Snow depth (SD) exhibits high spatiotemporal heterogeneity in Western Himalaya (WH), and its knowledge is essential for applications related to water resources, disaster management, climate, etc. However, due to inclement weather and rugged topographical conditions, only a sparse network of SD monitoring stations exists in WH. Spaceborne passive microwave (PMW) remote sensing datasets provides valuable information about SD; however, only a limited PMW SD studies that cover subregions of WH are available. The current study utilizes Extremely Randomized Trees (ERT) based machine learning technique to estimate daily SD over the entire WH region. The ERT SD model is developed using PMW brightness temperature datasets from Advanced Microwave Scanning Radiometer-2 (AMSR-2), snow cover duration (SCD), and other auxiliary parameters during the winter period between 2012-13 and 2019-20. The data between 2012-13 and 2017-18 is used for training the model, whereas the data between 2018-19 and 2019-20 is used for testing the model. The results demonstrate: (a) The ERT SD model has shown improved SD estimates compared to the available PMW remote sensing-based operational SD products and empirical PMW SD models. (b) in general, with an increase in SD, the mean absolute error of SD retrievals has increased in all SD products/models. (c) Unlike the operational AMSR2 SD product, and Northern Hemisphere Machine Learning SD product, the ERT SD model retrievals have shown better consistency with MODIS snow cover. (d) the developed model has shown a wider range in SD retrievals as compared to other products considered in this study.

The use of accurate streamflow estimates is widely recognized in the hydrological field. However, due to the model’s structural error, they often yield suboptimal streamflow estimates. Past studies have shown that soil moisture assimilation improves the performance of the hydrological model which often results in enhanced model estimates. Due to this reason, it is widely studied in the hydrological field.  However, the efficiency of the assimilation largely relies on the correct placement of the observation into the model. Ingesting futile observations often results in the degradation of model performance. On the contrary, performing assimilation only at those time steps when the assimilating variable is sensitive to the model output may yield desirable output. Further, it will avoid the assimilation of spurious observations. In this view, this study proposes a new approach where sensitivity-based sequential assimilation is performed on a conceptual Two Parameter Model (TPM). To demonstrate this approach, ASCAT soil moisture observations are assimilated into TPM using Ensemble Kalman Filter (EnKF) sequential approach. At first, the temporal evolution of the soil moisture sensitivity with respect to streamflow is established. Later, at those time steps when the soil moisture is sensitive, EnKF assimilation is performed. For this purpose, a moderately sized catchment in the Krishna basin, India is selected as the study area. Model calibration and validation are performed between 2000 to 2006 and 2007 to 2011 respectively. Model run without assimilation is considered as open-loop simulation. Streamflow simulation after assimilation showed a significant improvement when compared against the open-loop simulation. KGE value increased from 0.70 to 0.79 and PBIAS value reduced from 18.31 to 1.80. The highlighting factor is that only 39% of the total observations were used during the assimilation process. The initial results are encouraging and looks that the proposed approach shall be highly useful at those locations where data availability for assimilation purpose is a serious concern.  

The need for and the use of different data assimilation techniques to improve the quality of streamflow forecast is now well established. In this paper, the goal is to demonstrate the power of a new class of methods known as the Forward Sensitivity Method (FSM) which is based on the temporal evolution of model sensitivities with respect to the control variables consisting of initial conditions and parameters. FSM operates in two phases: The first phase provides a simple algorithm for placing observations at or near where the square of forward sensitivities attains their maximum values. Using only this selected subset of observations in a weighted least squares method, the second phase then provides an estimate of the unknown elements of the control variables. In this paper, FSM based assimilation is applied to a simple class of two parameter model in a medium-sized agriculture dominant watershed lying in the Krishna River Basin, India. Four assimilation scenarios were tested to determine the effect of assimilating only sensitive observations as well as the impact of temporally evolving initial condition sensitivity. Sensitivity results showed that observations during the monsoon time alone are enough for assimilation purposes, which has helped in reducing the computational time greatly. Assimilation and forecast results also indicated that the scenarios which assimilated only sensitive observations are better in estimating daily streamflow. From the obtained results, it is concluded that FSM based assimilation has significant potential to improve the streamflow simulations, especially in places where data availability remains a major challenge.

The accuracy of streamflow forecasts is important for efficient monitoring and mitigation of flood events. Unfortunately, the uncertainty in the model control variable which includes model parameters, initial and boundary conditions, propagates through the model, resulting in the degradation of streamflow forecast. Various studies in the past have shown the potential of soil moisture assimilation in hydrological models resulting in the improved forecast. Further, the efficiency of assimilation is based on the number and the distribution of observations used. This study proposes a new approach called Forward sensitivity method (FSM), which operates in two phases. By running the model and forecast sensitivity dynamics forward in time, the first phase places the observations at or near where the square of the forecast sensitivity with respect to the control takes maximum values. Then using only this subset of observations, the second phase estimates the unknown elements of the control by solving a resulting weighted least squares problem. The power of this approach is demonstrated by assimilating ASCAT soil moisture observations into a conceptual Two Parameter Model in a medium sized watershed lying in the Krishna River Basin, India. The model run extends for four monsoon years from June 2007 to June 2011 and two assimilation scenarios were tested. The first scenario uses all the observations, whereas, the second uses only sensitive observations during assimilation and the results were then compared against open loop simulation (model run without assimilation). Sensitivity results indicate that observations during monsoon time alone are sufficient for assimilation purpose, which accounts for only 37.42 percent of total observations. Also, the estimation and forecast results show improved streamflow performance when using only sensitive observations. From the results, it is concluded that FSM based assimilation can help in reducing the computation time greatly. Further, this study will be critically helpful in the places where data availability remains a major problem.

High-resolution sea surface temperature (SST) estimates are dependent on satellite-based infrared radiometers, which are proven to be highly accurate in the past decades. However, the presence of clouds is a big stumbling block when physical approaches are used to derive SST. This problem is more prominent across tropical regions such as Arabian Sea(AS) and Bay of Bengal(BoB), restricting the availability of high-resolution SST data for ocean applications. The previous studies for developing daily high-resolution cloud-free SST products mainly focus on fusion of multiple satellites and in-situ data products that are computationally expensive and often time consuming. At the same time, it was observed that the capabilities of data-driven approaches are not yet fully explored in the estimation of cloud-free high-resolution SST data. Hence, in this study an attempt has been made for the first time to estimate daily cloud free SST from a single sensor (MODIS Aqua) dataset using advanced machine learning techniques. Here, three distinct machine learning techniques such as Artificial Neural Networks (ANN), Support Vector Regression (SVR) and Random Forest (RF)-based algorithms were developed and evaluated over two different study areas within the AS and BoB using 10 years of MODIS data and in-situ reference data. Among the developed algorithms, the SVR-based algorithm performs consistently better. In AS region, while testing, the SVR-based SST estimates was able to achieve an adjusted coefficient of determination (R_adj^2) of 0.82 and root mean square error (RMSE) of 0.71°C with respect to the in situ data. Similarly, in BoB too, the SVR algorithm outperforms the other algorithms with R_adj^2 of 0.78 with RMSE of 0.88ºC. Further, a spatio-temporal and visual analysis of the results as well as an inter-comparision with NOAA AVHRR daily optimally interpolated global SST (a standard SST product available in practice) the suggest that the proposed SVR-based algorithm has huge potential to produce operational high-resolution cloud-free SST estimates, even if there is cloud cover in the image.