A random forest model for the classification of wheat and rye leaf rust symptoms based on pure spectra at leaf scale.

The pure spectra acquisition of plant disease symptoms is essential to improving the reliability of remote sensing methods in crop protection. The reflectance values read from the pure spectra can be used as valuable training data for development of algorithms designed for plant disease detection at leaf and canopy scale. The aim of this paper is to identify and distinguish spectrally the leaf rust symptoms caused by two closely related special forms (f. sp.) of Puccinia recondita f. sp. tritici on wheat and Puccinia recondita f. sp. recondita on rye at leaf scale. Spectral measurements were made with FieldSpec 3 spectrometer in the wavelength range of 350-2500 nm. The spectrometer was connected to a microscope by optical fiber. Raw spectra of uredinia, chlorotic discoloration, green leaves, senescent inoculated leaves and senescent uninoculated leaves of wheat and rye, all of which obtained for this study, were investigated with a view towards making an automized classification of plant species and their phases. The created Random Forest models were tested separately using pure spectra, and from these vegetation indices were derived as predictors. Three vegetation indices, namely CRI, PRI and GNDVI, appeared to be the most robust in terms of distinguishing uredinia from other symptoms on rye and wheat leaves. PRI, EVI, NDVI705, and GNDVI were the most suitable for distinguishing uredinia, chlorotic discoloration, and green leaf stages on rye. That tusk on wheat leaves can be recognized if seven indices (PRI, MSAWI, SAVI, NDVI, NDVI705, GNDVI and RVI) are used together. For the classification of all disease symptoms for both plant species, the most useful were wavelengths in the VIS range: 431-436, 696-703 and 646-686 nm. However, the ranges of SWIR wavelengths (1938, 1955) and NIR wavelengths (1099-1104) also have a high contribution to the discrimination accuracy of the model. In the classification of all disease symptoms, the most important vegetation indices were CRI, OSAVI, and GNDVI. Analysis of the results revealed the advantage of the model based on the selected spectral wavelengths (Hit Rate of 96.6%) in comparison with predictions based on vegetation indices alone (Hit Rate of 91.7%). Both approaches show the highly applicable character of utilizing high quality spectral products such as satellite images in reducing operational costs of crop protection.

Machine learning modeling of plant phenology based on coupling satellite and gridded meteorological dataset.

Changes in the timing of plant phenological phases are important proxies in contemporary climate research. However, most of the commonly used traditional phenological observations do not give any coherent spatial information. While consistent spatial data can be obtained from airborne sensors and preprocessed gridded meteorological data, not many studies robustly benefit from these data sources. Therefore, the main aim of this study is to create and evaluate different statistical models for reconstructing, predicting, and improving quality of phenological phases monitoring with the use of satellite and meteorological products. A quality-controlled dataset of the 13 BBCH plant phenophases in Poland was collected for the period 2007-2014. For each phenophase, statistical models were built using the most commonly applied regression-based machine learning techniques, such as multiple linear regression, lasso, principal component regression, generalized boosted models, and random forest. The quality of the models was estimated using a k-fold cross-validation. The obtained results showed varying potential for coupling meteorological derived indices with remote sensing products in terms of phenological modeling; however, application of both data sources improves models' accuracy from 0.6 to 4.6 day in terms of obtained RMSE. It is shown that a robust prediction of early phenological phases is mostly related to meteorological indices, whereas for autumn phenophases, there is a stronger information signal provided by satellite-derived vegetation metrics. Choosing a specific set of predictors and applying a robust preprocessing procedures is more important for final results than the selection of a particular statistical model. The average RMSE for the best models of all phenophases is 6.3, while the individual RMSE vary seasonally from 3.5 to 10 days. Models give reliable proxy for ground observations with RMSE below 5 days for early spring and late spring phenophases. For other phenophases, RMSE are higher and rise up to 9-10 days in the case of the earliest spring phenophases.

The variability of PM10 and PM2.5 concentrations in selected Polish agglomerations: the role of meteorological conditions, 2006-2016.

The research focuses on the analysis of PM10 and PM2.5 concentrations variability at 11 stations in selected urbanized areas of Poland (Tricity, Poznan, Lódz, Kraków). Methods comprised: the analysis of basic statistical characteristics in yearly/monthly/daily/hourly scale and threshold exceedance frequencies. Also, correlations between PM10 and meteorological variables were investigated. GEV distribution analysis allowed the estimation of the return levels of monthly maxima of PM10 and PM2.5. Results show that in Tricity there are fewer than 5 % of days with PM10 and PM2.5 threshold exceedance. In Kraków, the standards are only met during summer and the frequency of daily PM limit exceedance in winter was around 65-90 %. GEV analysis indicates that 10y return level of PM10 monthly maximum daily average do not usually exceed 250 µg/m3 at most of the stations (Kraków agglomeration is an exception here). In winter, the meteorological conditions unfavourable to the pollutant's dispersion comprise: high-pressure systems, stable equilibrium in the atmosphere and limited turbulence occur quite often together with low wind speed and reduced height of the planetary boundary layer.