A Deep and Wide Neural Network-based Model for Rajasthan Summer Monsoon Rainfall (RSMR) Prediction

03/03/2021
by   Vikas Bajpai, et al.
0

Importance of monsoon rainfall cannot be ignored as it affects round the year activities ranging from agriculture to industrial. Accurate rainfall estimation and prediction is very helpful in decision making in the sectors of water resource management and agriculture. Due to dynamic nature of monsoon rainfall, it's accurate prediction becomes very challenging task. In this paper, we analyze and evaluate various deep learning approaches such as one dimensional Convolutional Neutral Network, Multi-layer Perceptron and Wide Deep Neural Networks for the prediction of summer monsoon rainfall in Indian state of Rajasthan.For our analysis purpose we have used two different types of datasets for our experiments. From IMD grided dataset, rainfall data of 484 coordinates are selected which lies within the geographical boundaries of Rajasthan. We have also collected rainfall data of 158 rain gauge station from water resources department. The comparison of various algorithms on both these data sets is presented in this paper and it is found that Deep Wide Neural Network based model outperforms the other two approaches.

READ FULL TEXT VIEW PDF
10/22/2020

Prediction of Rainfall in Rajasthan, India using Deep and Wide Neural Network

Rainfall is a natural process which is of utmost importance in various a...
06/09/2021

Multistep Electric Vehicle Charging Station Occupancy Prediction using Mixed LSTM Neural Networks

Public charging station occupancy prediction plays key importance in dev...
01/21/2021

Filling the gap between GRACE- and GRACE-FO-derived terrestrial water storage anomalies with Bayesian convolutional neural networks

There is an approximately one-year observation gap of terrestrial water ...
03/31/2017

Comparison of multi-task convolutional neural network (MT-CNN) and a few other methods for toxicity prediction

Toxicity analysis and prediction are of paramount importance to human he...
11/19/2018

Automatic Three-Dimensional Cephalometric Annotation System Using Three-Dimensional Convolutional Neural Networks

Background: Three-dimensional (3D) cephalometric analysis using computer...
09/09/2020

Revealing Lung Affections from CTs. A Comparative Analysis of Various Deep Learning Approaches for Dealing with Volumetric Data

The paper presents and comparatively analyses several deep learning appr...
10/27/2020

A robust low data solution: dimension prediction of semiconductor nanorods

Precise control over dimension of nanocrystals is critical to tune the p...

1 Introduction

Understanding of rainfall characteristics is important for a variety of activities including efficient engineering, planning and management of water resources Halbe et al. (2013); Campling et al. (2001). In addition to this, rainfall play a major role in balancing of various activities such as, hydrologic cycle, water availability for terrestrial animals, agriculture and industrial processes. Rainfall and its estimation is not only important for India but is equally important for the entire globe Jang and Han (2011); Tierney et al. (2011); Clift and Plumb (2008); Fan et al. (2013); Ancy et al. (2014); Kerhoulas et al. (2017); He et al. (2006); Lau and Li (1984).

In India majority of the rain is received from the month of June to September (June-July-August-September) and that is why this period is called as Indian Summer Monsoon Rainfall (ISMR) or the Southwest monsoon rainfall. Cultivated land in India is majorly benefited by this ISMR Swaminathan (1998) which makes this season highly important and ultimately prediction and estimation of rainfall for this period also becomes equally essential. India receives nearly 80 percent rainfall during summer monsoon period Mooley and Parthasarathy (1984); Naidu et al. (2009); Mooley (1997) only. This summer monsoon rainfall fuhrer helps in predicting food grain production Parthasarathy et al. (1988) which ultimately contributes to country’s GDP222https://statisticstimes.com/economy/country/india-gdp-growth-sectorwise.php. Prediction and estimation of ISMR started way back from the year 1903 as people started believing on the importance of this monsoon rainfall Walker (1933).

In this work, our area of study is Rajasthan which is the largest state of India and the 60% of its area falls under the arid category which makes it very environmentally sensitive Dutta and Chaudhuri (2015). Even after being an arid to semi-arid zone, Rajasthan has observed several floods in the past Goyal et al. (2021); Yadav and Mangukiya (2021); Ray et al. (2019)and also observed several droughts KARAN (2016); Bokil (2000); Goel and Singh (2006); Mundetia et al. (2015); Parthasarathy et al. (1987). An early indication of the amount of monsoon rainfall a particular region is going to receive, can be very handy in terms of managing the water resource for the entire year. This early indication can give us an idea about the amount of availability of water in a particular reservoir. Now this reservoir which will cater to the needs of demand from people and industry in a particular area can be regulated and measures can be taken well in advance for proper water resource management for the monsoon and non-monsoon period.

There are several indicators on which rainfall depends, such as surface temperature, sea level, distance from sea, distance from mountain ranges etc. In this work, we propose a time series based approach for the prediction of rainfall for the months of June, July, August and September (summer monsoon months). For this we collected Indian Meteorological Department ( IMD hereafter) grided data of 118 years ( from the year 1901 to 2018) and station data of 61 years( from the year 1957 to 2017) from Water Resources Department, Rajasthan (WRD hereafter). In this work we design and analyze advance deep learning models to capture the patterns from this historical time series data for the prediction of Rajasthan summer monsoon rainfall (RSMR hereafter). For this we adapt and improve a model originally proposed by Cheng et al Cheng et al. (2016b) in the field of recommender systems. We name our proposed model as Deep and Wide Monsoon Rainfall Prediction Model (DWMRPM hereafter) and compared with advance deep-learning based models like multi-layer perceptron (MLP), one dimensional convolutional neural network (1D-CNN) based neural networks

1.1 Related Work

In past researchers have applied numericalDucrocq et al. (2002) and statistical modelsLi and Shao (2010); Montanari and Grossi (2008)

for rainfall prediction. But with gaining popularity of artificial intelligence and increasing machine computation power, training abundant data using machine learning and deep learning models are becoming the center of attraction for researchers

Zhang et al. (1998). One of the major reasons of scientists switching from traditional numerical approaches to artificial intelligence based approaches is that the statistical and numerical models fail to capture the dynamic nature of rainfall Singh et al. (2012a) whereas neural networks are quiet smart in capturing the hidden trends and seasonality existing in time series rainfall data. The numerical and statistical models were used majorly for two to three decades but these methods lacked forecasting accuracy Gadgil et al. (2005) resulting into failure in predicting major rainfall variations Kalsi et al. (2004); Sikka (2003). There are evidences from the past where these numerical methods failed Gadgil et al. (2002); Preethi et al. (2011) to predict the monsoon rainfall and severe droughts were observed.

