Fig 1 - uploaded by Kosuke Noborio
Content may be subject to copyright.
Decompositions of series A, H, R, and W (from left to right)
Source publication
To effectively protect plants from frost damage, an early alarm of frost can be helpful for growers. Frost is a localized phenomenon and can be quite variable across a small area, so predictive models developed with local data are preferred. As a climate phenomenon the occurrence of frost is closely related to multiple environment factors including...
Context in source publication
Context 1
... A, H, R, and W data are continuous values that behave as time series, while F is symbolic. The key idea of time series prediction is to capture varies self-correlation patterns hidden in a data series and try to extend the series to future so achieve prediction [15][16][17][18]. Fig. 1 shows additive decompositions of A, H, R, and W, in the period of 12/29/2016 17:08 to 1/7/2017 12:19 (total 12,672mins), where the horizontal scale is in days (1day = 1,440mins). The first three series show clearly seasonal components with strong self-correlation to the past, which makes possible for these series be extended to future ...
Similar publications
Selecting an evaluation metric is fundamental to model development, but uncertainty remains about when certain metrics are preferable and why. This paper introduces the concept of resolving power to describe the ability of an evaluation metric to distinguish between binary classifiers of similar quality. This ability depends on two attributes: 1. T...
A parameterization for the collision–coalescence process is presented based on the methodology of basis functions. The whole drop spectrum is depicted as a linear combination of two lognormal distribution functions, leaving no parameters fixed. This basis function parameterization avoids the classification of drops in artificial categories such as...
This paper proposes a numerical method to investigate the frictional characteristics of a fluid film journal bearing considering the elastohydrodynamic lubrication and the asperity contact force. We solved the average Reynolds equation by utilizing the FEM to determine the hydrodynamic force developed by the lubricant of the journal bearing. We als...
Report on model development in the project GlobalHydroPressure. Key characteristics of each case study in the project are briefly described. In all case studies, local models for assessing different hydrological pressures have been developed and evaluated. In this deliverable, besides a general description of the case study and its targeted hydrolo...
Traditional transit systems are usually composed of fixed routes and stops, which are suitable in densely populated areas. This paper presents a reformulation of the flexible transit model developed by Nourbakhsh and Ouyang (2012) to adapt it to many low demand cities in the world, especially those characterized by radial street patterns. Unlike tr...
Citations
... Diedrichs et al. [18] developed a component for an IoT-enabled frost prediction system, where they used machine learning algorithms trained by previous readings of temperature and humidity sensors to predict future temperatures. Ding et al. [19] propose the construction of predictive models using the support vector machine approach to capture possible causal relationships between several environmental factors and frost. Fuentes et al. [20] propose a neural network model, based on backpropagation, to predict the minimum air temperature of the following day from meteorological data using air temperature, relative humidity, radiation, precipitation, and wind direction and speed to detect the occurrence of radiative frost events. ...
A machine learning (ML)-based methodology for predicting frosts was applied to the southern and southeastern regions of Brazil, as well as to other countries including Uruguay, Paraguay, northern Argentina, and southeastern Bolivia. The machine learning model (using TensorFlow (TF)) was compared to the frost index (IG from the Portuguese: Índice de Geada) developed by the National Institute for Space Research (INPE, Brazil). The IG is estimated using meteorological variables from a regional weather numerical model (RWNM). After calculating the two indices using the ML model and the RWNM, a voting committee (VC) was trained to select between the computed outputs. The AdaBoost-Classifier algorithm was employed to implement the voting committee. The study area was subdivided into three distinct subregions: R1 (outside Brazil), R2 (the south of Brazil), and R3 (southeastern Brazil). Two forecasting time scales were evaluated: 24 h and 72 h. The 24 h forecasts from both approaches (TF and RWNM) exhibited a similar performance in terms of the number of accurate predictions. However, in the region covering Uruguay and northern Argentina, the TensorFlow model demonstrated superior frost prediction accuracy. Additionally, the TensorFlow model outperformed the RWNM for the 72 h forecast horizon.
... Diedrichs et al. [9] developed a component of an IoT-enabled frost prediction system, where they used 2 of 15 machine learning algorithms trained by previous readings of temperature and humidity sensors to predict future temperatures. Ding et al. [10] propose the construction of predictive models using the support vector machine approach to capture possible causal relationships between several environmental factors and frost. Fuentes et al. [12] propose a neural network model, based on a backpropagation type, to predict the minimum air temperature of the following day from meteorological data using air temperature, relative humidity, radiation, precipitation, and wind direction and speed to detect the occurrence of radiative frost events. ...
