No abstract

The study proposes Secondary Precipitation Estimate Merging using Machine Learning (SPEM2L) algorithms for merging multiple global precipitation datasets to improve the spatiotemporal rainfall characterization. SPEM2L is applied over the Krishna River Basin (KRB), India for 34 years spanning from 1985 to 2018, using daily measurements from three Secondary Precipitation Products (SPPs). Sixteen Machine Learning Algorithms (MLAs) were applied on three SPPs under four combinations to integrate and test the performance of MLAs for accurately representing the rainfall patterns. The individual SPPs and the integrated products were validated against a gauge-based gridded dataset provided by the Indian Meteorological Department. The validation was applied at different temporal scales and various climatic zones by employing continuous and categorical statistics. Multilayer Perceptron Neural Network with Bayesian Regularization (NBR) algorithm employing three SPPs integration outperformed all other Machine Learning Models (MLMs) and two dataset integration combinations. The merged NBR product exhibited improvements in terms of continuous and categorical statistics at all temporal scales as well as in all climatic zones. Our results indicate that the SPEM2L procedure could be successfully used in any other region or basin that has a poor gauging network or where a single precipitation product performance is ineffective.

No abstract

No abstract

Accurate estimation and prediction of drought events are highly essential for implementing effective planning and management strategies to handle this complex natural phenomenon. Application of machine learning algorithms (MLAs) for integrating satellite precipitation products (SPPs), unlike gauge observations, can furnish precise drought estimations. In this study, we have proposed and tested two approaches (pre and post‐integration of SPPs) that deal with the prediction of drought employing 13 MLAs. Three SPPs are integrated under four combinations (involves two and three datasets integration) employing pre and post integration approaches to predict Standardized Precipitation Index and Standardized Precipitation Evapotranspiration Index at various temporal scales (1, 3, 6 and 12‐month). From the overall results, Approach‐2 that involves estimation of drought before integration using MLAs proved effective than Approach‐1 (prediction of drought post‐integration). Neural Networks based Bayesian Regularization (NBR) under three dataset integration outperformed at all temporal scales and climatic zones of India when compared to the other 12 MLAs and two dataset integration combinations. The blended product (NBR) manifested enhancements in statistical results at all temporal scales and climatic zones. European Centre for Medium‐Range Weather Forecasts ReAnalysis (ERA‐5) dataset performed better in predicting drought events in more climatic zones compared to Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks‐Climate Data Record (PERSIANN‐CDR) and Climate Hazards Group InfraRed Precipitation with Station (CHIRPS) when compared against Indian Meteorological Department (IMD) dataset. In contrast, PERSIANN‐CDR proved effective in predicting drought at the country scale. ERA‐5 could be suitable for real‐time drought monitoring and prediction, whereas PERSIANN‐CDR can be used for retrospective drought analysis. The proposed approach and the best‐performed algorithm (NBR) can be extended and applied in any climatic region for enhancing the drought predictions where remotely sensed information are accessible even in regions with finite ground data availability.

No abstract

Semi-analytical algorithms (SAAs) invert spectral remote sensing reflectance (Rrs(λ), sr−1) to Inherent Optical Properties (IOPs) of an aquatic medium (λ is the wavelength). Existing SAAs implement different methodologies with a range of spectral IOP models and inversion methods producing concentrations of non-water constituents. Absorption spectrum decomposition algorithms (ADAs) are a set of algorithms developed to partition anw(λ), m−1 (i.e., the light absorption coefficient without pure water absorption), into absorption subcomponents of phytoplankton (aph(λ), m−1) and coloured detrital matter (adg(λ), m−1). Despite significant developments in ADAs, their applicability to remote sensing applications is rarely studied. The present study formulates hybrid inversion approaches that combine SAAs and ADAs to derive absorption subcomponents from Rrs(λ) and explores potential alternatives to operational SAAs. Using Rrs(λ) and concurrent absorption subcomponents from four datasets covering a wide range of optical properties, three operational SAAs, i.e., Garver–Siegel–Maritorena (GSM), Quasi-Analytical Algorithm (QAA), Generalized Inherent Optical Property (GIOP) model are evaluated in deriving anw(λ) from Rrs(λ). Among these three models, QAA and GIOP models derived anw(λ) with lower errors. Among six distinctive ADAs tested in the study, the Generalized Stacked Constraints Model (GSCM) and Zhang’s model-derived absorption subcomponents achieved lower average spectral mean absolute percentage errors (MAPE) in the range of 8–38%. Four hybrid models, GIOPGSCM, GIOPZhang, QAAGSCM and QAAZhang, formulated using the SAAs and ADAs, are compared for their absorption subcomponent retrieval performance from Rrs(λ). GIOPGSCM and GIOPZhang models derived absorption subcomponents have lower errors than GIOP and QAA. Potential uncertainties associated with datasets and dependency of algorithm performance on datasets were discussed.

No abstract