Pritpal Singh (2018a) has made an attempt to predict the ISMR using monthly monsoon rainfall values and applied fuzzy sets and artificial neural network (ANN). When the parameters on which the rainfall depends are very high then in order to predict ISMR, Saha et al. (2016) used auto encoder Ng and others (2011) for reducing the number of parameter and then predicted the ISMR. Saha et al. (2021) studied the climatic variables responsible for ISMR and used deep learning feature for monsoon rainfall prediction.This study also shows the monsoon deviation from long period average (LPA) rainfall. Johny et al used an adaptive Ensemble Model of ANN which was capable of capturing very low and very high rainfall in the Indian state of Kerala Johny et al. (2020). Dubey et. al Dubey (2015)

used three artificial neural network based algorithms ( feed-forward back propagation algorithm, layer recurrent algorithm and feed-forward distributed time delay algorithm) for rainfall prediction over the region of Pondicherry, India. Some amount of monsoon rainfall prediction is done by applying feed forward neural network

Chakraverty and Gupta (2008); Sahai et al. (2000); Singh and Borah (2013).

Fluctuations in the summer monsoon rainfalls can’t be captured efficiently by traditional linear statistical models Singh (2018b); Dash et al. (2018). This motivated us to use Deep Learning based model which are efficient in capturing this non-linearity and dynamic nature of ISMR. As per IMD weather forecasting manual 333https://imdpune.gov.in/Weather/Forecasting_Mannuals/IMD_IV-13.pdf, Indian rainfall is very well known for its variability in space and time. There is hardly any seasonal distribution of rainfall over entire India. At two different station locations which are a few miles apart, if we consider one day rainfall, we may observe that one station experiencing heavy rainfall whereas the other station may go completely dry. This kind of variation is not only found in monsoon rainfall period(June to September) but also during post monsoon period (October to December) as well.

A good amount amount of work has been done in the field of ISMR as presented above but at present to the best of our knowledge, no work is done in the field of RSMR prediction, which attracted the authors of this paper to explore this untouched area. An attempt to predict the agricultural drought index in Rajasthan is done by Dutta et al Dutta et al. (2013) using standardized precipitation index. In this proposed work, an extensive study is done in predicting RSMR for the first time. The good thing about Rajasthan is the strong Rain Gauge network from IMD, Water Resource Department, Rajasthan and the Revenue Department which has resulted into the abundant supply of rainfall data for analysis and prediction.

Research work done in the field of Rainfall Prediction and Estimation for the state of Rajasthan is very less. Vikas et al Bajpai et al. (2020) used the historical time-series data for daily rainfall prediction. However worked in analyzing the trends of rainfall in the state of RajasthanPingale et al. (2014); Yadav et al. (2018). Bryson and Swain (1981) made an effort to present the holocene variations of monsoon rainfall in Rajasthan. Meena et al. (2019) made an attempt to explore the spatial and temporal differences to identify trends in monthly, seasonal and annual rainfall over the Rajasthan region. They observed the prevailing homogeneity of rainfall at various stations in the state. In another work Singh et al. (2012b)

authors tried to estimate the one day maximum rainfall in Jhalrapatan, a city in the state of Rajasthan. Authors have done the probability analysis for this purpose.

Lal Meena et al. (2020) studied the rainfall pattern in Chaksu, Rajasthan.

Our objective is to predict the RSMR which starts in the month of June and ends in the month of September. In this work we propose a time series based prediction model which depends on the fundamental of present and future time series data dependency on past time series data Singh (2016). We adapt and improvise wide and deep learning model originally proposed by Cheng et al Cheng et al. (2016b)

for recommendations. Many authors h ave used this concept in different domains like regression analysis

Kim et al. (2020), quality prediction Ren et al. (2020), rainfall prediction Bajpai et al. (2020) etc. Wide networks are used for memorization and deep networks are used for generalization. In this work we propose a Deep and Wide Monsoon Rainfall Prediction Model (hereafter DWMRPM) to predict monsoon rainfall prediction in the Indian state of Rajasthan.

The rest of the paper is organized as follows. Section 2 explains the proposed model for summer monsoon rainfall in Rajasthan. Details of experimental evaluations, model training, results of rainfall prediction and comparison with other deep learning approaches is given in section 3. Finally we conclude the paper in Section 4 and provide avenues for future work.

1.2 Major Contributions

  1. In this work, we propose a novel architecture based on deep and wide neural network for the purpose of summer monsoon rainfall using historical time-series data. The model efficiently captures the dynamic nature of monsoon rainfall and works well in its prediction. To the best of our knowledge, we are the first who have tried to solve this challenging problem.

  2. We compare our work with various advanced deep learning algorithms for sequence prediction on two different types of datasets and have obtained very promising results.

  3. The algorithms we designed has the generalization ability and can be used to predict summer monsoon rainfall for atmospherically different regions of Rajasthan.

2 Deep & Wide Monsoon Rainfall Prediction Model (DWMRPM)

This section first provides a brief overview of the proposed approach and subsequently explain various steps involved in the prediction of Rajasthan summer monsoon rainfall (RSMR, hereafter).

2.1 Overview

In this work we address the problem of summer monsoon rainfall in Rajasthan, which is the largest state of India and is located in the North-Western part of the country. Rajasthan has very distinct physiographic characteristics. On one side it has India’s biggest desert area, called The Thar Dessert and on the other side this state has Eastern Plains and the ranges of Aravalli Hills Enzel et al. (1999). These ranges are in the direction of South-west monsoon, which is responsible for rainfall in the region Roy and Jakhar (2002). Atmospherically Rajasthan is divided into four zones: North West Desert Region, Central Aravalli Hill Region, Eastern Plains and South Eastern Plateau Region Upadhyaya (2014). Details of the districts, which come under the respective zones are given below:

  • []

  • North-West Desert Region: Jaisalmer, Jodhpur, Hanumangarh, Shriganganagar, Barmer, Churu, Nagaur, Pali, Sikar, Bikaner and Jhunjhunu

  • Central Aravalli Hill Region: Udaipur, Dungarpur, Sirohi, Jalore, Pali, Banswara, Bhilwara, Chittorgarh, Rajsamand and Ajmer

  • Eastern Plains: Alwar, Bharatpur, Tonk, Sawai Madhopur, Karauli, Jaipur, Dausa and Dhoulpur

