1 Reducing Evaporative Water Losses from Irrigation Ponds
2 Through the Re-Use of PET Bottles
1 2 3 3 K. Simon , A. H. Slocum , R. Shanbhag
4 ABSTRACT
5 Evaporation can be a source of water loss from artificial reservoirs used in agriculture.
6 Current methods of covering artificial reservoirs are too costly to use by poor, small-scale
7 farmers. This paper presents a method for using waste PET bottles to reduce the evaporative
8 losses from open tanks. This water-conservation method was tested using eight evaporation
9 pans with daily water-level measurements to record evaporation rate. Four pans were used
10 as controls, two were covered with empty waste PET bottles, and two were covered with
11 bottles partially filled with soil. The experiment showed an average reduction in evaporation
12 by 40% with the PET bottle treatment, with a 90% confidence of reducing evaporation by at
13 least 18%. The addition of soil did not affect the degree of evaporation reduction. Given the
14 local economics of the region surrounding Pune, India, it was found that this intervention
3 15 can save water at a cost of 0.09 USD/m .
16 BACKGROUND AND INTRODUCTION
17 Agriculture is the main source of water use, representing 70% of global water consump-
18 tion. In 2013, 25% of the world’s irrigated agricultural systems were withdrawing water
19 faster than the regional replenishment rate (Rengel 2013). This challenge of sustainably
20 managing water is acutely noticed in developing countries. In 2010, India consumed 761 x
9 3 21 10 m of water. 90% of that water was used in agriculture (FAO 2011).
1Engineering Systems Division, MIT, 77 Massachusetts Ave., Cambridge, MA 02139. E-mail: Kev- [email protected] 2Mechanical Engineering, MIT, 77 Massachusetts Ave., Cambridge, MA 02139 3Vigyan Ashram, Pabal, Dist. Pune, Maharashtra, India 412403
1 Simon, May. 20, 2014 9 3 22 Evaporation accounts for a little over 2% (16.95 x 10 m /year in 2011) of India’s effective
23 water consumption (Frenken 2011). This paper proposes the use of waste PET bottles as
24 floating covers to reduce water-scarcity as a less expensive and potentially scalable solution
25 to reduce evaporation from man-made irrigation storage systems by approximately 40%.
26 Discussions with farmers near Pabal, India identified concerns about the rate of water
27 evaporation from their water retaining ponds, often referred to as ‘irrigation tanks’ in India.
28 They noted that solutions to this problem already exist. Suspended and floating covers
29 have been used to reduce evaporation in industrial applications and with large reservoirs
2 30 (Yao 2010) These existing solutions cost between 8 and 30 USD/m of reservoir and are too
31 expensive. In addition to being too costly, suspended and floating covers are either unable
32 to handle high-winds or unable to capture rain. This paper proposes and tests a method
33 for using PET bottles as floating covers. These bottles are less expensive than existing
34 evaporation reduction methods. In preliminary tests, it was found that these bottle rotate,
35 exposing water film to air and increasing evaporation. To address this challenge, and to
36 prevent the bottle from being blown away in stronger winds, a small amount of soil was
37 added to each bottle in half of the treatments. Subsequently, it was discovered that square
38 bottles (with rounded corners) naturally resist rotation and may not need to be filled with
39 soil. In either case, bottles blown by the wind pile up on one side of a pond and fall back
40 in to a low energy state. When rain falls on the floating mass of bottles, it merely passes
41 between them.
42 There are popular concerns about chemicals leaching from PET bottles into water when
43 exposed to sunlight and heat. However, surveys of the literature on chemical leaching has
44 shown that the leaching of dangerous chemicals into irrigation ponds is well below the dan-
45 gerous limits (Gorbaty 2013).
46 MATERIALS AND METHODS
47 The experiment consisted of daily measurements taken from 8 evaporation pans, which
48 can be seen in Figure 1, from March 5th to May 24th 2014. This time-window was chosen
2 Simon, May. 20, 2014 49 because it is the hottest and driest in Maharasthra. Of those 8 pans, 4 were uncovered
50 controls, 2 were covered with empty PET bottles, and the last 2 were covered with 500 ml
51 PET bottles containing 10 g of soil. Each pan was made from a rolled piece of metal, welded
52 into a 400 mm tall, 1.5 meter diameter cylinder and lined with white tarpaulin. Those pans
53 were filled with approximately 270 mm of water and refilled when empty. A wire mesh was
54 placed over the evaporation pans to prevent the bottles from blowing away in high wind,
55 and to prevent animals drinking from the pans. The Vigyan Ashram, based near the village
56 of Pabal, outside of Pune, conducted the experiment by recording the depth of the water in
57 each pan every day at 5:00 pm IST. Depth measurements were taken with a ruler mounted
58 to a stand, for stiffness and repeatability, which was placed in the pan for each measurement.
