects in software engineering. This article presents the design of a web service providing software systems with semantic descriptions of data fostering a microservice architecture. It implements key concepts such as domain modelling, schema versioning and schema modularisation. It is evaluated and demonstrated in the context of a current environmental use case.
Large farmers’ datasets can help shed light on agroecological processes if used in the context of hypothesis testing. Here we used an anonymized set of data from the geoplatform Akkerweb to better understand the correlation between within-field elevation and normalized differential vegetation index (NDVI, a proxy for biomass). The dataset included 3249 Dutch potato fields, for each of which the cultivar, the field polygon, the year of cultivation and the soil type (clay or sandy) was known. We hypothesize that under dry conditions such correlation is negative, meaning that the lowest portions of the field have more biomass because of water redistribution. From the data, we observed that in dry periods, such as the summer of 2018, the correlation was negative in sandy soils. Furthermore, we observed that early cultivars show a weaker correlation between NDVI and elevation than late cultivars, possibly because early cultivar escape part of the long dry summer spells. We conclude that the correlation between NDVI and elevation may be a useful indicator of drought stress, and deviations from the norm may be useful to evaluate the resistance to drought of individual cultivars.
The recent popularity of artificial intelligence techniques and the wealth of free and open access Copernicus data have led to the development of new data analytics applications in the Earth Observation domain. Among them, is the detection of changes on image time series, and in particular, the estimation of levels and superficies of changes. In this paper, we propose an unsupervised framework to detect generic but relevant and reliable changes using pairs of Sentinel-2 images. To illustrate this method, we will present a scenario focusing on the detection of changes in vineyards due to natural hazards such as frost and hail.
The last decade infrastructure systems have been under strain around the globe. The 2008 financial crisis, the so-called fourth industrial revolution, ongoing urbanisation and climate change have contributed to the emergence of an infrastructural crisis that has been labelled as infrastructural gap. During this period, infrastructure systems have increasingly become sites of public contestation with significant effects on their operation and governance. At stake has been the issues of access to infrastructure, their social and environmental consequences and the ‘modern ideal’ embodied in the design of those socio-technical systems. With this paper we apply a cross-disciplinary methodology in order to document and define the practices of this new wave of infrastructural contestation, taking Greece in the 2008–2017 period as the case study. The synthesis of quantitative and qualitative datasets with ethnographic knowledge help us, furthermore, to record tendencies and patterns in the ongoing phenomenon of infrastructural contestation (This study is part of infra-demos project ( www.infrademos.net ), which is funded by a VIDI grant awarded by the Dutch Organisation of Science, PI: Prof. Dimitris Dalakoglou, Dept. of Social and Cultural Anthropology, Vrije Universiteit Amsterdam).
BEACON is a market-led project that couples cutting edge Earth Observation (EO) technology with weather intelligence and blockchain to deliver a toolbox for the Agricultural Insurance (AgI) sector with timely cost-efficient and actionable insights for the agri-insurance industry. BEACON enables insurance companies to exploit the untapped market potential of AgI, while contributing to the redefinition of existing AgI products and services. The Damage Assessment Calculator of BEACON employs remote sensing techniques in order to improve the quality and cost-effectiveness of agri-insurance by: (i) increasing the objectivity of the experts field inspections; (ii) reducing the cost of field visits and (iii) increasing farmers' confidence in the estimation results, given the significant economic impact of erroneous estimation. This paper provides an analysis of different type of EO data and remote sensing techniques implemented in the operational workflow of BEACON that can be used by AgI companies to provide safe and reliable results on storms, floods, wildfires and droughts damage on crops.
The purpose of this paper is to present the investigation of common requirements and needs of users across a diverse set of precision agriculture and livestock farming use cases that was based on a series of interviews with experts and farmers. The requirements were based on nine interviews that were conducted in order to identify common requirements and challenges in terms of data collection and management, Big Data technologies, High Performance Computing infrastructure and decision making. The common requirements that derived from the interviews and user requirement analysis per use case can serve as basis for identifying functional and non-functional requirements of a technological solution of high re-usability, interoperability, adaptability and overall efficiency in terms of addressing common needs for precision agriculture and livestock farming.
