You are on page 1of 10

Computers and Electronics in Agriculture 172 (2020) 105376

Contents lists available at ScienceDirect

Computers and Electronics in Agriculture


journal homepage: www.elsevier.com/locate/compag

Development of a closed-loop irrigation system for sugarcane farms using T


the Internet of Things
E. Wanga, , S. Attardb, A. Lintonc, M. McGlincheyd, W. Xianga, B. Philippaa, Y. Everinghama

a
James Cook University, 1 James Cook Dr, Douglas, QLD 4814, Australia
b
AgriTech Solutions, 343 Old Clare Road, Ayr, QLD 4807, Australia
c
Think Pink Hire, Home Hill, QLD 4806, Australia
d
SQR Software, Cascades Pietermaritzburg, 3202 Kwa-Zulu Natal, South Africa

ARTICLE INFO ABSTRACT

Keywords: Better irrigation practices, through increased water use efficiency, can deliver economic, environmental and
Irrigation management social benefits to agro-ecological systems. Increasingly, farmers world-wide are turning to automated irrigation
Sugarcane farming systems to save them a significant amount of time by remotely turning on and off pumps and valves.
Internet of Things Unfortunately, automated irrigation systems on their own do not provide insight on (i) the amount and timing of
irrigation required by the crop or, (ii) how irrigation schedules should change with soil type, farm management
and climate. To unravel these complex interactions, an irrigation decision support tool is needed. However, due
to the high frequency of irrigations across dozens of irrigation blocks, and the need to irrigate almost all year-
round, irrigation decision support tools can be very tedious and time-consuming for farmers to use daily. For
these reasons, many farmers will not adopt these tools, and in doing so, fail to optimise irrigation use efficiency
due to the multi-factorial nature of the farming system. This paper describes a cybernetic closed-loop solution
that was piloted on a sugarcane farm in north-eastern Australia. The solution seeks to improve irrigation
management by seamlessly integrating the WiSA automated irrigation system with the IrrigWeb irrigation de-
cision support tool. Specifically, an Uplink program and a Downlink program were implemented in the pilot
study. The Uplink program saved the farmer a significant amount of time. Instead of the farmer manually
entering in records, Uplink uploaded irrigation and rainfall data directly to the irrigation decision support tool.
The Downlink program calculated and applied irrigation schedules automatically using IrrigWeb, but also in-
corporated practical constraints, such as energy, pumping capability, irrigation priorities and farmer irrigation
preference. The simulation results demonstrated that the developed closed-loop solution could effectively
manage irrigation scheduling by incorporating irrigation decision support tools with practical constraints.
Systems that increase water use efficiency can deliver practical, profitable and environmental benefits to irri-
gated agricultural systems worldwide.

1. Introduction management in agriculture is essential for increasing crop yield and


decreasing costs while contributing to environmental sustainability.
Water is a highly valued commodity by multiple entities. This is The Internet of Things (IoT) (Atzori et al., 2010) emerges as the
especially the case in agriculture where crops require water for irriga- natural choice for smart irrigation management applications. IoT is the
tion, and municipalities compete for the same resource. Agriculture inter-networking of physical devices, vehicles or other items, embedded
accounts for nearly 70% of all freshwater withdrawals and over 90% of with electronics, software, sensors, actuators, and communication that
total consumptive water use (Siebert, 2010). Despite large investments enable these objects to “talk” and exchange data – or simply put,
and subsidies, irrigation performance indicators are falling short of making normal things smarter by connecting them (Atzori et al., 2010).
expectations for yield increases, area irrigated and technical efficiency IoT technology is driving the concept of precision agriculture (Dong
in water use (Turral et al., 2010). More than half of the water diverted et al., 2013). The IoT can transform the industry and enabling farmers
or pumped for irrigation is wasted (Schlager, 2001; Tiwari and Dinar, to contend with the enormous challenges they face (Salam, 2020).
2002; World Water Assessment Programme, 2012). Smart irrigation Combining the technologies of radio telemetry, sensors and big data


Corresponding author.
E-mail address: eric.wang@jcu.edu.au (E. Wang).

