Description: Socioeconomic livelihoods in the Horn of Africa (HA) are highly dependent on seasonal rainfall, which occurs during two main seasons: October-November-December (OND) and March-April-May (MAM). During the two last decades the HA region has been affected by severe and prolonged droughts, leading to acute food insecurity, shortage of drinking water, and increasing risk of disease. Sub-seasonal drought prediction over the HA, from two weeks to two months, is therefore crucial for decision making and early warnings across several sectors. The sub-seasonal prediction of high and low precipitation extremes (PEs) by dynamical forecast systems is challenging for both rainy seasons, but there may be potential for extending the current prediction timescale based on remote drivers. To investigate the sub-seasonal predictability of PEs during the OND season we build a Long Short-Term Memory (LSTM) Neural Network predicting biweekly precipitation tercile categories over the HA region. The LSTM is trained on observational and reanalysis data during the period 1981—2020 and provides predictions with lead times of one week to one month. The results show that floods can be more skillfully predicted than droughts for all lead times. Moreover, we use explainable AI methods to explore the contribution of remote drivers to the predictions and potential sub-seasonal forecast opportunities for PEs. Preliminary results show that the sea surface temperature over the tropical Pacific is important for the LSTM prediction, but further investigation is needed to determine more factors affecting the prediction skill for PEs over the HA region.

Description: Monthly precipitation prediction is of great significance to bridge the gap between short-term weather forecast and seasonal forecast. However, due to the complexity of the climate system, there is still a great deal of uncertainty in the prediction of precipitation at the monthly scale. In order to reduce the uncertainty of the monthly precipitation predictions results，we analyzed the corrective effect of BJP (Bayesian Joint Probabilistic ) and EMOS (Ensemble Model Output Statics) model on precipitation bias prediction using S2S (sub-seasonal to seasonal scale) near-real-time overall forecast data and re-forecast data from five global data centers, including ECMWF, NCEP, UKMO, JMA and KMA. According to the advantages of different models the integrated multi-structure monthly precipitation prediction model was constructed based on EM (Expectation-Maximum) algorithm. The prediction application of monthly precipitation was carried out during the flood season of 1981~2010 in the middle and lower reaches of the Yangtze River, and the result showed as follows. (1)The monthly precipitation prediction results of the ensemble model showed that the Nash efficiency coefficient reached more than 0.6 and the correlation coefficient was 0.83. It indicates that the prediction sequence and the measured sequence of the ensemble model had good consistency. (2) The average relative deviation of the monthly precipitation prediction results is 25%, which effectively reduces the uncertainty of the monthly precipitation prediction results of a single model compared with the monthly precipitation forecast results of a single model. The results provide scientific support for improving the accuracy of drought or flood prediction.

Description: Various paleoclimate records show that the end of interglacials of the late Pleistocene was marked by abrupt cooling events. Strong abrupt cooling occurring when climate was still in a warm interglacial condition is puzzling. Our transient climate simulations for the eleven interglacial (sub)stages of the past 800,000 years show that, when summer insolation in the Northern Hemisphere (NH) high latitudes decreases to a critical value (a threshold), it triggers a strong, abrupt weakening of the Atlantic meridional overturning circulation and consequently an abrupt cooling in the NH. The mechanism involves sea ice-ocean feedbacks in the Northern Nordic Sea and the Labrador Sea (Yin et al., 2021, doi: 10.1126/science.abg1737). The insolation-induced abrupt cooling is accompanied by abrupt changes in precipitation, vegetation from low to high latitudes and in particular by abrupt snow accumulation in polar regions. The timing of the simulated abrupt events at the end of interglacials is highly consistent with those observed in marine and terrestrial records, especially with those observed in high-resolution, absolutely-dated speleothem records in Asia and Europe, which validates the model results and reveals that the astronomically-induced slow variations of insolation could trigger abrupt climate events. Our results show that the insolation threshold occurred at the end of each interglacial of the past 800,000 years, suggesting its fundamental role in terminating the warm climate conditions of the interglacials. The next insolation threshold will occur in 50,000 years, implying an exceptionally long interglacial ahead.