  • South-Eastern Plateau Region: Kota, Bundi, Jhalawar and Baran

All these zones have different atmospheric and climatic conditions. The problem of predicting summer monsoon rainfall in Rajasthan is different from the prediction of Indian summer monsoon rainfall (ISMR, hereafter). Most of the time-series-based methods for predicting ISMR consider average monthly rainfall values by taking weighted average of the 306 well distributed rain-gauge stations in the non-hilly areas of Indian sub-continent Dash et al. (2018); Singh (2018b); Singh and Borah (2013); Sahai et al. (2000). Rajasthan being a dry state lies in arid and semi-arid zones and characterized by low and uneven rainfall Kulshreshtha et al. (2013), therefore, a dedicated system is required which can predict monsoon rainfall for different geographical regions separately. We use historical monthly rainfall data from two different sources to train and analyze the performance of our model in prediction of Rajasthan Summer Monsoon Rainfall. Details on the datasets are given in Section 2.2

For ISMR researchers used monthly rainfall values of June to September across all the years Dash et al. (2018) or just have captured the dependency of months of a single year Singh (2018a) In order to avoid loss of any information, we are using rainfall values of all the months of previous years for the prediction of rainfall for the months of June, July, August and September. For example in order to predict rainfall for the month of June 2019, we use rainfall values of all the months from May 2000 to May 2019.

In this work, we propose a deep and wide monsoon rainfall prediction model (DWMRPM) for the prediction of the total monthly rainfall intensity for the summer monsoons months of Rajasthan. The wide network is used to extract low-dimensional features. Here, instead of using a sequence of monthly rainfall values directly, we are using features obtained after applying a convolutional layer, as it is very effective in learning spatial dependencies in and between the series of data Van et al. (2020). High-dimensional features, on the other hand, are derived using Multi-layer perceptron (MLP) Pal and Mitra (1992) in which a sequence of rainfall intensity values are passed on to a deep network. In order to incorporate a geographical generalization ability in the model, so that a single model can be used to make rainfall predictions in different geographical conditions, information of geographical parameters (latitude and longitude) is included at the time of training. The operational steps involved in the development of our proposed DWMRPM for the prediction of rainfall are shown in Figure 1.

Figure 1: Overview of DWMRPM: The model takes sequence of monthly rainfall intensities and geographical parameters, namely latitude and longitude as input. After initial pre-processing, input is fed to a deep network, which is a multi-layer perceptron, and to a wide network, which is a convolutional network. The model is jointly trained considering the activation weights from both deep and wide networks simultaneously.

To evaluate the performance of the proposed method, we use two standard statistical metrics, namely mean absolute error (MAE) and root mean square error (RMSE). We compare our results with the advance deep learning models like MLP and one dimensional convolutional neural networks (1-DCNN) which are very popular for sequence based predictions.

2.2 Dataset description and pre-processing

In this work we have used Water Resources Department dataset and Indian Meteorological Department (IMD) gridded rainfall data with a high spatial resolution of Pai et al. (2014). From IMD data set, we selected the rainfall data of the Rajasthan meteorological sub-division ranging from N to N latitude and E to E longitude, for the period of 118 years from the year 1901 to 2018. It gave the rainfall data for 1008 rain-gauge stations. We have also collected the rainfall data from Rajasthan’s water resources department, for more than 500 rain-gauge stations, over a period of 61 years (from the year 1957 to 2017). The datasets were noisy in terms of negative and missing values. After initial level data pre-processing and cleansing steps, we selected 484 co-ordinates from IMD dataset of High Spatial Resolution of (0.25X0.25 degree) and 158 stations from Rajasthan’s water resources data for our analysis. The distribution of the selected stations from water resources data, over 33 districts are depicted on the map of Rajasthan in Figure 4

(a)
(b)
Figure 4: Map of Rajasthan showing distribution of (a) selected 484 coordinates from IMD gridded dataset at high spatial resolution of , (b) selected 158 rain-gauge stations from Water Resources Department, Rajasthan. The rainfall data obtained from these coordinates and rain-gauge stations are used for the prediction of Rajasthan summer monsoon rainfall (RSMR)

In this paper, authors have made use of both the Station data (data collected from various Rain Gauge Stations in Rajasthan) and the Gridded data Pai et al. (2014)

. The idea behind using both the data sets is that when only the station data is used, for experimentation, one uses the data for single point of scale whereas when the gridded data is used the application of different meteorological data for a region is applied depending upon the resolution. In large catchment areas where less number of Rain Gauges are installed, modeling may not be that much accurate, on the other hand gridded data is more continuous and may prove better than single point estimates. Gridded data contains the data from stations or satellites (in our case, it’s rain gauge station data) which undergoes interpolation over a grid. This interpolation needs careful analysis for biases and outliers

Rajeevan et al. (2005); Pai et al. (2014). Station data on the other hand is unbiased single point data. For our study, we have used quality controlled data sets from both the categories. If someone has enough single point station in the region under study, then the station data can be easily utilized but since the rain gauge distribution is not uniform ( as shown in Raj_WRD), specially in the dessert areas where the rain gauge station installation density is very low, combination of both the data set seems to be optimal. Another advantage of using gridded data is that it acts as a source of replacement to the data missing from the records of rain gauge stationsMeher and Das (2019). Any area or zone where the observed station data (point data) is comparatively less, interpolated gridded data can work as a potential alternative means Bandyopadhyay et al. (2018).

These datasets contained daily rainfall values from which we calculated monthly rainfall values for January to December. In order to provide rainfall pattern in Rajasthan, mean rainfall values for each month and the monsoon season (combined rainfall of June, July, August and September) from the year 1901-2018 for a randomly picked rain-gauge station is shown in Table 1. We also provide the minimum and maximum rainfall for each month and monsoon season, over the duration of 118 years. It can be observed that the significant amount of annual rainfall occurs in the monsoon months and the in the remaining months, the aggregate rainfall is very less.

Month Mean (mm) Maximum (mm) Minimum (mm)
IMD WRD IMD WRD IMD WRD
Jan 4.24 2.20 63.66 53.00 0 0
Feb 4.22 2.40 65.56 54.00 0 0
Mar 3.81 2.11 64.44 74.00 0 0
Apr 2.91 3.78 43.03 56.00 0 0
May 9.14 5.72 90.75 87.00 0 0
Jun 49.67 41.92 246.2 227.00 0 0
Jul 159.45 150.54 523.5 476.00 13.11 10
Aug 160.86 166.05 441 905.0 3.41 40.8
Sep 65.38 63.83 305.8 402.00 0 0
Oct 9.84 7.31 154.4 132.00 0 0
Nov 1.72 4.02 25.67 160.00 0 0
Dec 2.31 1.04 46.53 30.00 0 0
Overall Accuracy 868.63 460.05 1080.90 937.0 91.17 102.1
Table 1: Statistical summary of monthly data for the IMD dataset and the dataset from the water resource department of Rajasthan (WRD). Mean, maximum and minimum rainfall fall values of IMD are shown for a randomly picked coordinate at N and E for a period of 118 years from 1901 to 2018. The rainfall statistics for WRD dataset for a randomly picked rain-gauge station, situated at N and E is shown for a period of 61 years from the year 1957 to 2017.