https://doi.org/10.1016/j.compag.2020.105376
Received 4 December 2019; Received in revised form 17 March 2020; Accepted 17 March 2020
0168-1699/ © 2020 Elsevier B.V. All rights reserved.
E. Wang, et al. Computers and Electronics in Agriculture 172 (2020) 105376

analytics, the IoT-powered irrigation system can reduce water use and crop start dates, and weather information, the system cannot devise a
labour while improving crop quality and quantity (Vellidis, 2016). schedule to increase the chances of obtaining a larger crop with less
Sugarcane is the world’s largest cultivated crop by production water. To achieve this, the automated system would need to be in tune
quantity, with 1.9 billion tonnes produced on about 27.3 million hec- with the crop’s water needs, which can be simulated by the decision
tares in 2017, in more than 90 countries (Food and Agriculture support tool. For reasons outlined above, there are practical constraints
Organization, 2017). Sugarcane is a plant that originated in wet tropical which make this a feasible solution.
regions. Therefore, to achieve maximum productivity, it requires an In this paper, the development of an integrated modelling approach
abundant supply of water from either rainfall or irrigation. The need for involving the coupling of an automated irrigation system (WiSA) with
irrigation has been recognised in sugarcane for over 100 years and an irrigation decision support tool (IrrigWeb, www.irrigweb.com) to
varies by region. In Brazil, sugarcane is mainly cultivated under rainfed improve irrigation management is described. The approach offers scope
conditions, with only 1% of the crop area irrigated (Battie Laclau and for including closed-loop feedback to apply the right amount of water at
Laclau, 2009), while in some countries, such as India, 93% of the total the right time in a practical and effortless way.
sugarcane-cultivated area is under irrigation (Shrivastava et al., 2011). Specifically, an Uplink program is designed to connect WiSA to
In Australia and South Africa, about 40–60% of sugarcane areas depend IrrigWeb to allow effortless irrigation record keeping and data im-
on some type of irrigation (Inman-Bamber and Smith, 2005; Carr and porting. On the other hand, a Downlink program is developed to per-
Knox, 2011; Singels et al., 1998; M.P. Holden JR, 2013). form efficient irrigation scheduling to WiSA through IrrigWeb. The
Areas with low levels of effective rainfall see the greatest response integration of the Uplink and Downlink programs embedded in the
to applied irrigation, while areas with high amounts of effective rainfall proposed closed-loop irrigation system offers a novel solution using
are unlikely to benefit economically from irrigation. However, climate modern technology that can transform sugarcane irrigation manage-
change threatens the sustainability of most rainfed sugarcane farming ment in a way that is practical, profitable and environmentally friendly.
systems (Knox et al., 2010). Rainfed sugarcane farming systems are To test the unity of the framework and the impacts of the precision
gradually being replaced by irrigated farming systems. Furthermore, irrigation system, the methodology was applied to a case study su-
within a context of increasing water scarcity, climate change also ex- garcane farm in the Burdekin region, Australia’s largest sugarcane
acerbates the current supply-demand imbalance and has a significant district.
impact on irrigation water requirements. This poses severe challenges The remaining paper is organised as: the proposed system design is
to conventional low-efficiency irrigation systems and urges the transi- presented in Section 2. The results from the simulation are shown in
tion to climate-informed irrigation systems (Everingham et al., 2008). Section 3. Finally, the paper is concluded in Section 4.
Surface, overhead and drip irrigation methods are most commonly
used to irrigate sugarcane crops depending on physical characteristics 2. Materials and methods
of the location, economic considerations and social and other con-
siderations (Carr and Knox, 2011). Furrow irrigation is a form of surface 2.1. The test farm
irrigation and is the most widely used irrigation system for sugarcane
due to its simplicity and low cost (M. P. Holden JR, 2013). However, To evaluate the performance of the proposed closed-loop irrigation
the major drawback of furrow irrigation is low water use efficiency. system, a 100-ha test farm located near Home Hill, Queensland was
With each irrigation event, any excess irrigation is lost via runoff and/ selected for the 2018 growing season. This irrigation area is serviced by
or deep drainage. Significant economic benefit, through reduced water a combination of both pressurised (sub-surface drip) and furrow irri-
and energy costs, would be realised through improved irrigation effi- gation systems. The 15 irrigation blocks on this farm, are divided into
ciency. Sustainability of the region, through less deep drainage im- three hydraulic groups, as shown in Fig. 1. The downriver hydraulic
pacting upon the rising groundwater level and reduced agrichemicals group consisted of two irrigation blocks under furrow irrigation, the
and sediment loss to the environment, would also result through im- upriver hydraulic group of 9 irrigation blocks in furrow irrigation, and
proved irrigation management practices. the remaining four irrigation blocks under a pressurised irrigation
Most of the commercial automated irrigation systems offered by the system.
market are programmed to irrigate at time intervals for predefined Australian sugarcane cropping practices follow a 5- to 6-year cycle.
periods of time. The irrigation schedule is usually based on the user’s The crop is harvested on a ~12- to 13-month time period. In the
empirical knowledge of crop water needs, soil characteristic, and Burdekin region, the harvest season runs from June to November. After
weather factors (Romero et al., 2012). However, an efficient automated each harvest, the crop re-establishes from the same rootstock (ra-
system must describe all the main factors that are involved in the plant tooning) until the third or fourth ratoon whereupon the rootstock is
growing process. Automated irrigation systems on their own do not ploughed out and replanted. Table 1 shows the planting season and
provide (i) any insight on the amount and timing of irrigation required harvesting periods of the sugarcane crop in the test farm.
by the crop and (ii) how irrigation schedules should change with soil The simplest and most robust method for irrigation scheduling, i.e.
type, farm management and climate. All the automated irrigation sys- deciding when to irrigate and how much water to apply, is based on the
tems that have been developed so far do not use an agriculture/crop water balance (Singels and Smith, 2006). The soil water deficit (depth
model to calculate the quantity of water for irrigation. of water required to bring the soil to field capacity) is calculated using
To unravel these complex interactions, another smart tool is needed. ET (crop water demand) and rainfall data. The rules are based on the
Decision support tools such as APSIM (Keating, 2003), CANEGRO critical soil water deficit (SWD) (or SWD threshold used in IrrigWeb),
(Singels and Bezuidenhout, 2002), or WATERSENSE (Inman-Bamber the amount of water that the crop can extract in the rooting depth
et al., 2007, 2007) can provide accurate predictions of crop develop- before water stress occurs. By knowing the daily values of inflow
ment and soil water balance (Inman-Bamber and McGlinchey, 2003). (rainfall or irrigation) and outflow (crop water use), the daily water
Therefore, they are turned to as a solution to this problem (Chauhan balance can be calculated using the crop model. As soon as the accu-
et al., 2013). However, these decision support tools require frequent mulated SWD exceeds the critical value, irrigation water is supplied to
user-interactions to obtain precise irrigation scheduling. As such, the maintain an optimum water balance for plant growth.
time commitment to continuously update the information is a sig- The irrigation on a commercialised farm, however, is constrained by
nificant hurdle to ongoing use and ultimately leads to failed adoption many other factors. The three main factors are, (1) Energy tariff, which
practices. Automated irrigation, on the other hand, is becoming more restricts of the number of hours that could be used for irrigation each
attractive to farmers because of the conveniences it offers. However, day. The energy tariff used in the test farm restricts 10-hour off-peak
since automation does not consider farm characteristics, crop class, time during weekdays, and 24-hour off-peak time during weekends; (2)