Description: Understanding the sea ice variability and the mechanisms involved during warm periods of the Earth is essential for a better understanding of the sea ice changes at the present and in the future. Based on simulations with the model LOVECLIM, this study investigates the sea ice variations during the last nine interglacials and focuses on the inter-comparison between interglacials as well as their differences from the present and future. The results show that the annual mean Arctic sea ice variation is primarily controlled by local summer insolation, while the annual mean Southern Ocean sea ice variation is more influenced by the CO〈sub〉2〈/sub〉 concentration but the effect of local summer insolation can’t be ignored. As compared to the present, the last nine interglacials all have much less sea ice in the Arctic annually and seasonally due to high summer insolation. They also have much less Arctic sea ice in summer than the double CO〈sub〉2〈/sub〉 experiment, which makes to some degree the interglacials possible analogues for the future in terms of the changes of sea ice. However, compared to the double CO〈sub〉2〈/sub〉 experiment, the interglacials all have much more sea ice in the Southern Ocean due to their much lower CO〈sub〉2〈/sub〉 concentration, which suggests the inappropriateness of considering the interglacials as analogues for the future in the Southern Ocean. Our results suggest that in the search for potential analogues of the present and future climate, the seasonal and regional climate variations should be considered.

Description: The rapid convergence of PPP-RTK depends on the ionospheric correction including the accuracy and prior information. However, the traditional grid-based ionospheric model often uses fixed ionospheric prior information without taking into account the spatiotemporal diversity of the ionosphere, thus weakening the performance of PPP-RTK and limiting its application scenarios. In this study, a self-validation grid-based ionospheric model is proposed to improve the performance of PPP-RTK. A grid-based slant ionospheric model adapted to multi-scale networks is developed first with the careful consideration of receiver DCBs. Additionally, the ionosphere residuals obtained by self-validation of each reference station are assigned to the regional area based on distance and time, providing more reasonable ionospheric prior information for PPP-RTK. Then, PPP-RTK can achieve fast convergence with more reasonable ionospheric prior information. Experiments conducted under different ionospheric conditions demonstrate that the modified model significantly improves both the positioning accuracy and convergence time of PPP-RTK. During the ionosphere calm period, the average convergence time is reduced from 15.4s to 4.1s and the positioning accuracy is improved by 29.96% compared with the traditional grid model. Furthermore, during the ionosphere active period, the positioning accuracy is improved from (0.08, 0.10, 0.39) m to (0.05, 0.04, 0.12) m, with the improvement of 37.50%, 60.00%, 69.23% in the east, north and up directions, respectively.

Description: The accurate knowledge of tropospheric delay plays a key role in the rapid precise positioning. However, the tropospheric delay map with high spatial and temporal resolutions cannot be obtained from Global Navigation Satellite System (GNSS) observations alone, which hinders its further improvement of positioning performance, especially under complex terrains. In this study, a tropospheric model which integrates Global Forecast System (GFS) analysis and forecast grids, GNSS tropospheric delays, and multi-source meteorological data is established based on the four-dimensional variational assimilation method, in order to augment the capabilities of GNSS PPP-RTK under the circumstance of mountainous areas. The proposed tropospheric model is capable of predicting tropospheric delays with a spatial resolution of less than 5 km. The accuracy of tropospheric delays derived from this model is evaluated by comparing with the ERA5 products. Also, the performance of the tropospheric model augmented PPP-RTK is investigated in terms of the success rate of ambiguity fixing, convergence time, and positioning accuracy. The results show that the proposed tropospheric model exhibits excellent capabilities in improving the precision of tropospheric delays as well as the PPP-RTK positioning performance.

Description: Ocean surface winds are critical for shaping the Earth's weather and climate, and the CYGNSS mission, launched in 2016, is designed to monitor ocean wind. This study presents a novel deep learning model, CNN-LSTM, which retrieves ocean wind speed using CYGNSS observations. The model uses a Convolutional Neural Network (CNN) module to extract spatial features from a two-dimensional matrix of the delay-Doppler Map (DDM) around the Specular Point (SP) and a Long Short-Term Memory (LSTM) module to capture temporal features from a time series. The model's performance is evaluated against the fifth generation European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis (ERA5) products, and error analysis is performed to demonstrate the model's robustness at both spatial and temporal scales. Additionally, the study proposes the segmental modeling and calibration to address the issue of wind speed underestimation at high wind speeds. The study demonstrates the effectiveness and feasibility of the CNN-LSTM model in providing efficient processing of GNSS reflectometry observations and accurate global-scale ocean wind speed retrieval.

