Effectiveness of Common Preprocessing Methods of Time Series for Monitoring Crop Distribution in Kenya

Accurate crop identification and spatial distribution mapping are important for crop production estimation and famine early warning, especially for food-deficit African agricultural countries. By evaluating existing preprocessing methods for classification using satellite image time series (SITS) in Kenya, this study aimed to provide a low-cost method for cultivated land monitoring in sub-Saharan Africa that lacks financial support. SITS were composed of a set of MODIS Vegetation Indices (MOD13Q1) in 2018, and the classification method included the Support Vector Machine (SVM) and Random Forest (RF) classifier. Eight datasets obtained at three levels of preprocessing from MOD13Q1 were used in the classification: (1) raw SITS of vegetation indices (R-NDVI, R-EVI, and R-NDVI + R-EVI); (2) smoothed SITS of vegetation indices (S-NDVI); and (3) vegetation phenological data (P-NDVI, P-EVI, R-NDVI + P-NDVI, and P-NDVI-1). Both SVM and RF classification results showed that the “R-NDVI + R-EVI” dataset achieved the highest performance, while the three pure phenological datasets produced the lowest accuracy. Correlation analysis between variable importance and rainfall time series demonstrated that the vegetation index SITS during rainfall periods showed higher importance in RF classifiers, thus revealing the potential of saving computational costs. Considering the preprocessing cost of SITS and its negative impact on the classification accuracy, we recommend overlaying the original NDVI with the original EVI time series to map the crop distribution in Kenya.

Time-series analysis and statistical forecasting of daily rainfall in Kupang, East Nusa Tenggara, Indonesia: a pilot study

This pilot study presents a novel statistical time-series approach for analyzing daily rainfall data in Kupang, East Nusa Tenggara, Indonesia. By using the piecewise cubic hermite interpolation algorithm, we succeeded in filling in the null values in the daily rainfall time series. We then analyzed the monthly average and its pattern using the continuous wavelet transform (CWT) algorithm, which shows the strong annual pattern of rainfall in this region. In addition, we use the rainfall anomaly index (RAI) function to standardize daily rainfall as an indicator of dry/wet conditions in this region. Then we also use the daily RAI time-series objects from 1978 to 2020 for modeling and predicting daily RAI over the next year. The result is the root mean squared error (RMSE) of 0.8424041040593219. This Prophet model is also able to capture the linear trend of increasing drought throughout the study time period and the annual pattern of wet/dry conditions which is in accordance with previous study by Aldrian and Susanto (2003).

A Comparative Effectiveness of Hierarchical and Non-hierarchical Regionalisation Algorithms in Regionalising the Homogeneous Rainfall Regions

Descriptive data mining has been widely applied in hydrology as the regionalisation algorithms to identify the statistically homogeneous rainfall regions. However, previous studies employed regionalisation algorithms, namely agglomerative hierarchical and non-hierarchical regionalisation algorithms requiring post-processing techniques to validate and interpret the analysis results. The main objective of this study is to investigate the effectiveness of the automated agglomerative hierarchical and non-hierarchical regionalisation algorithms in identifying the homogeneous rainfall regions based on a new statistically significant difference regionalised feature set. To pursue this objective, this study collected 20 historical monthly rainfall time-series data from the rain gauge stations located in the Kuantan district. In practice, these 20 rain gauge stations can be categorised into two statistically homogeneous rainfall regions, namely distinct spatial and temporal variability in the rainfall amounts. The results of the analysis show that Forgy K-means non-hierarchical (FKNH), Hartigan- Wong K-means non-hierarchical (HKNH), and Lloyd K-means non-hierarchical (LKNH) regionalisation algorithms are superior to other automated agglomerative hierarchical and non-hierarchical regionalisation algorithms. Furthermore, FKNH, HKNH, and LKNH yielded the highest regionalisation accuracy compared to other automated agglomerative hierarchical and non-hierarchical regionalisation algorithms. Based on the regionalisation results yielded in this study, the reliability and accuracy that assessed the risk of extreme hydro-meteorological events for the Kuantan district can be improved. In particular, the regional quantile estimates can provide a more accurate estimation compared to at-site quantile estimates using an appropriate statistical distribution.

Analysis of rainfall and temperature trends of selected stations over North East India during last century

In this study, the trends of seasonal maximum and minimum temperatures and rainfall time series were investigated for 9 selected stations in the north eastern India with the available data stretching between the years 1913-2012.During the period under study the minimum temperature has increasing trends in almost all the stations of north east India except Cherrapunji where it shows decreasing trend in all the season of the year. In case of maximum temperature Cherrapunji, Guwahati and Imphal show increasing trends during all the seasons. Agartala and Shillong show increasing trend of maximum temperature during monsoon and post monsoon season. Dibrugarh and Pasighat show decreasing trend during pre monsoon season and increasing trend during all other seasons of the year. Gangtok shows decreasing trend of maximum temperature during all the seasons where as Silchar shows no trend in maximum temperature.Out of all the selected nine stations, most of the stations show either decreasing trend or no trend of rainfall except Guwahati which shows significant increasing trend of rainfall during post monsoon season.