2
E. Wang, et al. Computers and Electronics in Agriculture 172 (2020) 105376

Fig. 1. The layout of the test farm.

Pumping capacity, which restricts the number of paddocks that can be (https://www.irrigatewisa.com.au/)) to the irrigation decision support
irrigated at the same time. The irrigation system in the test farm can tool (e.g., IrrigWeb (http://www.irrigweb.com/)). The Downlink pro-
only irrigate one irrigation block at a time for the furrow irrigation gram automatically downloads, calculates and applies irrigation sche-
blocks. Therefore, the number of hours required to fill the soil capacity dules from the irrigation support tool to the automated irrigation
(shown in the last column of Table 1) will affect the irrigation sche- system. The framework of the closed-loop irrigation system is shown in
duling during the practical operation; and (3) Irrigation priority, which Fig. 2.
defines the sequence of irrigation when multiple irrigation blocks are
required to be irrigated at the same time (i.e., give priority to younger
crops). Ideally, the closed-loop irrigation system will closely follow the 2.3. IrrigWeb
irrigation schedule from IrrigWeb, while taking into the consideration
of the above practical constraints of the farm. IrrigWeb is a decision support tool for the sugar industry. IrrigWeb
The historical water balance data, including daily SWD, water ap- uses CANEGRO to simulate soil water deficit from weather data, farm
plied, crop water use and rainfall, in the 2018 growing season, are management and soil parameters, which is deemed to be a good re-
extracted from IrrigWeb. The purpose of the simulation is to demon- presentation of in-field measurements (Inman-Bamber and Smith, 2005;
strate the closed-loop irrigation can automatically upload the irrigation Inman-Bamber et al., 2007, 2007; Inman-Bamber and McGlinchey,
data to IrrigWeb, download the water balance data from IrrigWeb, 2003; Singels, 2010). Based on the simulation, IrrigWeb then provides
calculate the irrigation schedule for each irrigation block over the farmers with current and local advice on sugarcane development and
period of the whole growing season, and be able to successfully im- water use for their specific fields. The tool combines crop water-use
plement the irrigation schedule to the WiSA automated irrigation estimates with user-defined irrigation system constraints and crop-cycle
system. inputs to schedule future irrigation events. At the initialisation stage,
the user must input farm information, such as area, crop class, soil type,
row configuration, irrigation management rule, irrigation deficit and
2.2. Design and overview of the system harvest date. IrrigWeb acquires meteorological data daily from a third-
party source such as the Queensland Government’s SILO application
The framework for the closed-loop irrigation system is made of two (https://www.longpaddock.qld.gov.au/silo/), along with the weather
components, i.e., the Uplink and Downlink programs. The Uplink pro- data from local weather stations, and uses the CANEGRO model to
gram automatically extracts, calculates and uploads the daily irrigation calculate crop growth information, such as daily soil water deficit, crop
and rainfall data from the automated irrigation system (e.g., WiSA water use, crop stress and biomass. As for irrigation, IrrigWeb uses soil