59 Variations in the weather would not induce error in the mean-shift because the analysis
60 is based on the difference in evaporation rates between the control and treatments. Measure-
61 ment errors were symmetric about the true average evaporation rate because the evaporation
62 rate is calculated from daily measurements of the water height. The remaining sources of
63 error could be caused by albedo, surface area, or leakage. Using the same tarpaulin in each
64 pond ensures that their albedo is similar. The difference in pond surface area caused by
65 construction is small compared to the effect of PET bottle coverage. Leakage is the only
66 major source of error that is unaccounted for in the experimental design, which can be found
67 if the average water level drop increases suddenly. This could be improved in the future by
68 periodically and randomly changing which pond has which treatment.
69 Due to the non-normality of the data, which can be seen in Table 1 and Figure 3, the
6 70 Bootstrap method with 10 resamples was used to evaluate the confidence intervals on the
71 results (Efron 1979).
72 RESULTS
73 The cumulative evaporation rate can be seen in Figure 2. Two of the four control pans
74 (C1 and C2) have unmeasured evaporation rates between day 46 and day 52 because they
75 became empty and not refilled. Two other control pans (C3, and C4) were re-filled on day
3 Simon, May. 20, 2014 76 39, resulting in another unmeasured day. To maintain the parallelism of the results, those
77 days were excluded from the evaporation analysis.
78 The data has low skewness, because evaporation rates are calculated from a running
79 water-level, making the measurement error symmetric.
80 The evaporation rate data has high kurtosis. This quality is expected because of the
81 outlying evaporation rates caused by measurement error. These measurement errors are
82 identified by examining the net evaporation rate, and are likely the cause of the long tails.
83 This high kurtosis brings the normality of the data into question, requiring a Bootstrap
84 resampling to test confidence intervals. This kurtosis can be reduced in the future by using
85 a more consistent measurement method such as a sight glass.
86 One of the empty-bottle treatments began to leak after being refilled on day 51. Those
87 results were left from the final hypothesis testing. However, the data from the first 51 days
88 were used to show that the addition of dirt does not reduce the effectiveness of the solution.
89 The ANOVA-multicompare in Figure 4, performed in MATLAB, from the pre-leak data
90 showed that there is an insignificant difference between the two treatments, and a significant
91 difference (p = 0.1) between the bottle treatments and the control.
92 The collected data has a mean-shift from 6.0 mm/day with the control to 3.8 mm/day
93 with the sand-bottle treatment. This is a 2.2 mm/day reduction in evaporation, amounting
∆x 94 to a 37% mean shift (%reduced = ). It was found, within a 90% confidence interval, that xcontrol
95 the average evaporation rate can be reduced by at least 1.1 mm/day (18%). The histogram
96 of the resampling that was used can be seen in Figure 5. The control evaporation rate of 6.0
97 mm/day closely matches the data presented by the Indian government (Sinha 2006), so the
98 analysis in this paper will be based on the Indian government’s data for yearly evaporation
99 rates in Pabal: 2250 mm/year. It is assumed that the fractional reduction in evaporation
100 remains constant.
101 ANALYSIS
102 The value of this evaporation reduction method is dependent on the economics of PET
4 Simon, May. 20, 2014 103 waste and the price or value of water. This section considers the costs and benefits associated
104 with using waste PET bottles as floating covers by examining the price of irrigation water
105 and the price of PET bottles. In this paper, the value of the intervention is calculated as
106 the difference between the cost of water saved, and the cost of covering a pond with PET
107 bottles.
108 Price of Water
109 There are many ways to determine the value of water (market price, cost of production,
110 social value). The most simple metric is the price paid for irrigation water by farmers. This
111 metric does not capture the value of water in water strained regions, where increased water
112 could improve or save the yield of a crop, but it does provide a baseline.
113 There is a large variability in the pricing of water schemes (Cornish 2004; Saleth 1997).
114 The cost of pumped groundwater is used as the benchmark for the price of water. A farmer
3 115 in Pabal, Maharashtra who consumes 100 m of water per day on average consumes approx-
116 imately 300 USD/year in electricity. A pump capable of providing that amount of water
117 would likely cost about 150 USD and needs to be replaced every 4 years on average. This
3 118 farmer would consume about 10,000 m of water over the course of 100 days of irrigation each
119 year. Not accounting for the cost of digging a well, the estimated cost of pumped irrigation
3 120 water in Pabal which will be used in this paper is 0.035 USD/m of water. This value is
3 121 reasonable because it is close to 0.032 USD/kWh; the energy cost of lifting 1 m an average
122 of 30m with a 20% efficient pump and an electricity cost of 0.08 USD/kWh. It should be
th 123 noted that other farmers in the region receive subsidized irrigation and only pay 1/10 the
124 electricity rate that this farmer does.
