Introduction about water and estimating
crop water requirements
The main constituents of the Earth hydrosphere are the
water, which is the transparent, inorganic, odorless, and colorless chemical
substance. It is very significant for
all the known forms of life, even though it is providing organic nutrients and calories.
In the world economy, water plays a significant role. By the human, approximately
70% of the water is used, and it goes to agriculture. Mostly in the oceans and
the sea, the water covers approximately 70% of the Earth surface. So water is the basic need of the crops, and
proper supply of water to different crops according to their condition is very
important to determine because water is the necessity of a life of all elements.
The crop water requirements are defined as “ The
water depth in (mm) which is required to meet the consumptions of water by the Evapotranspiration’s
() through the disease of the free crop by growing the large fields under
the conditions of non-restricting oil, and it involves the fertility and soil
water plus it obtaining the full production of potential below the growing environments.
It is also very important to estimate the water requirement of every crop
because different crops have different production processes, and they need
water according to their required production processes. However, it is also
important to understand what water requirement for crops is. Whereas the
Evapotranspiration’s () is defined as the rate of Evapotranspiration’s () for the given crop is influenced through the growing stages of the
conditions of environments as well as the management of the crops. And the rate
of Evapotranspiration’s () in () could be computed as;
Whereas = latent
heat of vaporizations in ;
and is the net balance of energy ; ; is the mean air density ; specific heat constant ; the slope of saturation vapor ; aerodynamic resistance (Pereira & et al, 2013).
The purpose of the paper is to demonstrate the feasibility of
the Artificial Natural network (ANN), which is being applied for the estimation
of the crop water requirements. The crop water requirements are estimated by the use
of the FAO model crop because it is a major crop in the various regions of the
states. It is also used the Penman-Monteith method in the below equation for determining the reference of crop evapotranspiration and also gives
the matches values through the actual needs of the crop water and the use the
data of worldwide (Kumari & et al, 2017). The evapotranspiration is calculated a
s;
Method on Machine
Learning
Ø Deep learning
of
Method on Machine Learning
For the data analysis as
well as image processing Deep learning is the constitution of modern techniques
by the promising result of the great potential. In the different domains, deep
learning has been successfully applied, and it is entered in the domain of
agriculture. According to the investigations, there is 40 research that
employed deep learning methods for different agriculture and the production of the
challenges. Deep learning belongs to machine learning by the computational
fields, and it is similar to the ANN. Thus the Deep learning is regarding the
“deeper” neutral network, which also provides the hierarchical representations
of data through the means of different convolutions. It also presents the
larger learning capabilities as well as higher performance along with the precisions.
The strong benefits of deep learning are feature learning, which has the
automatic extractions of features from the raw data, then by these features
form high levels of the hierarchy, which is being formed through the compositions
of the lower levels of the features (Kamilaris, 2018).
Ø Long Short Term
Memory (LSTM) neural network
LSTM is the method which is
related to the memory cell as well as in what way it is related to the old
context by the new
context . Whereas is the memory
cell . then the input gate of the discharge level of the data has the memory
cell, which results is in the ( input gate ) then this input is used for the next
sequences. The first step to determine the LSTM is the explicit formula of the LSTM-RNN
is (Sudriani & et al, 2019);
Whereas the
There are some definitions
for the above equations is shown in the below screenshots;
The ions of the nutrients
solution are influenced through the accumulations by the time series factors like
the environment greenhouse, and the supply of water, as well as drainage of
water along with the plant growth. By using the LSTM methods, estimate the crop
requirements by the ANN and the meteorological variables.
Ø Gated recurrent units(GRU)
GRU is the newly developed variations for the
LSTM units, where both of the variants have the RNN. Then by the empirical
evidence, these methods have been proved through the effective and wide variety
of machine learning tasks like natural language processing. The GRU –RNN also
reduces the gating signals by the two forms of the LSTM models, then these two
gates are also known as the update gate . (Dey & et al, 2017) Then the model of
the GRU RNN is presented as;
Whereas
Then these two gates presented as;
The application of deep learning is also predicting
the water level for the estimation of the crop requirements. The forecasting
model GRU is developed the base the RNN
to predict the water level from one to the four-time steps.