We have considered time-series values of monthly rainfall and geographical parameters like latitude and longitude for the prediction of rainfall during the monsoon months in different regions of Rajasthan. The rainfall intensity values ranges from 0 mm to more than 800 mm while coordinate values of latitude and longitude lies between N to N and to E, respectively. Since the data is of different dimensions and dimensional units, therefore we normalize the data to make it dimensionally uniform. When the magnitude of different parameters in a dataset is different, the parameters with higher values suppresses the role of the parameters with lower values in model training. To handle this issue, we use the min-max normalization method to convert all rainfall intensity values to number between 0 and 100 (latitude and longitude values are already in this range). The mathematical representation of the min-max normalization method is as follows:

where, is the normalized value of the monthly rainfall intensity value, represents a value in the original dataset, and are the maximum and minimum intensity values, respectively. Normalization can also help in improving the learning capability of the model and in reducing the computational complexity Shanker et al. (1996).

2.3 Model Description

We use deep and wide neural network-based architecture Bajpai et al. (2020) for the purpose of summer monsoon rainfall prediction in the Indian state of Rajasthan. The following paragraphs explains the major components of the model.

2.3.1 The Wide Component: Convolutions

The wide component is used to memorize certain combinations of monthly rainfall events, which is beyond the capabilities of the deep model. It is a generalized linear model of type . In the model proposed by Cheng et al Cheng et al. (2016a), cross-product feature transformations were used as the wide component. In this work we use convolutional network as wide component. The basic components of a general CNN consists of 2 types of layers, namely convolutional layer and pooling layer Gu et al. (2018). The convolutional layer is composed of several convolutional kernels, which capture and learn the correlation of spatial features by computing different feature maps. The output of one dimensional convolutional layer with input size is:

where, is the layer number, is the kernel from the neuron at layer to the neuron at layer , , activations, bias at layer.

Convolutional layer is followed by a pooling layer that is used to realize shift invariance by reducing the resolution of the feature maps. As demonstrated by Van et al. (2020), 1D CNN performs well in regression type of problems and can learn to find the correlation in between the series very efficiently. Therefore, instead of using raw features in the wide part of the network, we use a convolutional layer to capture such combinations. In addition to this, to make our model more generalized with respect to different atmospheric conditions, we are using geographical parameters namely, longitude and latitude while designing and developing our model (Figure 5).

Figure 5:

Selected architecture of DWMRPM for prediction of Rajasthan Summer Monsoon Rainfall. There are two major components: 1.The Deep component consists of mainly an input layer and 3 ReLU layers. 2. The wide component consists of a convolutional layer followed by a global average pooling layer.A sequence of monthly rainfall intensity values after normalization and values are fed to deep and wide components separately

2.3.2 The Deep Component: Multi-layer Perceptron

The deep component is a feed-forward neural network, specifically a multi-layer perceptron, as shown in Figure 5. Sequence of monthly rainfall intensity values are given as input, which are then fed into hidden layers of a neural network in the forward pass. Typically, each hidden layer computes:

where, is the layer number and

is the activation function, rectified linear units (ReLUs) in our case,

, , and are the activations, bias and model weights at layer.

2.3.3 Joint training of the model

The model is trained using the joint training approach that optimizes all parameters simultaneously by taking into account the output of the deep and wide components and their weighted sum. It helps in providing an overall prediction, which is based on aforementioned components, also depicted in Figure 5.

where, is the prediction,

are the output vectors of two sub-models namely wide-convolutional model and deep model respectively, and

are their respective weight vectors to be trained.

3 Experimental evaluations

3.1 Implementation details

All the experimental programs are coded using Keras

Chollet and others (2015)

API of TensorFlow framework

Abadi et al. (2016); Gulli and Pal (2017). The hardware setup includes computer processor from Intel with i7-8750H configuration supported by 32GB RAM. The upcoming sections and subsections describe the designing and implementation setup of proposed approach and baseline approaches followed by results obtained.

For prediction of monthly rainfall of monsoon season, we consider different training windows of lengths ranging from 2 years to 10 years. We found that 9 years training window gives most accurate prediction results for the monsoon months of June, July August and September.

3.1.1 Training, validation and test sets

We use two type of datasets, one from the Indian Meteorological Department (IMD) and the other from the Water Resource Department (WRD). In case of WRD, monthly rainfall values from the year 1957 to 1986 are considered for the purpose of training. Validation of the model is done on the dataset considering the years starting from 1987 to 1997 and finally we test the model on the dataset containing monthly rainfall values in the interval of the year 1998 to 2017. In case of IMD dataset, training is done by considering values from the year 1901 to 1980 and validation is done from the year 1981 to 1995 and finally testing is done on the rainfall intensity values from the year 1996 to 2018.

3.1.2 Evaluation metrics

As shown by Glorot and Bengio (2010) and He et al. (2015a)

, to evaluate the overall accuracy of predictions, we use root mean square error (RMSE) and mean absolute error (MAE) as the basic evaluation metrics. Low value of RMSE and MAE means better prediction accuracy of the model.

(1)
(2)

where, N represents the number of samples, is the actual rainfall of the sample and is the corresponding prediction.

3.1.3 Model Training

We optimize various hyper parameters like the batch size, number of hidden layers, number of neuron and the dropout rates using trial-and-error method. The network configuration of DWMRPM used in our experiments is shown in Figure 5. The input to the model is the normalized sequence of monthly rainfall intensity values and actual coordinate values (latitude and longitude). The deep part is a Multi-layer perceptron with an input layer; 3 hidden layers containing 300, 200 and 100 neural units with ReLU as the activation function; and finally a dense output layer.In order to prevent over-fitting of the model, dropout layers srivastava2014dropout with dropout rate 0.3 are added after each hidden layer. The wide part contains a convolutional layer with 100 filters, each of size 1 x 5, followed by a global average pooling layer. The outputs of both the wide and deep networks are concatenated and the model is trained using the joint-training approach, as explained in Section 2.3.3. We use Adam optimizer Kingma and Ba (2014)

for training with Mean Square Error (MSE) as loss function, which is calculated as follows:

Here, N represents the number of samples, is the actual rainfall of the sample and is the corresponding prediction. The goal of the model is to find optimized parameters that minimizes MSE

where, is the total number of trainable parameters. Weights of the network are initialized using He initializationHe et al. (2015b)

. Model is trained for 200 epochs with batch size equals to 8.