Description: As one of the most significant circulation systems over the Northern Hemisphere in the cold season, the East Asian winter monsoon (EAWM) has been broadly investigated from the seasonal-mean perspective, while subseasonal variations in the EAWM still remain ambiguous. Based on Season-reliant Empirical Orthogonal Function (S-EOF) analysis, this study shows that the subseasonal strength reversal of the EAWM (SR-EAWM), featuring a weaker (or stronger) EAWM in early winter (December) and a stronger (or weaker) EAWM in late winter (January-February), is a distinct leading mode of the month-to-month variation of the EAWM. The weak-to-strong SR-EAWM is characterized by an anomalous low over Eurasia and a weakened East Asian major trough (EAT) in early winter, with an intensified Siberian High and a deepened EAT in late winter. The SR-EAWM is preceded by surface air temperature anomalies over Davis Strait (DST) and those over central-eastern North America (CENAT) in September-October. The DST mainly influences the SR-EAWM in early winter through a “sea ice bridge” of the November Baffin Bay sea ice concentration anomaly (BBSIC). The BBSIC could intensify the DST in December by altering surface heat flux, thus exciting a downstream atmospheric response and modulating the strength of the EAT in early winter. The CENAT affects the SR-EAWM in late winter by inducing an “ocean bridge” of the western North Atlantic sea surface temperature anomaly (WNASST). The WNASST can persist into late winter and then significantly affects the SR-EAWM by regulating Eurasian circulation anomalies and the downstream EAT. The bridge roles of the BBSIC and WNASST can be further verified by a linear baroclinic model. Finally, two physical-empirical models are established using the DST/BBSIC and the CENAT indices, respectively. Both exhibit promising prediction skills. The results highlight that the DST, BBSIC, and CENAT are crucial predictability sources for the SR-EAWM.

Description: The wet tropospheric correction (WTC) retrieved from the onboard calibration microwave radiometer (CMR) of Haiyang-2A (HY-2A) is critical in monitoring the global sea level. However, the CMR WTC became significantly biased from June 2017 due to the failure of the 18.7-GHz band, which caused massive errors in the sea surface height (SSH) measurements. We investigate the accuracy of the CMR WTC derived from the two remaining bands to address this problem. A comprehensive evaluation using multisource data demonstrates that the dual-band + backscattering coefficient (BC) algorithm achieves comparable accuracy to the three-band algorithm, and it does not suffer from any large errors when the equipment works well. Hence, we calibrated the HY-2A CMR data with the dual-band + BC algorithm when the 18.7-GHz band failed, and the accuracy of the CMR WTC is improved from 2.34 to 1.39 cm compared with European Center for Medium-Range Weather Forecasts (ECMWF) ERA5 data. In addition, the SSH measurements are improved significantly by a maximum of 2 cm in mean value using the dual-band + BC WTC during the failure period of HY-2A CMR. Compared with Jason-3 SSH measurements, the HY-2A with dual-band + BC shows a slightly larger difference than HY-2A with three-band by 0.1 cm in rms. This method prolongs the operational lifetime of the HY-2A CMR and could be used in the reprocessing of HY-2A observations.

Description: Traditional methods for precipitation nowcasting such as optical flow methods give prediction results by extrapolating the radar echo maps, which doesn’t consider well the generation and disappearance of convection. With the development of deep learning, extrapolating the radar echo maps using deep learning methods has become an important way to predict precipitation. The Global Navigation Satellite System (GNSS) can accurately sense water vapor with high temporal resolutions and short latency, which makes it possible to improve the accuracy of precipitation nowcasting by integrating GNSS and radar observations. In this study, we choose GNSS stations in the German region and make use of the radar products provided by the German Weather Service (DWD) to give a precipitation prediction with deep learning methods. The U-Net, which takes four consecutive radar composite grids and GNSS zenith total delays (ZTDs) as separate input channels (t-15, t-10, t-5 minutes, and t, where t is the time of the nowcast) to produce a nowcast at time t+5 minutes. Verification experiments on several rainfall events are carried out, and the results suggest that the proposed GNSS and radar fusion model reveals an enhanced accuracy of precipitation nowcasting when compared to the conventional method based on radar echo maps only, when the routine verification metrics of Mean Absolute Error (MAE) and Critical Success Index are taken into consideration.