This is a preprint. The final version has been published in Meteorology and can be accessed here: https://doi.org/10.3390/meteorology4010006 .
A machine learning (ML)-based methodology for predicting frosts is applied to the southern and southeastern regions of Brazil, as well as to other countries including Uruguay, Paraguay, northern Argentina, and southeastern Bolivia. The machine learning model (using TensorFlow (TF)) was compared to the frost index IG (from the Portuguese: Índice de Geada) developed by the National Institute for Space Research (INPE, Brazil). The IG index is estimated using meteorological variables from a regional weather numerical model (RWNM). After calculating the two indices using the ML model and the RWNM, a voting committee (VC) was trained to select between the computed outputs. The AdaBoostClassifier algorithm was employed to implement the voting committee. The study area was subdivided into three distinct subregions: R1 (outside Brazil), R2 (the southern of Brazil), and R3 (the southeastern of Brazil). Two forecasting time scales were evaluated: 24 hours and 72 hours. The 24-hour forecasts from both approaches (TF and RWNM) exhibited similar performance in terms of the number of accurate predictions. However, in the region covering Uruguay and northern Argentina, the TensorFlow model demonstrated superior frost prediction accuracy. Additionally, the TensorFlow model outperformed the RWNM for the 72-hour forecast horizon.
... From expert users, we obtained relevant information about the specific needs of the zone being analyzed. For example, they started from the assumption that temperatures have a strong influence on frosts, as described in some works in the literature [12,22]. However, expert users were not aware of the specific influences of other weather factors, such as wind and rain. ...
... However, they also knew that this is not always the case, and it depends on the wind speed and rainfall. Even in the related works presented in [22,24], the authors analyzed the influences of these factors without finding strong relations or conclusive results that influence frost. ...
... On the one hand, we compared our case studies to related works in Table 1. Like the works [14,[22][23][24], we defined a main objective to analyze the influence of weather variables on late frosts. We used similar weather variables as relevant features, concluding that temperature was the most influential factor for frost occurrence. ...
The large amount of available data, generated every second via sensors, social networks, organizations, and so on, has generated new lines of research that involve novel methods, techniques, resources, and/or technologies. The development of big data systems (BDSs) can be approached from different perspectives, all of them useful, depending on the objectives pursued. In particular, in this work, we address BDSs in the area of software engineering, contributing to the generation of novel methodologies and techniques for software reuse. In this article, we propose a methodology to develop reusable BDSs by mirroring activities from software product line engineering. This means that the process of building BDSs is approached by analyzing the variety of domain features and modeling them as a family of related assets. The contextual perspective of the proposal, along with its supporting tool, is introduced through a case study in the agrometeorology domain. The characterization of variables for frost analysis exemplifies the importance of identifying variety, as well as the possibility of reusing previous analyses adjusted to the profile of each case. In addition to showing interesting findings from the case, we also exemplify our concept of context variety, which is a core element in modeling reusable BDSs.
... Among the most recent methodologies used to climate prediction and specifically for frost events prediction, supervised machine learning techniques such as artificial neural network (Diedrichs et al., 2018;Latif et al., 2020), decision tree (Lee et al., 2016), random forest (Diedrichs et al., 2018) and support vector machine (Ding et al., 2019) have been used. These techniques have been used to identify when a reduction in temperature below 0°C is going to occur, and to predict the temperature behavior and the minimum value that temperature will reach. ...
... The hours of the day for which the SelectKBest function detected the highest score were for temperature between 02:00 and 10:00, RH between 10:00 and 16:00, DP between 01:00 and 04:00 and PAR between 9:00 and 16:00.Otherwise,Lee et al. (2016) established a positive relationship between DP and RH with frost occurrence probability. In addition,Ding et al. (2019) found that air temperature, net radiation and wind speed had a negative correlation with frost(-0.37, -0.2, and -0.28, respectively) and positive correlation with relative humidity (0.35), with a data frequency of 60 data per hour (one record per minute). ...
... -0.2, and -0.28, respectively) and positive correlation with relative humidity (0.35), with a data frequency of 60 data per hour (one record per minute). Other authors have considered wind speed and minimum temperature for frost prediction because when speed exceeds the very low threshold value, it prevents the formation of a thermal inversion layer near the ground(Ding et al., 2019;Fuentes et al., 2018). However,Ding et al. (2019) have found lower correlation values between wind speed and frost events. ...