Table 1
Planting seasons and harvesting periods of sugarcane crops in the test farm. P = plant crop Ri is the ratoon number.
Irrigation block Area Crop Row Config. Prev. Harvest Date Dry Off Date Harvest Date Irrig. Type Irrig. Rule Irrig. hours

D1 14.11 R3 1.8 m Dual 15/08/17 24/05/18 26/07/18 Drip


D2 14.09 P 1.8 m Dual 22/04/17 19/06/18 3/08/18 Drip
D4 2.00 R2 1.8 m Dual 15/08/18 25/09/18 5/10/18 Drip
Set 1 5.32 P 1.8 m Dual 20/04/17 5/09/18 28/09/18 Furrow 30 mm/3-day 6.8
Set 2 8.50 R2 1.8 m Dual 26/09/17 29/10/18 7/11/18 Furrow 30 mm/3-day 10.9
Set 3 2.73 R1 1.8 m Dual 15/08/17 28/08/18 26/09/18 Furrow 30 mm/3-day 3.5
Set 4 3.32 R1 1.8 m Dual 15/08/17 28/08/18 26/09/18 Furrow 30 mm/3-day 4.3
Set 5 4.80 R1 1.8 m Dual 1/10/17 11/09/18 28/09/18 Furrow 30 mm/3-day 6.2
Set 6 3.93 R2 1.8 m Dual 2/10/17 1/09/18 5/10/18 Furrow 30 mm/3-day 5.0
Set 7 3.81 R3 1.8 m Dual 25/09/17 25/09/18 5/10/18 Furrow 30 mm/3-day 4.9
Set 8 3.58 R3 1.8 m Dual 25/09/17 25/09/18 5/10/18 Furrow 30 mm/3-day 4.6
Set 9 7.09 R1 1.8 m Dual 18/07/17 18/07/18 26/07/18 Furrow 80 mm/7-day 24.2

3
E. Wang, et al. Computers and Electronics in Agriculture 172 (2020) 105376

Fig. 2. The framework of the closed-loop irrigation system.

water deficit and estimated crop water use to provide the amount of 2.5. Uplink
water required for each irrigation block. Farmers can then use this in-
formation to plan their irrigation management and apply irrigation The Aqualink software runs the WiSA irrigation system records and
schedules to their automated irrigation tools. stores all the irrigation information on a local computer. It uses the
NexusDB Database (https://www.nexusdb.com) to store the informa-
tion of pumps, valves and flow meter, the operation records of pumps
2.4. WiSA and valve, and the flow meter readings. The Uplink program is designed
to extract these data from the database, calculate the daily water ap-
WiSA is an automated irrigation system that has been successfully plied for each irrigation block, and upload the irrigation and rainfall
deployed in the Burdekin region (Gillies et al., 2017). Each farm is record to IrrigWeb, as shown in Fig. 4.
divided into multiple irrigation blocks, where one or two irrigation The information flow of the Uplink program includes: (1) WiSA
blocks are connected to the same field unit. The WiSA auto-irrigation passes the data from all the connected devices, including pumps, valves,
system uses wireless communications to monitor water flow, water flow meters, pressure transducers and moisture probes to Aqualink
pressure, soil moisture, weather information to control pumps and every 3 s; (2) Aqualink stores the incoming data from WiSA in the
valves. The framework of a WiSA automated irrigation system is shown database with information, such as (a) start and finish time for pumps
in Fig. 3, which consists of: (1) Base station: conveys information be- and valves; and (b) timestamp and flow rates for flow meters; (3) The
tween field units and the computer; (2) Pump unit: receives the control Uplink program interrogates Aqualink database on an hourly basis,
signal from the base station, and controls the on/off of pumps; (3) Field extracts the data for pumps, valves and flow meters, and calculates the
unit: (a) collects and convey sensor data to the base station; and (b) irrigation and rainfall information each day; (4) The Uplink program
receives the control signal from the base station, and controls the on/off uploads the irrigation and rainfall records for the last seven days to an
of valves; (4) Weather station: collects weather information (e.g., IrrigWeb server on an hourly basis; and (5) IrrigWeb requests the seven-
temperature, wind speed and rainfall); and (5) Aqualink: a software day irrigation and rainfall records on the server and updates each
program that manages data storage and display, irrigation manage- corresponding user’s IrrigWeb.
ment, and irrigation report.

Fig. 3. WiSA automated irrigation system.

4
E. Wang, et al. Computers and Electronics in Agriculture 172 (2020) 105376

Fig. 4. The workflow of the Uplink program.

