Based on the collected weather data from the agricultural Internet of Things (IoT) system, changes in the weather can be obtained in advance, which is an effective way to plan and control sustainable agricultural production. However, it is not easy to accurately predict the future trend because the data always contain complex nonlinear relationship with multiple components. To increase the prediction performance of the weather data in the precision agriculture IoT system, this study used a deep learning predictor with sequential two-level decomposition structure, in which the weather data were decomposed into four components serially, then the gated recurrent unit (GRU) networks were trained as the sub-predictors for each component. Finally, the results from GRUs were combined to obtain the medium- and long-term prediction result. The experiments were verified for the proposed model based on weather data from the IoT system in Ningxia, China, for wolfberry planting, in which the prediction results showed that the proposed predictor can obtain the accurate prediction of temperature and humidity and meet the needs of precision agricultural production.

No abstract

The global ionospheric storm in March 2015 was investigated by using data from over 3000 GPS stations worldwide. In this study, total electron content (TEC), rate of TEC (ROT), and ROT's standard deviation rate of the TEC index, as well as the second‐order difference operator TECT, were considered as main characteristic methods to distinguish ionosphereic disturbances. The results show that (1) based on the multiple methods above, we all observed that for the first time, there were three equatorward traveling ionospheric disturbances (TIDs) in the main phase of this storm. In North America, the disturbance zone expanded to ~40°N; the disturbance periods and AE peak stages were roughly synchronous. We suggest that these three TIDs were induced by the propagation of atmospheric gravity waves to low latitudes under the action of AE. (2) The most intense positive storm occurred over South America and the South Atlantic (over 300% enhancement; 00:00–05:00 UT on 18 March), whereas a negative storm was observed in the corresponding region of the Northern Hemisphere. Such inverse hemispheric asymmetry in intensity and structure can be explained by the variations of the thermospheric composition, the IMF By component, and the geomagnetic intensity. (3) On 18 March, a negative storm dominated globally (except at certain low latitudes), and tended to propagate equatorward and decay with time, which could be largely attributed to the storm circulation theory. And the evolution of the negative storm was further characterized by the foF2 variations of ionosondes.

The three-dimensional computerized ionospheric tomography (3DCIT) technique is used to reconstruct the spatial distribution of storm-enhanced density (SED) based on the global positioning system total electron content measurements over the North American area during the 17 March 2013 storm. The reconstruction results are carefully validated with observations from three ionosonde stations, the constellation observing system for meteorology, ionosphere, and climate (COSMIC) radio occultations, and the Millstone Hill incoherent scatter radar. The electron density profiles from the 3DCIT reconstruction show a good agreement with the ionosonde and COSMIC electron density profiles. The 3DCIT-derived electron density difference between the storm day of 17 March and the quiet day of 16 March also captures the similar SED plume signature that was observed by the Millstone Hill incoherent scatter radar. The 3DCIT reconstruction allows us for the first time to unveil the 3-D configuration of the SED plume and its spatiotemporal evolution. It was found that the SED plume first appeared around 400 km and then expanded downward to~300 km as well as upward to~500 km over the course of a 3-hr period from 19 to 22 UT on 17 March. Our study also showed that the density enhancement within the SED plume occurred mostly above the storm time F layer peak height.