208
S. Namany and T. Al-Ansari
Risk assessment is broadly defined as the steps followed to identify the probability
of potential losses, through the analysis of vulnerabilities and hazards governing the
surrounding environment of a system, livelihoods or people [92]. The aim of risk
assessment and prediction within resource management to determine and predict
the possible risks and uncertainties that can cause systemic failures and reduce the
system’s adaptive capacity. In addition, it can enable the effective evaluation of
existing control actions and re-adjust according to the predicted risks. Any failure
to account for risks in one of the resource sectors can cause a cascade failure due
to the inherent interlinkages between the resource systems. One example of cascade
failure can be illustrated with the energy to food interaction. For instance, volatility
in gas prices, which is considered a risk in the energy sector, can influence the
energy production, which can alter the supply of the electricity, a critical component
for irrigation technologies, production of fertilisers and to support farms [93]. This
sequence of disturbances can reduce crop production and engender food shortages.
To avoid such undesirable results, considering volatility in prices and uncertainties
while making decisions can reduce potential losses. One of the tools commonly
used to capture risks associated with the food sector and its associated energy and
water sectors is the Monte Carlo simulation [94]. Liu et al. utilised the Monte Carlo
simulation with a stochastic programing model to assess the impact of water inflow
uncertainty on the planning of irrigation water and crops. The purpose of the study
is to alleviate risk of food shortages due to uncertain water availability [95]. Monte
Carlo also supports the assessment of financial risks associated with food-related
investments and projects. In this regard, Kadigi et al. adopted a Monte Carlo simulation that incorporates the risk preferences for decision-makers in the deployment of
enhanced rice farming practices and technologies. To this end, the economic feasibility of each project is assessed and used as a criterion to select the most economically
profitable in terms of rice productivity [96]. In addition to simulation, data-driven
models are often used to account for risks and uncertainty in modelling resource
systems. Machine learning is one category of data-driven models that has shown its
effectiveness in accurately modelling large-scale intricate problems. While simulation tools begin with unknown input variables, representing the source of uncertainty,
the model that transforms inputs into outputs, in the form of mathematical relationships and equations, is known in advance. Alternatively, machine learning inputs
are known in advance while the model is unknown. In this case, historical data is
used to draw the relationship between the inputs and outputs, making the model the
only source of uncertainty. In other words, simulation models use the knowledge of
experts about models to predict outputs, while machine learning design the models
to predict results [97]. Using machine learning for food security purposes, Kumar
et al. used a machine learning algorithm to predict crop yields. The purpose of the
model is to determine the set crops with the highest productivity or yield while
considering a range of factors, such as weather conditions, crop type, soil type and
water density [98]. Similarly, Kuwata and Shibasaki used a deep learning approach, a
sub-category of machine learning to predict crop yields, presenting a more accurate
version of the previous crop yield model by reducing reliance on input data [99].
Another application of machine learning in agriculture is presented by Bagheri et al.
S. Namany and T. Al-Ansari
Risk assessment is broadly defined as the steps followed to identify the probability
of potential losses, through the analysis of vulnerabilities and hazards governing the
surrounding environment of a system, livelihoods or people [92]. The aim of risk
assessment and prediction within resource management to determine and predict
the possible risks and uncertainties that can cause systemic failures and reduce the
system’s adaptive capacity. In addition, it can enable the effective evaluation of
existing control actions and re-adjust according to the predicted risks. Any failure
to account for risks in one of the resource sectors can cause a cascade failure due
to the inherent interlinkages between the resource systems. One example of cascade
failure can be illustrated with the energy to food interaction. For instance, volatility
in gas prices, which is considered a risk in the energy sector, can influence the
energy production, which can alter the supply of the electricity, a critical component
for irrigation technologies, production of fertilisers and to support farms [93]. This
sequence of disturbances can reduce crop production and engender food shortages.
To avoid such undesirable results, considering volatility in prices and uncertainties
while making decisions can reduce potential losses. One of the tools commonly
used to capture risks associated with the food sector and its associated energy and
water sectors is the Monte Carlo simulation [94]. Liu et al. utilised the Monte Carlo
simulation with a stochastic programing model to assess the impact of water inflow
uncertainty on the planning of irrigation water and crops. The purpose of the study
is to alleviate risk of food shortages due to uncertain water availability [95]. Monte
Carlo also supports the assessment of financial risks associated with food-related
investments and projects. In this regard, Kadigi et al. adopted a Monte Carlo simulation that incorporates the risk preferences for decision-makers in the deployment of
enhanced rice farming practices and technologies. To this end, the economic feasibility of each project is assessed and used as a criterion to select the most economically
profitable in terms of rice productivity [96]. In addition to simulation, data-driven
models are often used to account for risks and uncertainty in modelling resource
systems. Machine learning is one category of data-driven models that has shown its
effectiveness in accurately modelling large-scale intricate problems. While simulation tools begin with unknown input variables, representing the source of uncertainty,
the model that transforms inputs into outputs, in the form of mathematical relationships and equations, is known in advance. Alternatively, machine learning inputs
are known in advance while the model is unknown. In this case, historical data is
used to draw the relationship between the inputs and outputs, making the model the
only source of uncertainty. In other words, simulation models use the knowledge of
experts about models to predict outputs, while machine learning design the models
to predict results [97]. Using machine learning for food security purposes, Kumar
et al. used a machine learning algorithm to predict crop yields. The purpose of the
model is to determine the set crops with the highest productivity or yield while
considering a range of factors, such as weather conditions, crop type, soil type and
water density [98]. Similarly, Kuwata and Shibasaki used a deep learning approach, a
sub-category of machine learning to predict crop yields, presenting a more accurate
version of the previous crop yield model by reducing reliance on input data [99].
Another application of machine learning in agriculture is presented by Bagheri et al.