TuberZone: Development of an innovative spatial crop model and decision support system for improved potato agronomy

农业现在是一个数据丰富的环境。大量的近端和远程传感器捕捉农业生产系统的许多不同方面，特别是种植系统。如今，种植者能够记录和改变大多数农艺投入或操作的比率。然而，种植者很少使用他们可以支配的能力，因为他们无法将可用的数据流转化为信息，然后转化为良好的农艺决策。错误的分析会产生错误的决策。正因为如此，种植者对根据他们不太了解的信息做出决策持谨慎态度。一个明确的、潜在的非常重要的空间数据使用方式是在作物模型中。作物模型对于农业界预测作物在不同情景(替代管理和/或季节气候变化演变)下的发展情况是非常宝贵的。虽然存在许多成熟的、可靠的作物模型，但这些模型是建立在对一个点进行建模的假设基础上的，这是对田地或农场的平均反应。它们不是为高分辨率空间建模而设计的，当用作高分辨率空间模型时通常会崩溃。本项目的目标是将点作物模型与空间数据相结合，以生成用于马铃薯生产的有效空间作物模型。这将侧重于预测块茎大小分布(TSD)和管理TSD的各种驱动因素(环境和管理)。通过赋予现有作物模型以空间信息，就有可能将种植者/农学家直接从数据分析和决策中剔除。专家知识将在作物模型中获取，但空间数据和最终用户之间没有直接的参与，从而消除了这一错误和混乱的根源。因此，空间作物模型是一种对原始空间数据进行空间数据融合和增值的方法。该模型提供了一个相对简单的综合空间产出(推荐的可变费率管理操作)，种植者可以获得采用。该模型还允许估计不确定性(以及操作)，以帮助种植者通过不同的管理进行风险评估。从学术角度来看，要实现这一目标，需要研究和发展几个问题。这些措施包括：1)填补马铃薯田作物变异量(大小和空间结构)的知识空白。可用的空间研究很少&这些信息是在可感知的边界范围内正确地将任何空间模型参数化所必需的。2)理解所观察到的作物产量变化的驱动因素。观测到的变异性可以与土壤和天气变化的空间信息以及管理决策联系起来。这有助于为产量决定因素的空间模型提供信息。3)空间元模型的发展。空间裁剪模型依赖于现有点裁剪模型的输出被用作空间元模型的输入。空间元模型是一个新概念。它需要对投入进行标准化，特别是在空间足迹方面，对邻里相互作用进行正确的参数化，并对空间模型中每一点的不确定性进行正确的建模。正确的数据处理&上面第1)点和第2点的知识将确保正确地设计和填充元模型。该项目汇集了英国在马铃薯生产(SAC、SRUC、McCains)、供应链和加工(McCains)、马铃薯生产机械(Grimme)和精密农业服务(SE)方面的领先行业专业知识，以及精准农业(NewCastle Uni)和作物建模(NewCastle Uni)和作物建模(NewCastle Uni)领域的领先学术研究人员。该财团处于有利地位，能够交付项目并满足行业需求。

Agriculture is now a data-rich environment. A multitude of proximal & remote sensors capture many different aspects of agriculture production systems, particularly cropping systems. Nowadays, growers are able to record & change the rates of most agronomic inputs or operations. However, growers rarely use the capabilities at their disposal because they are unable to translate the available data streams into information & then into good agronomic decisions. Incorrect analysis generates incorrect decisions. Because of this, growers are wary to adopt decisions based on information that they do not understand well.One clear, potentially very important way in which these spatial data can be used is within crop models. Crop models are invaluable to the agricultural community to predict how crops develop under different scenarios (alternative management and/or evolving in-season climate variations). While many well developed & well credential crop models exist, these are built on an assumption of modelling a point, which is an average response for a field or farm. They are not designed for high-resolution spatial modelling & usually collapse when used as such.The objective for this project is to integrate a point crop model with spatial data to generate an effective spatial crop model for potato production. This will have an emphasis on predicting tuber size distribution (TSD) & managing the various drivers (environmental & managerial) of TSD. By empowering an existing crop model with spatial information, it is possible to remove the grower/agronomist directly from the data analysis & the decision-making. Expert knowledge will be captured within the crop model, but there is no direct involvement between the spatial data & the end-users, removing this source of error and confusion. The spatial crop model is therefore a method for spatial data-fusion & value-adds to the original spatial data. The model provides a relatively simple integrated spatial output (recommended variable-rate management operations) that the grower can access for adoption. The modelling also allows estimates of uncertainty (as well as an operation) to assist growers in risk assessment with differential management. From an academic perspective, a few issues need to be researched & developed to achieve this. These include;1) Filling the knowledge gap on the amount (magnitude & spatial structure) of crop variability in potato fields. There are very few spatial studies available & this information is needed to correctly parameterise any spatial model within sensible boundary limits.2) Understanding the drivers of the observed variation in crop production. The variability observed can be linked to spatial information on soil & weather variations, as well as management decisions. This helps to inform the spatial model of the yield determinant factors.3) Development of a spatial meta-model. The spatial crop model relies on the output from an existing point crop model being used as an input into a spatial meta-model. The spatial meta-model is a new concept. It requires standardisation of inputs, particularly in regards their spatial footprint, correct parameterisation of neighbourhood interactions & correct modelling of the uncertainty at each point in the spatial model. Correct data processing & the knowledge from Points 1) & 2) above will ensure that the meta-model is correctly designed & populated. It will be validated against field experiments in the latter stages of the project.The project brings together leading UK industry expertise in potato production (SAC, SRUC, McCains), supply chains & processing (McCains), machinery for potato production (Grimme) & precision agricultural services (SE), as well as leading academic researchers in the area of precision agriculture (Newcastle Uni) & crop modelling (Newcastle Uni, Mylnefield Research Services). This consortium is well placed to deliver the project & deliver it to the needs of the industry.