125 Other regions in Maharasthra, Aurangabad in this case, have reported groundwater ir-
3 126 rigation costs as great as 0.50 USD/m (Foster 2008). Furthermore, irrigation wells run dry
127 for many farmers in Pabal after April. Some farmers overcome this challenge by importing
128 water from nearby reservoirs to irrigate high-value crops such as mango-trees. The cost of
3 129 importing water via truck for one farmer was 11.60 USD for one 10 m tank truck, and sets
5 Simon, May. 20, 2014 130 the upper bound for acceptable water costs.
131 Price of PET Bottles
132 The net value of using PET bottles for irrigation needs to include the value of those
133 PET bottles, either as the opportunity cost or the price to purchase. This determines the
134 cost-benefit of using these bottles for evaporation reduction.
135 The price of recycled PET bottles in India has been reported to vary from 0.03 USD/kg to
136 0.02 USD/bottle (Dasgupta 2008). It was found that wholesalers in Pune typically purchase
137 PET scrap for between 0.50 and 0.67 USD/kg (Gorbaty 2013). The value 0.58 USD/kg
138 will be used in this calculation. The PET bottle used in this calculation weighs 20 g. The
139 common PET bottle, seen in Figure 6, has a height of 267 mm and a width of 76 mm.
2 140 This gives a cross-section of approximately 0.02 m . Under these assumptions, the cost of
2 141 covering an irrigation tank is approximately 0.82 USD/m .
142 The recorded mean-reduction in evaporation demonstrated in this paper is approximately
143 40 %. The average annual evaporation rate presented above comes from a report published
144 by the Government of India’s Central Water Commission (Sinha 2006).
145 If it takes 1 minute to add soil to a bottle, and a typical unskilled laborer is paid 0.30
146 USD/hour, it will cost an additional 0.005 USD/bottle. The cost of transporting bottles can
147 vary significantly, and is left out of this calculation. Note that many regions that use water
148 tanks as storage will be in places with low population density and low income. This fact will
149 require that the bottles be shipped in. The regional cost of transporting whole PET bottles
150 will need to be considered. One estimate of shipping costs is provided below.
151 A summary of the analysis presented in this paper, with additional upper and lower
152 bounds for estimations, is in Table 2. The estimated cost to conserve water with PET bottle
3 153 floating covers, assuming that the bottles will last for five years, is 0.09 USD/m of water.
154 This is greater than the estimated cost of pumping water in Pabal; slightly less than the cost
155 of pumping water in Aurangabad; and less than cost of importing water. There are many
156 places in India where water cannot be found in ground-wells during the summer season. In
6 Simon, May. 20, 2014 157 those places the PET bottles could enable increased yield in addition to reduced costs. Pabal
158 is one of those locations.
159 PET bottle availability
160 The availability of PET bottles affects how broad of an impact that this innovation could
161 have. It is estimated that Pune produces about 1,168 tons of PET bottles/year (Gorbaty
162 2013). The most common irrigation tank size in Maharashtra is 30 m wide by 30 m long,
2 163 and 3 meters deep. With a surface area of 900 m , it would take approximately 45,000 1L
164 bottles, or 900 kg, to cover one typical tank. If all PET bottle waste produced by Pune were
165 used as floating covers, they could cover approximately 1,300 tanks each year. There are a
166 total of 208,000 irrigation tanks in India (Vaidyanathan 2001). If the bottles last 10 years,
167 16 cities of equivalent size to Pune would be required to cover all of the irrigation tanks in
168 India.
169 Shipping the bottles have been estimated to cost 0.66 USD to ship 2000 bottles 1 km
170 (Gorbaty 2013). At 0.00033 USD/bottle/km, the transportation costs for bottles with 0.02
2 2 171 m cross-sectional area are 0.0165 USD/m /km. At this rate, shipping the bottles 50 km to
3 172 an irrigation tank doubles the cost of the intervention, putting it at 0.18 USD/m . Adapting
173 the solution to keep the bottle sources as local as possible will be important to preserving
174 the economic value of this intervention.
175 CONCLUSION AND FUTURE WORK
176 This paper identifies a method for putting a common waste to use for reducing water
177 evaporation by at least 18% (p = 0.1) with a mean reduction of 40%. The effect of using
178 PET bottles as floating covers has been demonstrated in the village of Pabal, India, near
179 Pune. It has been shown that the cost of this intervention can be greater or less than the
180 value of the saved water depending on the context. Compared to other interventions, this
rd 181 can save water at 1/3 the cost or less (Sinha 2006).