Paper
2
Reservoir
inflow estimating by deep learning
Introduction
of Method on Machine Learning
Importance
of water
It
can be noted that water is one of the most important elements on earth. It is
also playing a vital role in the human body. The next thing is that there is no
survival without water. We can easily survive without food for some weeks, but
it will be extremely difficult to live for 2 days without water. This is
because every tissue and system in the body requires water for proper
functioning. Moreover, water is also
involved in carrying all nutrients all over the body and also providing oxygen
to the whole brain. Furthermore, water is also absorbing some important
minerals for our bodies. The next thing is that water can also flush out toxin
material and waste. This element is also helping to maintain the temperature of
the body. Additionally, water is extremely important for the soil to grow at a
certain temperature. If there is no water, then it will become extremely
difficult for the plant to perform the process of photosynthesis. Another thing
is that water is also one of the most important lubricants for the movement of
muscles. This shows that the value of water is extremely high. This is why the
world is working on solving the issues related to saving this element for
future use (Hans Lambers, 2008).
Importance
of Forecast of inflow of Method on Machine Learning
In
the flood season, there is a need for reservoir inflow forecasting that is
extremely important for proper management of the reservoir. This type of
forecasting is extremely important for such a season. This forecasting must be
accurate, and on time, this is because it will allow the dam managers to
release water on time to control the flood in the downstream areas. It can be
seen that in the past few years, there are different methodologies and models
are presented for the inflow forecasting with accurate and absolute results.
There are also some defects in some models for forecasting. For that case,
there is a need to overcome the weakness with perfection. Then a variation
analog method is implemented that can be considered as analog-based
forecasting. Due to this forecasting, it will become extremely easy to save
different downstream areas (Amnatsan, Yoshikawa, & Kanae, 2018).
Importance
of Deep learning
Before
going brief in the subject, it is important to get useful information related
to the importance of deep learning. It can be noted that its importance is
extremely recognized by the average person. Besides this, according to the
different researchers, its benefits are playing a game-changing role. The next
thing is that in real-world applications, it is extremely important because it
is capable of proving to be practical. This system is also involved in using
unsupervised learning to increase its important benefits. In contrast, the
machine learning program is involved in depending on various labels for the
processing of data in an efficient way. This shows that it will require extra
labor to process data, but on the other hand, deep learning can easily learn
with the help of structureless data (Ian Goodfellow, 2016).
There
are different methods for reservoir inflow estimating through deep learning.
All of these methods are given below.
Bidirectional
LSTM of Method on Machine Learning
It
can be noted that it is one of the most advanced types of RNN. It is totally
related to such information that flows from one part to the other and vice
versa. The world bidirectional shows that the information moves in both
directions. The next thing is that this method is useful for time series
forecasting. It can be noted that there are different types of LSTM models, and
all of these models can be used for solving the time series forecasting
problem. But it can be seen that if there is a need for bidirectional LSTM,
then it will solve the sequence prediction problem for the whole system. It is
extremely important for the whole system because it will allow the model to
learn useful information related to the input sequence. Then it will help the
system in both directions. For the implementation of the model, it must be
important that the input values must be arranged in series form. This model
will check each and every value in both directions. Then by combining these
values, it will become extremely easy to solve the time-variant problem in the
reservoir inflow before flooding (Brownlee, 2020).
Gated
recurrent unit of Method on Machine Learning
The
gated recurrent unit is also one of the efficient methods of deep learning. For
efficient use of the water resources for humans and also other organization,
the reservoir is one of the most important engineering measures. For that case,
there is a need for an effective operating plan for implementing its functions.
Now for exploring its main application, there are different methods applied
related to deep learning, and the Gated recurrent unit is one of them. This
unit is implemented for the prediction of outflow through the reservoir. Now
there is a need for such a forecasting model that is completely based on the
recurrent neural network. This model will help to predict the level from
four-time ahead.
This will provide efficient progress for
controlling the flow of water and refrain from floods. The next thing is that
this method is also considered as a simple variation of the LSTM network. This
is because both are involved in producing the same kind of results. But there
is some similarity present between these models. This is the reason why this
system is involved in solving the gradient problem in a proper way. This
structure can be demonstrated in the figure given below
According
to the researcher, it can be noted that there is no cell that contains
separated memory like considered with LSTM cell. Moreover, the next thing is
that there are only two gating layers are present in GRU. The one is related to
update, and the next one is the reset gate. On the other hand, the reset gate
is involved in determining the information from a previous memory. Also, the
update gate works similarly like in the LSTM cell. Now the hidden state in the
GRU is defined in the form of differential equations that are given below (Le, Ho, & Lee, 2019)
Long
short-term memory of Method on Machine Learning
In
the above section, there is proper information related to the importance of
water. But if there is an excessive amount of water assembles at a given place,
then it will result in flooding. From refraining from such an issue there is a
need for forecasting for integrating the water resource in an effective way.