The use of land cover mappings built using remotely sensed imagery data has become increasingly popular in recent years. However, these mappings are ultimately only models. Consequently, it is vital for one to be able to assess and verify the quality of a mapping and quantify uncertainty for any estimates that are derived from them in a reliable manner. For this, the use of validation sets and error matrices is a long standard practice in land cover mapping applications. In this paper, we review current state of the art methods for quantifying uncertainty for estimates obtained from error matrices in a land cover mapping context. Specifically, we review methods based on their transparency, generalisability, suitability when stratified sampling and suitability in low count situations. This is done with the use of a third-party case study to act as a motivating and demonstrative example throughout the paper. The main finding of this paper is there is a major issue of transparency for methods that quantify uncertainty in terms of confidence intervals (frequentist methods). This is primarily because of the difficulty of analysing nominal coverages in common situations. Effectively, this leaves one without the necessary tools to know when a frequentist method is reliable in all but a few niche situations. The paper then discusses how a Bayesian approach may be better suited as a default method for uncertainty quantification when judged by our criteria.
It is evident that machine learning algorithms are being widely impacting industrial applications and platforms. Beyond typical research experimentation scenarios, there is a need for companies that wish to enhance their online data and analytics solutions to incorporate ways in which they can select, experiment, benchmark, parameterise and choose the version of a machine learning algorithm that seems to be most appropriate for their specific application context. In this paper, we describe such a need for a big data platform that supports food data analytics and intelligence. More specifically, we introduce Agroknow’s big data platform and identify the need to extend it with a flexible and interactive experimentation environment where different machine learning algorithms can be tested using a variation of synthetic and real data. A typical usage scenario is described, based on our need to experiment with various machine learning algorithms to support price prediction for food products and ingredients. The initial requirements for an experimentation environment are also introduced.
Yield is one of the key indicators in agriculture. The most common practices provide only one yield value for a whole field according to the weight of the harvested crop. On the contrary, precision agriculture techniques discover spatial patterns within a field to minimise the environmental burden caused by agricultural activities. Field harvesters equipped with sensors provide more detailed and spatially localised values. The measurements from such sensors need to be filtered and interpolated for the purposes of follow-up analyses and interpretations. This study verified the differences between three methods of interpolation (Inverse Distance Weighted, Inverse Distance Squared and Ordinary Kriging) derived from field sensor measurements that were (1) obtained directly from the field harvester, (2) processed by global filters, and (3) processed by global and local filters. Statistical analyses evaluated the results of interpolations from three fully operational Czech fields. The revealed spatial patterns, as well as recommendations regarding the suitability of the interpolation methods used, are presented at the end of this paper.
In many application domains sensor data contributes an important part to the situation awareness required for decision making. Examples range from environmental and climate change situations to industrial production processes. All these fields need to aggregate and fuse many data sources, the semantics of the data needs to be understood and the results must be presented to the decision makers in an accessible way. This process is already defined as the "sensor to decision chain" [11] but which solutions and technologies can be proposed for implementing it? Since the Internet of Things (IoT) is rapidly growing with an estimated number of 30 billion sensors in 2020, it offers excellent potential to collect time-series data for improving situational awareness. The IoT brings several challenges: caused by a splintered sensor manufacturer landscape, data comes in various structures, incompatible protocols and unclear semantics. To tackle these challenges a well-defined interface, from where uniform data can be queried, is necessary. The Open Geospatial Consortium (OGC) has recognized this demand and developed the SensorThings API (STA) standard, an open, unified way to interconnect devices throughout the IoT. Since its introduction in 2016, it has shown to be a versatile and easy to use standard for exchanging and managing sensor data. This paper proposes the STA as the central part for implementing the sensor to decision chain. Furthermore, it describes several projects that successfully implemented the architecture and identifies open issues with the SensorThings API that, if solved, would further improve the usability of the API.