2.5.1. Water applied calculation irrigation event with the timestamp of each flow meter record (Fig. 5b,
The amount of water applied to each irrigation block can be cal- Step 6 and 7).
culated as the recorded flow rate on the valve flow meter multiplied by the
record interval then multiplied by the number of records. Since in the test 2.6. Downlink
farm, there are no flow meters at each valve, the flow rate for each
valve can be approximated to be proportional to the area. IrrigWeb can accurately estimate the amount of water required for
Mathematically, we can generalise the calculation of the amount of each irrigation block for the next 7 days by simulating the plant growth.
water applied and develop an algorithm to implement this calculation However, the amount of water required suggested by IrrigWeb is an
in the Uplink program. Let Ti, n {Ti,1, Ti,2, , Ti, N |i {1, I }, n {1, N }} ideal irrigation value, while the actual water that can be applied will be
and Fi, n {Fi,1, Fi,2, , Fi, N |i {1, I }, n {1, N }} be the timestamp and constrained by many practical factors, such as irrigation system design,
flow rate for the nth record of flow meter i , where I is the total energy constraint and irrigation priorities. Hence, the Downlink pro-
number of flow meters in the hydraulic group, and N is the gram is designed to take the scientific simulation results and combine
total number of records for flow meter i . Defined with these practical constraints to generate an efficient and practical
by T jS, m {T jS,1, T jS,2, , T jS, M |j {1, J }, m {1, M }} and TjF, m irrigation schedule, as shown in Fig. 6.
{TjF,1, TjF,2, , TjF, M |j {1, J }, m {1, M }} the start time and finish time
for the mth irrigation event of valve j in one day, where J is the total 2.6.1. Irrigation scheduling
number of irrigation blocks and M is the total number of irrigation The Downlink program downloads the soil water deficit (denoted by
events for valve j . Here, the normalised flow rate for record Fi, n is de- Sj, d 1, j {1, J } ) of yesterday and expected daily crop water use of
fined as the recorded flow rate divided by the sum of areas Aj of irri- today (denoted by Cj, d ) from IrrigWeb for each irrigation block. The soil
gation blocks j opened at this timestamp, represented as follows: water deficit at the end of today can be estimated as:
Fi, n
Fi, n = (litre/second/ha), JO = {j|T jS, m < Ti, n < TjF, m} Sj, d = Sj, d 1 Cj, d (4)
j JO
Aj (1)
and the water required is
The total amount of water applied (litre) in one irrigation event Wj, m
Wj, d = Sj, d THj (5)
for an irrigation block can be calculated as the summation of the nor-
malised recorded flow rate which satisfies Ti, n {T jS, m, TjF, m} multiplied where THj is the soil water deficit threshold for valve j . The required
by the record interval (which is set to 5 min in Aqualink), then multi- number of hours Hj, d for each irrigation block can then be calculated as:
plied by the area of irrigation block j , which can be represented as:
Hj, d = Wj, d/ Fj (6)
Wj, m = Aj
I
i=1 ( N
n=1 )
Fi, n × 300 (litre)
(2) where Fj is the design flow rate for valve j .
The irrigation shift is determined by the following factors, the
while the timestamp Ti, n of each normalised flow rate record Fi, n is
number of available off-peak hours, the number of required irrigation hours
within {T jS, m, TjF, m} . Thus, the total amount of water applied (mm) for
of each irrigation block, the capacity (i.e., the rated flow rate) of the pump,
irrigation block j in one day can be calculated as:
the design flow rates of each irrigation block, and the irrigation priority in
M Wj, m this hydraulic group.
Wj = (mm)
m 10000Aj (3) Denoted by Pi by the pumping capacity of hydraulic group i , J the
total number of irrigation blocks in this hydraulic group, Td the off-peak
starting time of day d and Hd the number of off-peak hours. As men-
2.5.2. Water applied software implementation tioned previously, the optimum irrigation scheduling is to prioritise the
The water applied software implementation is shown in Fig. 5. The younger crop in the same hydraulic group. Therefore, at the start of
normalised flow rate (litre/second/hectare) is calculated using the re- each irrigation schedule calculation, all the irrigation blocks in the
corded flow rate at a timestamp divided by the summation of the areas same hydraulic group are rearranged according to their ages (or ratoon
of the irrigating sets at this timestamp (Fig. 5a, Step 11). This is number) (Fig. 7, Step 3). The algorithm then finds the best subset of
achieved by comparing the timestamp for each flow meter record with valves that need irrigation and can be irrigated at the same time
the start time and the finish time of each irrigation event (Fig. 5a, Step 6 (meaning that the overall design flow rate is lower than the pumping
and 7). With the normalised flow rate, the amount of water applied for capacity) (Fig. 7, Step 4). The irrigation event is then created for a
an irrigation event is the sum of water flow during each recorded in- duration of the minimum hour among the required hours in this subset
terval (300 s) multiplied by the area of this irrigation block, which is and the available number of hours for irrigation (Fig. 7, Step 5, 6 and
achieved by comparing the start time and the finish time of each 7). The available hours, the required hours, and the starting time are