(a)
(b)
Figure 8: Architecture of the baseline approaches, selected after experimentation with various hyper-parameters. (a) Network configuration of multi-layer perceptron, and (b) Network configuration of CNN.

3.1.4 Baseline approaches

In order to establish the competence of our proposed approach, we have compared the results obtained from the proposed DWMRPM with the results of two advance deep learning approaches: MLP and 1-DCNN. These approaches are working well but not at par with our proposed approach. We have used the same sets of both the data sets for all the approaches in order to avoid any discrepancies that may arise by using different set of datasets for training and testing the models. The network architecture of the baseline approaches, which is selected (after experimenting with various hyper-parameters) for the comparative analysis with the proposed method is explained in the subsequent paragraphs. In all these approaches, we use Adam optimizer for training and MSE as loss function. Input sequence length is 108 (9 years). (Details in Figure 8)

  • []

  • Multi-layer perceptron (MLP): The network architecture for MLP is shown in Figure (a)a.Sequence of rainfall is normalized and concatenated with latitude and longitude. It contains 3 hidden ReLU layers with 300, 200 and 100 units of neurons respectively.

  • Convolutional Neural Network (CNN): The network architecture selected for CNN is given in Figure (b)b. Sequence of rainfall is normalized and concatenated with latitude and longitude. The setup has two convolutional layers with 100 filter size of 1x5 each followed by Global Average Pooling layer.

3.2 Results and discussion

In the following subsections, we present the results of experimental analysis and comparison of the proposed method with the baseline approaches described in Section 3.1.4.

3.2.1 Forecasting accuracy of DWMRPM

As mentioned in Section 2.1, Rajasthan is divided into four atmospheric zones, each of which having huge difference in their climatic and physiographic properties. To evaluate the effectiveness and accuracy of the proposed model, we apply it on each zone separately. The prediction results on one of the randomly picked rain-gauge stations from each zone is given in Table 2 and graphical representation is shown in Figure 13. Here we have used station data from WRD. IMD gridded data is not used in this case because the dataset is generated by interpolation which may have biases and outliers Rajeevan et al. (2005); Pai et al. (2014).

3.2.2 Generalization ability of DWMRPM

In order to verify generalization ability of our model, we use it for monsoon rainfall prediction in each zone separately. The prediction results for each zone, on the basis of two evaluation criteria i.e., MAE and RMSE (Section 3.1.2) on WRD dataset are shown in Table 2 and Figure 13.

Zone Name Latitude Longitude June July August September
MAE RMSE MAE RMSE MAE RMSE MAE RMSE
North-West Desert 2912’N 7314’E 2.4164 2.7660 5.1061 5.8879 4.8080 5.6060 2.8041 3.3493
Central Aravalli Hill Region 2604’N 7446’E 3.1414 3.7925 7.7769 9.2311 11.5339 13.3020 4.7901 6.8593
Eastern Plain 2641’N 7514’E 3.1223 3.7269 6.5146 7.8600 9.5284 10.7145 2.8730 3.4377
South-Eastern Plateau Region 2518’N 7557’E 6.2528 7.5165 8.3010 10.3782 9.0789 10.7850 5.0400 5.8975
Table 2: Prediction results of DWMRPM on the four rain-gauge stations from WRD dataset. Each station is randomly picked from four different atmospheric zones.
(a)
(b)
(c)
(d)
Figure 13: Prediction results of DWMRPM for four rain-gauge stations, each picked from a different atmospheric zone (Section 3.2.2). Results are from year 2016 to 2017. (a) Prediction results of rain-gauge station situated at 2912’N, 7314’E in North-West dessert region, (b) Prediction results of rain-gauge station situated at 2604’N, 7446’E in Central Aravalli hill region, (c) Prediction results of rain-gauge station situated at 2641’N, 7514’E in Eastern plains region, and (d) Prediction results of rain-gauge station situated at 2518’N, 7557’E in South-Eastern plateau region.

It can be observed that a single model is working well in rainfall forecasting for different geographical conditions ranging from plains and plateaus to desserts and hills.

3.2.3 Comparison with baseline approaches

To establish the significance of present work, we compare the results of our model with the baseline approaches separately using the IMD gridded dataset and WRD station dataset.Table 3 and Table 4 show the comparison of the proposed DWMRPM with other approaches in the prediction of monsoon rainfall for the months of June, July, August and September on WRD and IMD datasets respectively. Overall accuracy of the model in the prediction of rainfall for the monsoon months is also given. Qualitative analysis for the comparison on different datasets is shown in Figure 17 and Figure 21

Month MLP 1-DCNN DWMRPM
RMSE MAE RMSE MAE RMSE MAE
June 7.0382 5.8610 7.7567 5.2118 6.550 4.550
July 12.3831 9.2600 14.2568 10.3249 11.0974 8.7081
August 15.4046 12.3992 15.4564 10.4677 13.7013 10.4781
September 8.1199 9.5221 7.9481 5.9679 6.5770 5.0796
Overall Accuracy 10.2014 7.0106 11.8901 7.9931 9.9637 7.2052
Table 3: Comparison of the proposed DWMRPM with one Dimensional Convolutional Neural Networks (1-DCNN) and Multilayer Perceptron (MLP) on WRD dataset using for each month of the monsoon season.
Month MLP 1-DCNN DWMRPM
RMSE MAE RMSE MAE RMSE MAE
June 6.8156 5.7843 5.9660 5.0429 4.0239 3.1371
July 12.9685 9.9012 12.7962 8.5378 11.7024 7.8652
August 13.8754 12.8701 13.3874 12.6042 12.6878 12.1112
September 6.5700 5.8955 4.6474 4.52739 3.8953 4.0184
Overall Accuracy 11.5009 8.4529 9.5039 4.6780 9.0598 4.2830
Table 4: Comparison of the proposed DWMRPM with one Dimensional Convolutional Neural Networks (1-DCNN) and Multi-layer Perceptron (MLP) on IMD dataset using for each month of the monsoon season.
(a)
(b)
(c)
Figure 17: Comparison of DWMRPM and two deep-learning approaches on WRD dataset for a randomly selected rain-gauge station situated at 2518’N, 7557’E. The results are for the months of June, July, August and September of 20 years (June 1998 to September 2017) (a) Prediction results of MLP, (b) Prediction results of one dimensional CNN and, (d) Prediction results of the proposed DWMRPM.
(a)
(b)
(c)
Figure 21: Comparison of DWMRPM and two deep-learning approaches on IMD dataset for a randomly coordinates situated at 2525’N, 7550’E. The results are for the months of June, July, August and September of 23 years (June 1996 to September 2018) (a) Prediction results of MLP, (b) Prediction results of one dimensional CNN and, (d) Prediction results of the proposed DWMRPM.