In the tropic, many crops are distributed in the highlands of provinces of the Andean regions at heights of 2,500 m asl and constitute the areas with the highest susceptibility to the frost events occurrence. The study objective was to propose an early frost prediction model based on the relationships between frost events and climatic variables, modeled with machine learning methods. The climatic variables were obtained from thirteen meteorological stations located inside flower crops and distributed in nine municipalities of the Cundinamarca Department. The variables registered were temperature, relative humidity, dew point, photosynthetically active radiation, and precipitation, entered as explanatory variables of frost events. The metrics used for predictive performance evaluation of the five machine learning methods examined were precision, recall, true negative rate, accuracy, and F1 score. The variables’ climatic behavior of previous hours to a frost event are low humidity, wind speed and cloudiness, and high thermal radiation. The fourth of the five trained models performed well due to their classification evaluation metrics, greater than 91%. The cross-validation and statistical analysis demonstrated the higher accuracy of the GBDT model on frost events detection.
Keywords
forecast; artificial neural networks; gradient boosting; climatic variables
... A forecast horizon is helpful due to the information that could be brought by a system before a possible frost event [5]. Minimum night temperature estimations are applied to obtain this information using traditional-empirical methods [7] to deterministic equations [8] and machine learning methods [9]. The application of machine learning methods could be divided into classification and regression tasks. ...
... In [9], [10] it is indicated that the use of wind-related variables can be problematic in some datasets due to abrupt changes that cannot be represented at an adequate resolution. Because of this, along with the cost and lack of access to the instrumentation within the context of the project, led to discard wind-related variables. ...
... Addressing this complexity requires the use of advanced computer models, field observations, and an understanding of meteorological patterns-aspects that have garnered the attention of researchers from a wide range of scientific disciplines [25], [26]. Most research regarding frost prediction relies on simulating partial differential equations or conventional statistical models to anticipate weather conditions. ...
Precision Agriculture (PA), also known as Smart Farming, has emerged as an innovative solution to address contemporary challenges in agricultural sustainability. A particular sector within PA, Precision Viticulture (PV), is specifically tailored for vineyards. The advent of the Internet of Things (IoT) has facilitated the acquisition of higher-resolution meteorological and soil data obtained through in situ sensing. The integration of Machine Learning (ML) with IoT-enabled farm machinery stands at the forefront of the forthcoming agricultural revolution. This data allows ML-based forecasting as an alternative to conventional approaches, providing agronomists with predictive tools essential for improved land productivity and crop quality. This study conducts a thorough examination of vineyards with a specific focus on three key aspects of PV: mitigating frost damage, analyzing soil moisture levels, and addressing grapevine diseases. In this context, several ML-based models are proposed in a real-world scenario involving a vineyard located in southern Italy. The test results affirm the feasibility and efficacy of the ML models, demonstrating their potential to revolutionize vineyard management and contribute to sustainable agricultural practices.
... All analyzed stations in the studied area, except for Nehbandan, which generally does not experience frost days, were characterized using the firstorder Markov chain, indicating that frost days depend on past weather conditions. Ding et al. [13] predicted the possibility of future frost with an SVM model using the historical values of temperature, humidity, and radiation. Temperature is a key factor in frost prediction models, with humidity helping generate an early warning for a relatively long period, such as within 2 or 3 h. ...
Regional accuracy was examined using extreme gradient boosting (XGBoost) to improve frost prediction accuracy, and accuracy differences by region were found. When the points were divided into two groups with weather variables, Group 1 had a coastal climate with a high minimum temperature, humidity, and wind speed and Group 2 exhibited relatively inland climate characteristics. We calculated the accuracy in the two groups and found that the precision and recall scores in coastal areas (Group 1) were significantly lower than those in the inland areas (Group 2). Geographic elements (distance from the nearest coast and height) were added as variables to improve accuracy. In addition, considering the continuity of frost occurrence, the method of reflecting the frost occurrence of the previous day as a variable and the synthetic minority oversampling technique (SMOTE) pretreatment were used to increase the learning ability.
... When Frost occurs during reproductive and vegetative growth, it can affect seedling survival rate from medium to extreme (Barlow et al. 2015). Reports in December 2018 represented rising frost injury to many trees in vast areas of Central Europe due to global warming (Ding, Noborio, and Shibuya 2019). Fresh Plaza is a website that shares many news and articles on a global scale in agriculture. ...