For that case, there is a need flood forecasting network, and it can be done
easily through the help of long short-term memory neural network. In that
particular network, there is information related to daily discharge, and
rainfall data can be used as the input. The next thing is that the
characteristics data set may affect the working of the model.
The
next most important thing about this model is that its predictive ability is
extremely impressive. According to the
information, this method was presented by Schmidhuber and Hochreiter.
Furthermore, this method is solving the problems that are present in the RNN.
This can be done through the help of additional interaction. This means that
LSTM is one of the upgraded types of RNN. Moreover, this model is organized and
arranged in the form of a chain. But on the other hand, the repeating module
has various structures. The next thing is that it also has four interacting
layers that contain a unique method for communication. This system can be
defined in the form figure
The
next thing is that a proper LSTM network
consists of different memory blocks that are named as cells. All of these two
streams are transferred to the next
cell, hidden, and state cell. For the main chain of data flow, state cell is
used, and the next thing is that the data can be easily added and remove from
the cell in a proper way. The proper LSTM network can be constructed by
identifying the information that is not applied from the cell. In that network,
the forget gate is playing an important role, and this can be defined as
For
updating the new cell, this information can be updated in the new cell through
Simple
recurrent neural network of Method on Machine
Learning
It
can be noted that for forecasting the flood, the simplest model was presented
in the late 1980s. it is also known as the traditional method for forecasting.
The structure of this method consists of different layers that include input,
hidden, and output layers. All of these layers are extremely important for
estimating the time. The next thing is that it also contains chain-like structures.
Moreover, the next thing is that it also contains a feedback loop, and this
loop is involved in accepting the sequence of inputs (Le, Ho, Giha, & Sungho, 2019).
This
can be explained with the help of figure
References of Method on Machine Learning
Amnatsan, S., Yoshikawa, S., & Kanae, S. (2018). Improved
Forecasting of Extreme Monthly Reservoir Inflow Using an Analogue-Based
Forecasting Method: A Case Study of the Sirikit Dam in Thailand. Water 2018.
Brownlee, J. (2020). How
to Develop LSTM Models for Time Series Forecasting. Retrieved from
https://machinelearningmastery.com/how-to-develop-lstm-models-for-time-series-forecasting/
Dey, R., & et al.
(2017). Gate-variants of Gated Recurrent Unit (GRU) neural networks. IEEE 60th
International Midwest Symposium on Circuits and Systems (MWSCAS).
Hans Lambers, F. S.
(2008). Plant Physiological Ecology. Springer Science & Business Media,.
Ian Goodfellow, Y. B.
(2016). Deep Learning. MIT Press.
Kamilaris, A. (2018).
Andreas Kamilaris Deep learning in agriculture: A survey. Computers and
Electronics in Agriculture, 70–90.
Kumari, M., & et al.
(2017). Crop Water Requirement, Water Productivity and Comparative Advantage of
Crop Production in Different Regions of Uttar Pradesh, India. International
Journal of Current Microbiology and Applied Sciences, 6(7), 2043-2052.
Le, X.-H., Ho, H. V.,
& Lee, G. (2019). APPLICATION OF GATED RECURRENT UNIT (GRU) NETWORK FOR
FORECASTING RIVER WATER LEVELS AFFECTED BY TIDES. Proceedings of the 10th
International Conference on Asian and Pacific Coasts.
Le, X.-H., Ho, H. V.,
Giha, L., & Sungho, J. (2019). Application of Long Short-Term Memory (LSTM)
Neural Network for Flood Forecasting. water 2019.
Pereira , L., & et
al. (2013). Crop Water Requirements. Reference Module in Earth Systems and
Environmental Sciences. Change History.
Sudriani, Y., & et
al. (2019). Long short term memory (LSTM) recurrent neural network (RNN) for
discharge level prediction and forecast in Cimandiri river, Indonesia. IOP
Conference Series: Earth and Environmental Science