4 Conclusion and Future Work

This paper has presented a deep and wide neural network based model for the prediction of Rajasthan Summer Monsoon Rainfall (RSMR). Rainfall data is collected from Water Resource Department, Rajasthan and gridded data of resolution 0.25 X 0.25 degrees from Indian Meteorological Department (IMD). This model has the added advantage of exploiting the benefits from both the interpolated gridded data set and the unbiased single point station data set as well. Results obtained by DWRM are compared with baseline approaches like MLP and CNN. It is observed that for RSMR, the deep and wide model works better than other approaches. In future we may apply similar technique for the prediction of summer monsoon rainfall in other states in India as well as abroad. We plan to add more number of rainfall indicators and explore the possibilities of improving the accuracy of the current method.

5 Acknowledgments

This work is in collaboration with Water Resources, Government of Rajasthan. We are thankful to Indian Meteorological Department (IMD) and Special Project Monitoring Unit, National Hydrology Project, Water Resources Rajasthan, Jaipur, India for providing us the Rainfall data for this study.

6 Declaration

  • []

  • Funding: Not Applicable

  • Conflicts of interest/Competing interests: The authors certify that they have NO affiliations with or involvement in any organization or entity with any financial interest or non-financial interest in the subject matter or materials discussed in this manuscript.

  • Availability of data and material: Available on request.

References

  • M. Abadi, P. Barham, J. Chen, Z. Chen, A. Davis, J. Dean, M. Devin, S. Ghemawat, G. Irving, M. Isard, et al. (2016) Tensorflow: a system for large-scale machine learning. In 12th USENIX symposium on operating systems design and implementation (OSDI 16), pp. 265–283. Cited by: §3.1.
  • S. Ancy, R. Kumar, R. Asokan, and R. Subhashini (2014) Prediction of onset of south west monsoon using multiple regression. In Proceedings of IEEE International Conference on Computer Communication and Systems ICCCS14, pp. 170–175. Cited by: §1.
  • V. Bajpai, A. Bansal, K. Verma, and S. Agarwal (2020) Prediction of rainfall in Rajasthan, India using deep and wide neural network. External Links: 2010.11787 Cited by: §1.1, §1.1, §2.3.
  • A. Bandyopadhyay, G. Nengzouzam, W. R. Singh, N. Hangsing, and A. Bhadra (2018) Comparison of various re-analyses gridded data with observed data from meteorological stations over india. EPiC Series in Engineering 3, pp. 190–198. Cited by: §2.2.
  • M. Bokil (2000) Drought in rajasthan: in search of a perspective. Economic and Political Weekly, pp. 4171–4175. Cited by: §1.
  • R. A. Bryson and A. Swain (1981) Holocene variations of monsoon rainfall in rajasthan. Quaternary Research 16 (2), pp. 135–145. Cited by: §1.1.
  • P. Campling, A. Gobin, and J. Feyen (2001) Temporal and spatial rainfall analysis across a humid tropical catchment. Hydrological processes 15 (3), pp. 359–375. Cited by: §1.
  • S. Chakraverty and P. Gupta (2008) Comparison of neural network configurations in the long-range forecast of southwest monsoon rainfall over india. Neural Computing and Applications 17 (2), pp. 187–192. Cited by: §1.1.
  • H. Cheng, L. Koc, J. Harmsen, T. Shaked, T. Chandra, H. Aradhye, G. Anderson, G. Corrado, W. Chai, M. Ispir, R. Anil, Z. Haque, L. Hong, V. Jain, X. Liu, and H. Shah (2016a) Wide & deep learning for recommender systems. In Proceedings of the 1st Workshop on Deep Learning for Recommender Systems, pp. 7–10. Cited by: §2.3.1.
  • H. Cheng, L. Koc, J. Harmsen, T. Shaked, T. Chandra, H. Aradhye, G. Anderson, G. Corrado, W. Chai, M. Ispir, et al. (2016b) Wide & deep learning for recommender systems. In Proceedings of the 1st workshop on deep learning for recommender systems, pp. 7–10. Cited by: §1.1, §1.
  • F. Chollet et al. (2015) Keras. GitHub. External Links: Link Cited by: §3.1.
  • P. D. Clift and R. A. Plumb (2008) The asian monsoon: causes, history and effects. Vol. 288, Cambridge University Press Cambridge. Cited by: §1.
  • Y. Dash, S. K. Mishra, S. Sahany, and B. K. Panigrahi (2018) Indian summer monsoon rainfall prediction: a comparison of iterative and non-iterative approaches. Applied Soft Computing 70, pp. 1122–1134. Cited by: §1.1, §2.1, §2.1.
  • A. D. Dubey (2015) Artificial neural network models for rainfall prediction in pondicherry. International Journal of Computer Applications 120 (3). Cited by: §1.1.
  • V. Ducrocq, D. Ricard, J. Lafore, and F. Orain (2002) Storm-scale numerical rainfall prediction for five precipitating events over france: on the importance of the initial humidity field. Weather and Forecasting 17 (6), pp. 1236–1256. Cited by: §1.1.
  • D. Dutta, A. Kundu, and N. Patel (2013) Predicting agricultural drought in eastern rajasthan of india using ndvi and standardized precipitation index. Geocarto International 28 (3), pp. 192–209. Cited by: §1.1.
  • S. Dutta and G. Chaudhuri (2015) Evaluating environmental sensitivity of arid and semiarid regions in northeastern rajasthan, india. Geographical Review 105 (4), pp. 441–461. Cited by: §1.
  • Y. Enzel, L. L. Ely, S. Mishra, R. Ramesh, R. Amit, B. Lazar, S. Rajaguru, V. Baker, and A. Sandler (1999) High-resolution holocene environmental changes in the thar desert, northwestern india. Science 284 (5411), pp. 125–128. Cited by: §2.1.
  • L. Fan, S. Shin, Q. Liu, and Z. Liu (2013) Relative importance of tropical sst anomalies in forcing east asian summer monsoon circulation. Geophysical Research Letters 40 (10), pp. 2471–2477. Cited by: §1.
  • S. Gadgil, M. Rajeevan, and R. Nanjundiah (2005) Monsoon prediction – why yet another failure?. Current Science 88 (9), pp. 1389–1400. External Links: ISSN 00113891, Link Cited by: §1.1.
  • S. Gadgil, J. Srinivasan, R. S. Nanjundiah, K. K. Kumar, A. Munot, and K. R. Kumar (2002) On forecasting the indian summer monsoon: the intriguing season of 2002. Current Science 83 (4), pp. 394–403. Cited by: §1.1.
  • X. Glorot and Y. Bengio (2010) Understanding the difficulty of training deep feedforward neural networks. In In Proceedings of the International Conference on Artificial Intelligence and Statistics (AISTATS’10). Society for Artificial Intelligence and Statistics, pp. 249–256. Cited by: §3.1.2.
  • A. Goel and R. Singh (2006) Climatic variability and drought in rajasthan. In Advances in Geosciences: Volume 4: Hydrological Science (HS), pp. 57–67. Cited by: §1.
  • H. R. Goyal, K. K. Ghanshala, and S. Sharma (2021) Recommendation based rescue operation model for flood victim using smart iot devices. Materials Today: Proceedings. Cited by: §1.
  • J. Gu, Z. Wang, J. Kuen, L. Ma, A. Shahroudy, B. Shuai, T. Liu, X. Wang, G. Wang, J. Cai, et al. (2018) Recent advances in convolutional neural networks. Pattern Recognition 77, pp. 354–377. Cited by: §2.3.1.
  • A. Gulli and S. Pal (2017) Deep learning with keras. Packt Publishing Ltd. Cited by: §3.1.
  • J. Halbe, C. Pahl-Wostl, J. Sendzimir, and J. Adamowski (2013) Towards adaptive and integrated management paradigms to meet the challenges of water governance. Water Science and Technology 67 (11), pp. 2651–2660. Cited by: §1.
  • J. He, M. Wen, L. Wang, and H. Xu (2006) Characteristics of the onset of the asian summer monsoon and the importance of asian-australian “land bridge”. Advances in Atmospheric Sciences 23 (6), pp. 951–963. Cited by: §1.
  • K. He, X. Zhang, S. Ren, and J. Sun (2015a)