5
E. Wang, et al. Computers and Electronics in Agriculture 172 (2020) 105376

Fig. 5. The algorithm to calculate the amount of water applied for each irrigation block.

updated after each irrigation shift (Fig. 7, Step 8, 9 and 10), and the demand had been reduced. The result clearly proves that the proposed
algorithm continues until there are no available hours left for irrigation closed-loop irrigation system can effectively manage irrigation sche-
(Fig. 7, Step 2). duling by incorporating the crop water balance with practical con-
straints on the test farm.
3. Results and discussion Fig. 9 shows the water balance of two Downriver irrigation blocks
using the unconstrained and constrained schedules. Due to the small
In this section, we will present the simulation results using the data number of irrigation blocks in this hydraulic group and their small
from the 2018 growing season of the test farm. Two irrigation schedules sizes, the irrigation demand for these two irrigation blocks can always
are compared: the unconstrained schedule from the decision support be met. Hence, the SWD and water applied charts for both schedules are
tool - IrrigWeb (which ignores practical and economic constraints), and nearly identical.
the constrained schedule created by the proposed closed-loop irrigation Fig. 10 shows the water balance of two selected Upriver irrigation
system (which accounts for energy tariffs, infrastructure constraints, blocks using the unconstrained and constrained schedules. Since there
and a user-defined irrigation priority that allocates to plant cane ahead are nine irrigation blocks in this group, the irrigation demand cannot
of older ratoons. These experiments demonstrate the applicability of always be fulfilled in the limited irrigation hours during weekdays. The
our methods to realistic scenarios on the test farm. constrained schedule prioritises fields according to irrigation avail-
Fig. 8 shows the water balance of two selected drip irrigation blocks ability, and most irrigation events for Set 5 are divided into two-day
using the unconstrained and constrained schedules. It can be clearly events, as can be seen in Fig. 10. However, the SWD is not much af-
observed that the water balance charts using the two schedules are fected due to the effective irrigation management of the proposed
identical for D4, which demonstrates that the proposed system is able to closed-loop irrigation system, and very close irrigation SWD charts can
meet the irrigation water demand of D4 when it is within the con- be seen for all the irrigation blocks. It is noted that Set 9 is relatively a
straints. On the other hand, the SWD for D1 using the constrained large block compared to the other sets in the Upriver hydraulic group,
schedule exhibits a notable difference with the one using the un- which runs on a 7-day irrigation schedule and requires 24-hour irri-
constrained schedule during December 2017 and February 2018, as gation. Therefore, the irrigation must be split into three events during
highlighted in Figs. 8-D1. The primary reason for that is because D1 is weekdays and can only be fulfilled in one event during weekends.
an R2 crop, while D4 is an R1 crop. Therefore, due to the low rainfall Despite this, the constrained schedule can still maintain a relatively
and high irrigation demand in the two periods, the proposed system close SWD with the unconstrained schedule, as can be seen in Fig. 10-
chose to irrigate D4 first and delay the irrigation for D1. However, the Set 9.
soil water deficit for D1 was brought back shortly after the water IrrigWeb calculates a potential crop water use and does not take into

Fig. 6. The Downlink program downloads the soil water deficit and daily crop water use from IrrigWeb every hour and calculates the water required for each
paddock. Then the Downlink program automatically arranges the irrigation scheduling for each irrigation block based on the energy constraint, irrigation system
constraint and farmers’ preference.

6
E. Wang, et al. Computers and Electronics in Agriculture 172 (2020) 105376

Fig. 7. The algorithm for calculating irrigation shifts.

Fig. 8. Water balance of two drip irrigation blocks using the unconstrained and constrained schedules. The top part of the chart presents the rainfall amount
throughout the season (in the green bar), the water applied using the unconstrained schedule (in the red bar), and the irrigation amount using the constrained
schedule (in the blue bar). The bottom part of the chart presents the soil water deficit result using the unconstrained schedule (in the solid red line) and the soil water
deficit result using the constrained schedule (in blue dash line). The soil water deficit threshold (in the black line) is the threshold set up by farmers to trigger an
irrigation event. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

7
E. Wang, et al. Computers and Electronics in Agriculture 172 (2020) 105376

Fig. 9. Water balance of two Downriver irrigation blocks using the unconstrained and constrained schedules.