Frostbite and frost is one of the problems that endanger the health of crops and can ruin plants and fruits. Soil temperature is the most significant factor that influences the freezing depth. Therefore, monitoring and predicting this characteristic is crucial for frostbite protection. This study aims to predict soil temperature on cold days to prevent frostbite injury in crops. For this matter, we used the registered and logged hourly data by the HOBO U30 data logging device and predicted the soil temperature from air temperature, soil water content, and relative humidity. We used 80% of the data set for the training data and assigned the other 20% to the test data. RMSE and MSE were two of the evaluation criteria of the neural network in this study. Also, we calculated P-value and T-value for statistical hypothesis testing. In another approach for weighting the neural network, we used evolutionary algorithms such as Genetic Algorithm and Particle Swarm Optimization instead of the gradient-based methods. According to the results, Multi-layer perceptron neural network with the respective values 0.082 and 0.0068 for RMSE and MSE in training data and 0.085 and 0.0073 for RMSE and MSE in testing data proved to have a better performance in the soil temperature prediction compared to the ANN-GA and ANN-PSO models. Farmers, botanical researchers, and policymakers in food security can use these results.
... The minimum temperature is generally considered to be the most important factor affecting frost occurrence [57,58]. In previous studies, scientists used the threshold and duration of the minimum temperature to assess the severity of frost [16,57]. ...
Tea trees are the main economic crop in Zhejiang Province. However, spring cold is a frequent occurrence there, causing frost damage to the valuable tea buds. To address this, a regional frost-hazard early-warning system is needed. In this study, frost damage area was estimated based on topography and meteorology, as well as longitude and latitude. Based on support vector machine (SVM) and artificial neural networks (ANNs), a multi-class classification model was proposed to estimate occurrence of regional frost disasters using tea frost cases from 2017. Results of the two models were compared, and optimal parameters were adjusted through multiple iterations. The highest accuracies of the two models were 83.8% and 75%, average accuracies were 79.3% and 71.3%, and Kappa coefficients were 79.1% and 67.37%. The SVM model was selected to establish spatial distribution of spring frost damage to tea trees in Zhejiang Province in 2016. Pearson’s correlation coefficient between prediction results and meteorological yield was 0.79 (p < 0.01), indicating consistency. Finally, the importance of model factors was assessed using sensitivity analysis. Results show that relative humidity and wind speed are key factors influencing accuracy of predictions. This study supports decision-making for hazard prediction and defense for tea trees facing frost.
... Later year in the field, Shibuya et al. (2020) detected dew, frozen dew, and hoarfrost under different weather conditions using TDR techniques. With time-series data on environmental factors and TDR-measured hoarfrost, Ding et al. (2019) reported that machine learning techniques enabled hoarfrost forecast three-hour in advance. ...
Time domain reflectometry (TDR) is the most widely used non-destructive method to determine the water content of soils and other porous media. TDR equipment can be automated and multiplexed to acquire accurate and rapid waveforms (return signal) without safety concerns associated with radioactive methods (e.g., neutron probe and Gamma-ray probe). Two key steps are required for TDR applications: (1) Obtain and analyze TDR waveforms using travel-time and signal attenuation analysis to determine dielectric permittivity and electrical conductivity, respectively. (2) Calibrate to determine a new- or apply an existing (e.g., Topp et al. (1980)) relationship between the derived soil dielectric permittivity and the volumetric water content of the porous medium of interest. A majority of researchers and practitioners focus on step two and additionally develop new mathematical models to get better estimates of water content. Although there are reviews of TDR principles and applications in soil science, there is a lack of information on how TDR can disclose critical information in porous media beyond average soil water content. Therefore, we present a newly expanded review of TDR applications in porous media including soils, plants, snow, food stuffs, and concrete. We begin by reviewing TDR basics, including principles, probe design, commercially available equipment, and graphical and numerical methods as well as available software for waveform analysis. Applications of TDR to estimate volumetric water content in various types of porous media, the latest techniques available to derive spatial variability of soil water distributions along a single TDR probe are included, followed by TDR waveform based analyses to estimate electrical conductivity (EC), wetting/drying and freezing/thawing fronts, and snow depth. The combination of TDR measurements coupled with other methods (e.g., gypsum/ceramics and heat pulse method) to determine a wide range of soil physical properties (e.g., soil water retention curve, thermal properties, and hydraulic conductivity) and fluxes (e.g., soil heat flux, liquid water flux, and vapor flux) are also included. The study concludes with a discussion of limitations and future perspectives on various TDR applications.