    Delving deep into rectifiers: surpassing human-level performance on imagenet classification

    .
    In

    Proceedings of the 2015 IEEE International Conference on Computer Vision (ICCV)

    ,
    pp. 1026–1034. Cited by: §3.1.2.
  • K. He, X. Zhang, S. Ren, and J. Sun (2015b)

    Delving deep into rectifiers: surpassing human-level performance on imagenet classification

    .
    In Proceedings of the IEEE international conference on computer vision, pp. 1026–1034. Cited by: §3.1.3.
  • J. Jang and S. Han (2011) Importance of monsoon rainfall in mass fluxes of filtered and unfiltered mercury in gwangyang bay, korea. Science of the total environment 409 (8), pp. 1498–1503. Cited by: §1.
  • K. Johny, M. L. Pai, and S. Adarsh (2020) Adaptive eemd-ann hybrid model for indian summer monsoon rainfall forecasting. Theoretical and Applied Climatology, pp. 1–17. Cited by: §1.1.
  • S. Kalsi, H. Hatwar, N. Jayanthi, S. Subramanian, B. Shyamala, M. Rajeevan, and R. Jenamani (2004) Various aspects of unusual behaviour of monsoon 2002. India Meteorol. Dep. Monogr 2, pp. 97. Cited by: §1.1.
  • D. KARAN (2016) AN empirical study of impact of drought on agricultural produce and use of inputs in rajasthan. EIJFMR. Cited by: §1.
  • L. P. Kerhoulas, T. E. Kolb, and G. W. Koch (2017) The influence of monsoon climate on latewood growth of southwestern ponderosa pine. Forests 8 (5), pp. 140. Cited by: §1.
  • M. Kim, S. Lee, and J. Kim (2020) A wide & deep learning sharing input data for regression analysis. In 2020 IEEE International Conference on Big Data and Smart Computing (BigComp), pp. 8–12. Cited by: §1.1.
  • D. P. Kingma and J. Ba (2014) Adam: a method for stochastic optimization. arXiv preprint arXiv:1412.6980. Cited by: §3.1.3.
  • S. Kulshreshtha, S. Sharma, and B. Sharma (2013) The majestic rajasthan: an introduction. In Faunal heritage of Rajasthan, India, pp. 3–37. Cited by: §2.1.
  • A. Lal Meena, P. Bisht, et al. (2020) Study of rainfall pattern in chaksu tehsil, jaipur, rajasthan, india. . Cited by: §1.1.
  • K. Lau and M. Li (1984) The monsoon of east asia and its global associations—a survey. Bulletin of the American Meteorological Society 65 (2), pp. 114–125. Cited by: §1.
  • M. Li and Q. Shao (2010) An improved statistical approach to merge satellite rainfall estimates and raingauge data. Journal of Hydrology 385 (1-4), pp. 51–64. Cited by: §1.1.
  • H. M. Meena, D. Machiwal, P. Santra, P. C. Moharana, and D. Singh (2019) Trends and homogeneity of monthly, seasonal, and annual rainfall over arid region of rajasthan, india. Theoretical and Applied Climatology 136 (3), pp. 795–811. Cited by: §1.1.
  • J. K. Meher and L. Das (2019) Gridded data as a source of missing data replacement in station records. Journal of Earth System Science 128 (3), pp. 1–14. Cited by: §2.2.
  • A. Montanari and G. Grossi (2008) Estimating the uncertainty of hydrological forecasts: a statistical approach. Water Resources Research 44 (12). Cited by: §1.1.
  • D. Mooley and B. Parthasarathy (1984) Fluctuations in all-india summer monsoon rainfall during 1871–1978. Climatic change 6 (3), pp. 287–301. Cited by: §1.
  • D. Mooley (1997) Variation of summer monsoon rainfall over india in ei-niños. Mausam. Cited by: §1.
  • N. Mundetia, D. Sharma, et al. (2015) Analysis of rainfall and drought in rajasthan state, india. Global Nest J 17 (1), pp. 12–21. Cited by: §1.
  • C. Naidu, K. Durgalakshmi, K. Muni Krishna, S. Ramalingeswara Rao, G. Satyanarayana, P. Lakshminarayana, and L. Malleswara Rao (2009) Is summer monsoon rainfall decreasing over india in the global warming era?. Journal of Geophysical Research: Atmospheres 114 (D24). Cited by: §1.
  • A. Ng et al. (2011)