consideration the constraint of pumping capacity, and so assumes that insight on the best timing of irrigation or the best amount of water to
each field is irrigated timely. Due to this fact, the unconstrained sche- apply for a specific crop and soil types. Decision support tools, on the
dule results in a significant increase in peak-hour usage. Table 2 shows other hand, take all these variables and provide an ideal suggestion on
the comparison of the number of off-peak and peak hours used and when to irrigate and how much to apply. However, these tools require
corresponding energy cost using the two schedules. Although the water frequent user-interaction to obtain precise irrigation scheduling. More
use using the two schedules are identical, the constrained schedule importantly, they are not tailored to handle the variabilities and con-
created by the closed-loop irrigation system reduces the peak-hour straints of real-world irrigation systems, which ultimately leads to
usage significantly, which generates economic benefits for farmers failed adoption practices. There is an urgent need to integrate these two
(with an overall AUD 2692.8 cost saving for the test farm). It can also be platforms to deliver precise, practical and effect irrigation scheduling
observed from the table that the simulated crop yield from both sche- for farmers.
dules are nearly identical, demonstrating an increase in profitability This project designed, tested and implemented a closed-loop irriga-
using the constrained schedule since costs are reduced while production tion system to apply the right amount of water at the right time in a
remains unchanged. practical and effortless way through the integration of automation with
an irrigation decision support system. In the first stage of this project,
the Uplink program was developed to automatically log the irrigation
4. Conclusion and rainfall data to IrrigWeb, from the WiSA irrigation system. In the
second stage of this project, a Downlink program was developed to
Better irrigation practices, through increased water use efficiency, connect IrrigWeb to WiSA, which can download, extract, calculate and
can deliver economic, environmental and social outcomes to agro- apply irrigation schedules automatically. The Downlink program suc-
ecological systems. The automated irrigation system has naturally been cessfully incorporated the IrrigWeb generated soil-water deficit for all
a solution to improve irrigation practices due to its remote-control fields and produced realistic and effective schedules by incorporating
capability. However, most automated systems do not provide any

8
E. Wang, et al. Computers and Electronics in Agriculture 172 (2020) 105376

Fig. 10. Water balance of two selected Upriver irrigation blocks using the unconstrained and constrained schedules.

Table 2
Energy, water and yield comparison for the 2018 growing season.
Group IrrigWeb Closed-loop Irrigation

Peak Off-peak Energy cost Water (ML) Yield (t/ Peak Off-peak Energy cost Water (ML) Yield (t/ Cost Saving (AUD)
hours hours (AUD) ha) hours hours (AUD) ha)

Drip 481 2451 6896.4 686.1 106.3 0 2932 5305.1 686.1 106.1 1591.3
Upriver 290 1148 3564.7 336.7 194.5 0 1438 2603.4 336.7 194.5 961.3
Downriver 42 827 1713.5 203.48 156.7 0 869 1573.3 203.48 156.8 140.2
Overall 813 4426 12174.6 1226.28 152.5 0 5239 9481.8 1226.28 152.5 2692.8

* The cost calculation is based on the 2017/18 energy tariff rate of Tariff 62 from Ergon Energy, Queensland, Australia. https://www.ergon.com.au/retail/business/
tariffs-and-prices/farming-tariffs.
* It is assumed that the pump capacity is 10 kW, under a constant rate of 65 L/s.

practical constraints and thus, overcame a deficiency in the IrrigWeb CRediT authorship contribution statement
decision support system to make it more life-like and realistic. The si-
mulation results demonstrate the proposed closed-loop irrigation E. Wang: Conceptualization, Methodology, Software, Writing -
system not only can effectively manage irrigation scheduling by in- original draft. S. Attard: Conceptualization, Data curation, Validation,
corporating crop water balance with practical constraints, but also Writing - review & editing. A. Linton: Resources. M. McGlinchey: Data
generates economic benefits for farmers. curation. W. Xiang: Writing - review & editing. B. Philippa:

9
E. Wang, et al. Computers and Electronics in Agriculture 172 (2020) 105376

Validation. Y. Everingham: Conceptualization, Supervision, Writing - furrow irrigation in the sugar industry. Aust. Soc. Sugar Cane Technol. 1–7.
review & editing, Funding acquisition, Project administration. Inman-Bamber, N.G., Smith, D.M., 2005. Water relations in sugarcane and response to
water deficits. F. Crop. Res. 92 (2–3), 185–202.
Inman-Bamber, N.G., McGlinchey, M.G., 2003. Crop coefficients and water-use estimates
Declaration of Competing Interest for sugarcane based on long-term bowen ratio energy balance measurements. F. Crop.
Res. 83, 125–138.
Inman-Bamber, N.G., Attard, S.J., Verrall, S.A., Webb, W.A., Baillie, C., 2007. A web-
The authors declare that they have no known competing financial based system for scheduling irrigation in sugarcane. In: XXVI Congress, International
interests or personal relationships that could have appeared to influ- Society of Sugar Cane Technologists, ICC, Durban, South Africa, 29 July - 2 August,
2007, pp. 459–464.
ence the work reported in this paper. Keating, B.A., et al., 2003. An overview of APSIM, a model designed for farming systems
simulation. Eur. J. Agron. 18 (3–4), 267–288.
Acknowledgement Knox, J.W., Rodríguez Díaz, J.A., Nixon, D.J., Mkhwanazi, M., 2010. A preliminary as-
sessment of climate change impacts on sugarcane in Swaziland. Agric. Syst. 103 (2),
63–72.
This research is funded by the Australian Government’s the National M.P. Holden, J.R., 2013. Irrigation of sugarcane manual.
Environmental Science Program through its Tropical Water Quality Romero, R., Muriel, J.L., García, I., Muñoz de la Peña, D., 2012. Research on automatic
Hub. The authors would also like to thank the Burdekin sugarcane irrigation control: State of the art and recent results. Agric. Water Manag. 114, 59–66.
Salam, Abdul, 2020. Internet of Things in Agricultural Innovation and Security. In:
farmers in our consultative group, Burdekin Productivity Services, and Internet of Things for Sustainable Community Development, pp. 71–112.
Sugar Research Australia. We would also like to thank SQR Software Schlager, E., 2001. Managing water for people and nature. Science (80-) 292 (5519),
Solutions, NQ Dry Tropics, and WiSA for their in-kind contribution. 1071–1072.
Shrivastava, A.K., Srivastava, A.K., Solomon, S., 2011. Sustaining sugarcane productivity
under depleting water resources. Curr. Sci. 101 (6), 748–754.
References Siebert, S., et al., 2010. Groundwater use for irrigation - A global inventory. Hydrol. Earth
Syst. Sci. 14 (10), 1863–1880.
Singels, A., et al., 2010. Modelling crop growth and crop water relations in South Africa:
Atzori, L., Iera, A., Morabito, G., 2010. The Internet of Things: A survey. Comput. Netw.
Past achievements and lessons for the future. South African J. Plant Soil 27 (1),
54 (15), 2787–2805.
49–65.
Battie Laclau, P., Laclau, J.P., 2009. “Growth of the whole root system for a plant crop of
Singels, A., Bezuidenhout, C.N., 2002. A new method of simulating dry matter parti-
sugarcane under rainfed and irrigated environments in Brazil. F Crop. Res. 114 (3),
tioning in the Canegro sugarcane model. F. Crop. Res. 78 (2–3), 151–164.
351–360.
Singels, A., Smith, M.T., 2006. Provision of irrigation scheduling advice to small scale
Carr, M.K.V., Knox, J.W., 2011. The water relations and irrigation requirements of sugar
sugarcane farmers using a web based crop model and cellular technology: a South
cane (Saccharum officinarum): a review. Exp. Agric. 47 (1), 1–25.
African case study. Irrig. Drain. 55, 363–372.
Chauhan, Y.S., Wright, G.C., Holzworth, D., Rachaputi, R.C.N., Payero, J.O., 2013.
Singels, A., Kennedy, A., Bezuidenhout, C., 1998. Irricane: A simple computerised irri-
AQUAMAN: A web-based decision support system for irrigation scheduling in pea-
gation scheduling method for sugarcane. Proc. S Afr. Sug. Technol. Ass. 1–6.
nuts. Irrig. Sci. 31 (3), 271–283.
Tiwari, D., Dinar, A., 2002. Role and use of economic incentives in irrigated agriculture.
Dong, X., Vuran, M.C., Irmak, S., 2013. Autonomous precision agriculture through in-
World Bank Tech. Pap. 1 (1), 1–75.
tegration of wireless underground sensor networks with center pivot irrigation sys-
Turral, H., Svendsen, M., Faures, J.M., 2010. Investing in irrigation: Reviewing the past
tems. Ad Hoc Netw. 11 (7), 1975–1987.
and looking to the future. Agric. Water Manag. 97 (4), 551–560.
Everingham, Y., Baillie, C., Inman-Bamber, G., Baillie, J., 2008. Forecasting water allo-
Vellidis, G., et al., 2016. Development and assessment of a smartphone application for
cations for Bundaberg sugarcane farmers. Clim. Res. 36 (3), 231–239.
irrigation scheduling in cotton. Comput. Electron. Agric. 127, 249–259.
Food and Agriculture Organization, Food and Agriculture Organization Corporate
World Water Assessment Programme, The United Nations World Water Development
Statistical Database, 2017. [Online]. Available: http://www.fao.org/faostat/.
Report 3: Water in a Changing World, 2012.
Gillies, M., Attard, S., Jaramillo, A., Davis, M., Foley, J., 2017. Smart automation of

10

You might also like