Monthly rainfall forecasting by a hybrid neural network of discrete wavelet transform and deep learning

Actual rainfall forecast is critical to the management and allocation of water resources. In recent years, deep learning has been proved to be superior to traditional forecasting methods when predicting rainfall time series with high temporal and spatial variability. In this study, the discrete wavelet transform (DWT) and two typical deep learning approaches, namely long-short term memory (LSTM) and dilated causal convolutional neural network (DCCNN), are integrated innovatively and the hybrid model (DWT-CLSTM-DCCNN) is used for monthly rainfall forecasting for the first time. Monthly rainfall time series of four major cities in China (Beijing, Tianjin, Chongqing and Guangzhou) are used as the dataset of DWT-CLSTM-DCCNN. Firstly, two methods of sample construction are used to train DWT-CLSTM-DCCNN and their effects on the model performance are analyzed. Then, LSTM and the dilated causal convolutional network (DCCNN) are established as the benchmark models, and their forecast accuracy is compared with that of DWT-CLSTM-DCCNN. From the results of the evaluation criteria such as mean absolute error (MAE), root mean squared error (RMSE) and Nash-Sutcliffe model efficiency coefficient (NSE) as well as the fitting curve for forecasted rainfall, it can be concluded that the DWT-CLSTM-DCCNN developed in this study outperforms the benchmark models in model accuracy, peak and mutational rainfall capturing ability. Compared with the previous studies, DWT-CLSTM-DCCNN is proven to be better peak capture and more suitable for long-term rainfall forecasting.

Study of rainfall departure over catchments of Bihar plains

Station wise daily rainfall data of sixty years is used to study rainfall departure and variability in Kosi, Kamala/Bagmati/Adhwara and Gandak/Burhi Gandak catchments during monsoon season. Station and catchment wise rainfall time series have been made to compute rainfall departure and Coefficient of Variation (CV). Southern Oscillation Index (SOI), Multivariate ENSO Index (MEI) and ENSO strength based on percentile analysis are used to ascertain their impact on rainfall distribution in the category as excess, normal, deficient and scanty. Results indicate that the variability is greater over Kosi as compared to the other catchments. Probability of normal rainfall is found 0.75 and there is no possibility of scanty rain over the catchments during El Nino and La Nina year. Similarly probabilities of normal, deficient, excess rainfall are found as 0.67, 0.18 and 0.15 respectively during mixed year. SOI has emerged as principal parameter which modifies the departure during El Nino and La Nina year. MEI along with ENSO strength are more prominent during mixed year particularly to ascertain deficient and excess rain in weak and strong- moderate La Nina respectively .

Comparing methods to place adaptive local RTC actuators for spill volume reduction from multiple CSOs

The selection of flow control device (FCD) location is an essential step for designing real-time control (RTC) systems in sewer networks. In this paper, existing storage volume-based approaches for location selection are compared with hydraulic optimisation-based methods using genetic algorithm (GA). A new site pre-screening methodology is introduced, enabling the deployment of optimisation-based techniques in large systems using standard computational resources. Methods are evaluated for combined sewer overflow (CSO) volume reduction using the CENTAUR autonomous local RTC system in a case study catchment, considering overflows under both design and selected historic rainfall events as well as a continuous 3-year rainfall time series. The performance of the RTC system was sensitive to the placement methodology, with CSO volume reductions ranging between −6 and 100% for design and lower intensity storm events, and between 15 and 36% under continuous time series. The new methodology provides considerable improvement relative to storage-based design methods, with hydraulic optimisation proving essential in relatively flat systems. In the case study, deploying additional FCDs did not change the optimum locations of earlier FCDs, suggesting that FCDs can be added in stages. Thus, this new method may be useful for the design of adaptive solutions to mitigate consequences of climate change and/or urbanisation.

Stochastic simulation of reference rainfall scenarios for hydrological applications using a universal multifractal approach

Hydrological applications such as storm-water management or flood design usually deal with and are driven by region-specific reference rainfall regulations or guidelines based on Intensity-Duration-Frequency (IDF) curves. IDF curves are usually obtained via frequency analysis of rainfall data using which the exceedance probability of rain intensity for different durations are determined. It is also rather common for reference rainfall to be expressed in terms of precipitation P, accumulated in a duration D (related to rainfall intensity ), with a return period T (inverse of exceedance probability). Meteorological modules of hydro-meteorological models used for the aforementioned applications therefore need to be capable of simulating such reference rainfall scenarios. The multifractal cascade framework, since it incorporates physically realistic properties of rainfall processes (non-homogeneity or intermittency, scale invariance and extremal statistics) seems to suit this purpose. Here we propose a discrete-in-scale universal multifractal (UM) cascade based approach. Daily, Hourly and six-minute rainfall time series datasets (with lengths ranging from 100 to 15 years) over three regions (Paris, Nantes, and Aix-en-Provence) in France that are characterized by different climates are analyzed to identify scaling regimes and estimate corresponding UM parameters (α, C1) required by the UM cascade model. Suitable renormalization constants that correspond to the P, D, T values of reference rainfall are used to simulate an ensemble of reference rainfall scenarios, and the simulations are finally compared with datasets. Although only purely temporal simulations are considered here, this approach could possibly be generalized to higher spatial dimensions as well.

Comparative Evaluation of the Rainfall Erosivity in the Rieti Province, Central Italy, Using Empirical Formulas and a Stochastic Rainfall Generator

Soil erosion caused by intense rainfall events is one of the major problems affecting agricultural and forest ecosystems. The Universal Soil Loss Equation (USLE) is probably the most adopted approach for rainfall erosivity estimation, but in order to be properly employed it needs high resolution rainfall data which are often unavailable. In this case, empirical formulas, employing aggregated rainfall data, are commonly used. In this work, we select 12 empirical formulas for the estimation of the USLE rainfall erosivity in order to assess their reliability. Moreover, we used a Stochastic Rainfall Generator (SRG) to simulate a long and high-resolution rainfall time series with the aim of assessing its application to rainfall erosivity estimations. From the analysis, performed in the Rieti province of Central Italy, we identified three equations which seem to provide better results. Moreover, the use of the selected SRG seems promising and it could help in solving the problem of hydrological data scarcity and consequently guarantee major accuracy in soil erosion estimation.