    Sparse autoencoder

    .
    CS294A Lecture notes 72 (2011), pp. 1–19. Cited by: §1.1.
  • D. Pai, L. Sridhar, M. Rajeevan, O. Sreejith, N. Satbhai, and B. Mukhopadhyay (2014) Development of a new high spatial resolution (0.25 0.25) long period (1901–2010) daily gridded rainfall data set over india and its comparison with existing data sets over the region. Mausam 65 (1), pp. 1–18. Cited by: §2.2, §2.2, §3.2.1.
  • S. K. Pal and S. Mitra (1992) Multilayer perceptron, fuzzy sets, classifiaction. . Cited by: §2.1.
  • B. Parthasarathy, A. Munot, and D. Kothawale (1988) Regression model for estimation of indian foodgrain production from summer monsoon rainfall. Agricultural and Forest Meteorology 42 (2-3), pp. 167–182. Cited by: §1.
  • B. Parthasarathy, N. Sontakke, A. Monot, and D. Kothawale (1987) Droughts/floods in the summer monsoon season over different meteorological subdivisions of india for the period 1871–1984. Journal of Climatology 7 (1), pp. 57–70. Cited by: §1.
  • S. M. Pingale, D. Khare, M. K. Jat, and J. Adamowski (2014) Spatial and temporal trends of mean and extreme rainfall and temperature for the 33 urban centers of the arid and semi-arid state of rajasthan, india. Atmospheric Research 138, pp. 73–90. Cited by: §1.1.
  • B. Preethi, J. Revadekar, and R. Kripalani (2011) Anomalous behaviour of the indian summer monsoon 2009. Journal of earth system science 120 (5), pp. 783–794. Cited by: §1.1.
  • M. Rajeevan, J. Bhate, J. Kale, and B. Lal (2005) Development of a high resolution daily gridded rainfall data for the indian region. Met. Monograph Climatology 22, pp. 2005. Cited by: §2.2, §3.2.1.
  • K. Ray, P. Pandey, C. Pandey, A. Dimri, and K. Kishore (2019) On the recent floods in india. Current science 117 (2), pp. 204–218. Cited by: §1.
  • L. Ren, Z. Meng, X. Wang, R. Lu, and L. T. Yang (2020) A wide-deep-sequence model-based quality prediction method in industrial process analysis. IEEE Transactions on Neural Networks and Learning Systems 31 (9), pp. 3721–3731. Cited by: §1.1.
  • A. B. Roy and S. R. Jakhar (2002) Geology of rajasthan (northwest india) precambrian to recent. Scientific Publishers. Cited by: §2.1.
  • M. Saha, P. Mitra, and R. S. Nanjundiah (2016) Autoencoder-based identification of predictors of indian monsoon. Meteorology and Atmospheric Physics 128 (5), pp. 613–628. Cited by: §1.1.
  • M. Saha, A. Santara, P. Mitra, A. Chakraborty, and R. S. Nanjundiah (2021) Prediction of the indian summer monsoon using a stacked autoencoder and ensemble regression model. International Journal of Forecasting 37 (1), pp. 58–71. Cited by: §1.1.
  • A. Sahai, M. Soman, and V. Satyan (2000) All india summer monsoon rainfall prediction using an artificial neural network. Climate dynamics 16 (4), pp. 291–302. Cited by: §1.1, §2.1.
  • M. Shanker, M. Y. Hu, and M. S. Hung (1996) Effect of data standardization on neural network training. Omega 24 (4), pp. 385–397. Cited by: §2.2.
  • D. Sikka (2003) Evaluation of monitoring and forecasting of summer monsoon over india and a review of monsoon drought of 2002. Proceedings-Indian National Science Academy Part A 69 (5), pp. 479–504. Cited by: §1.1.
  • A. Singh, M. A. Kulkarni, U. Mohanty, S. Kar, A. W. Robertson, and G. Mishra (2012a) Prediction of indian summer monsoon rainfall (ismr) using canonical correlation analysis of global circulation model products. Meteorological Applications 19 (2), pp. 179–188. Cited by: §1.1.
  • B. Singh, D. Rajpurohit, A. Vasishth, and J. Singh (2012b) Probability analysis for estimation of annual one day maximum rainfall of jhalarapatan area of rajasthan, india. Plant Archives 12 (2), pp. 1093–1100. Cited by: §1.1.
  • P. Singh and B. Borah (2013) Indian summer monsoon rainfall prediction using artificial neural network. Stochastic environmental research and risk assessment 27 (7), pp. 1585–1599. Cited by: §1.1, §2.1.
  • P. Singh (2016) Applications of soft computing in time series forecasting. Springer. Cited by: §1.1.
  • P. Singh (2018a) Indian summer monsoon rainfall (ismr) forecasting using time series data: a fuzzy-entropy-neuro based expert system. Geoscience Frontiers 9 (4), pp. 1243–1257. Cited by: §1.1, §2.1.
  • P. Singh (2018b) Rainfall and financial forecasting using fuzzy time series and neural networks based model. International Journal of Machine Learning and Cybernetics 9 (3), pp. 491–506. Cited by: §1.1, §2.1.
  • M. Swaminathan (1998) Padma bhusan prof. P. Koteswaram First Memorial Lecture-23rd March, pp. 3–10. Cited by: §1.
  • J. E. Tierney, J. M. Russell, J. S. S. Damsté, Y. Huang, and D. Verschuren (2011) Late quaternary behavior of the east african monsoon and the importance of the congo air boundary. Quaternary Science Reviews 30 (7-8), pp. 798–807. Cited by: §1.
  • H. Upadhyaya (2014) Vulnerability and adaptation to climate change in the context of water resource with reference to rajasthan. Ph.D. Thesis, The IIS University. Cited by: §2.1.
  • S. P. Van, H. M. Le, D. V. Thanh, T. D. Dang, H. H. Loc, and D. T. Anh (2020) Deep learning convolutional neural network in rainfall–runoff modelling. Journal of Hydroinformatics 22 (3), pp. 541–561. Cited by: §2.1, §2.3.1.
  • G. T. Walker (1933) Seasonal weather and its prediction. Nature Publishing Group. Cited by: §1.
  • S. M. Yadav and N. K. Mangukiya (2021) Semi-arid river basin flood: causes, damages, and measures. In Proceedings of the Fifth International Conference in Ocean Engineering (ICOE2019), pp. 201–212. Cited by: §1.
  • S. K. Yadav, S. Nath, and S. Gautam (2018) Analysis of rainfall variability in western rajasthan, india. Journal of Pharmacognosy and Phytochemistry 7 (4), pp. 1592–1595. Cited by: §1.1.
  • G. Zhang, B. Eddy Patuwo, and M. Y. Hu (1998) Forecasting with artificial neural networks:: the state of the art. International Journal of Forecasting 14 (1), pp. 35–62. External Links: ISSN 0169-2070 Cited by: §1.1.