182 Due to the cost of PET bottles, this solution will be effective in regions where water is
183 scarce and waste PET bottles are available. In poor areas with low-population density, the
7 Simon, May. 20, 2014 184 bottles will need to be shipped to the irrigation tanks, adding a cost to the solution that will
185 need to be taken into account.
186 This paper also discusses the nature of data collected remotely with high measurement
187 induced kurtosis. The measurement method described above produces statistically useful
188 data with a simple, and low-cost setup when combined with a bootstrap analysis. This
189 setup can be scaled to engage farmers in different regions of India to test and compare this
190 and other evaporation reduction methods.
191 While this paper focuses on the use of waste PET bottles as floating covers in India, this
192 solution will have even more value in parts of the world where water is more expensive and
193 evaporation rates are very high. The yearly evaporation estimate in Pune is 2250 mm/year.
194 In places like Australia, which also use open water storage tanks, annual evaporation can be
195 as great as 3,000 mm/year (Craig 2005), further increasing the value of this intervention. The
196 application of this solution to other water scarce regions in the world should be investigated
197 further.
rd 198 Filling the bottles with soil accounts for 1/3 of the cost of the bottles. In some regions,
199 the soil filling may be unnecessary if the winds are not strong enough to blow the bottles
200 away. Square bottles (which have more resistance to rotation) could be used without soil
201 if the dynamics of bottle rotation, which can bring a film of water to the surface which
202 then evaporates, are better understood. A further benefit of reducing evaporation is that
203 many farmers currently do not use the last bit of water in their pond because evaporation
204 has caused such a dramatic increase in salt or other chemicals. Additional future work will
205 include studying the effects of reducing the concentration of these undesired chemicals in the
206 water by reducing evaporation.
207 ACKNOWLEDGEMENTS
208 We would like to thank Vishal Jagtap, Dr. Yogesh Kulkarny, and the rest of the Energy &
209 Environment section of the Vigyan Ashram for conducting the evaporation pan experiment,
210 evaluating water quality, helping our understanding of small scale farming around Pune, and
8 Simon, May. 20, 2014 211 their excellent hospitality.
9 Simon, May. 20, 2014 212 APPENDIX I. REFERENCES
213 Craig, I., Green, A., Scobie, M., and Schmidt, E. (2005) “Controlling Evaporation Loss from
214 Water Storages.” National Centre for Engineering in Agriculture (NCEA) Publication No
215 1000580/1.
216 Cornish, G. and Bosworth, B. (2004) “Annex 1.” Water Charging for Irrigation,
217
218 Dasgupta, B. and Khurana, S. (2008) “Waste Management of PET Bottles.” Journal of
219 Environmental Research And Development vol. 2, no. 4, 862-867.
220 Efron, B. (1979) “Bootstrap Methods: Another Look at the Jacknife.” The Annals of Statis-
221 tics. 1979
222 Food and Agriculture Organization of the United Nations (FAO) (2011) “India,” Aqua-
223 stat,
224 (Oct. 20, 2014).
225 Frenken, K. (2011) “Water Resources” Irrigation in Southern and Eastern Asia in figures.
226 FAO, 43.
227 Gorbaty, E. (2013) “Development of an efficient off-grid pumping system and evaporation
228 reduction strategies to increase access to irrigation for smallholder farmers in India.” M.S.
229 thesis, MIT, Cambridge, MA.
230 Rengel, Z. (2013) “Current State and Future Potential of Global Food Production and
231 Consumption.” Improving Water and Nutrient-use Efficiency in Food Production Systems.
232 Oxford: Wiley-Blackwell, 10.
233 Saleth, R. (1997) “Water pricing Experinces: an international perspective.” World Bank.
234 Sinha, S. (2006) “Evaporation Control in Reservoirs.” Government of India, Central Water
235 Commission, Basin Planning and Management Organisation, New Delhi.
236 Foster, S. (2008) “Ground water Use in Aurangabad - A Survey and Analysis of Social Signif-
237 icance and Policy Implications for a Medium-sized Indian City” Sustainable Groundwater
238 Management: Lessons from Practice.
10 Simon, May. 20, 2014 239 Vaidyanathan, A. (2001) “Tanks of South India.” New Delhi: Centre for Science and Envi-
240 ronment.
241 Yao, X., Zhang, H., Lemckert, C., Brook, A., and Schouten, P. (2010) “Evaporation Re-
242 duction by Suspended and Floating Covers: Overview, Modeling, and Efficiency.” Urban
243 Water Security Research Alliance Technical Report No. 28. Griffith University, Queens-
244 land, Australia.
11 Simon, May. 20, 2014 245 List of Tables
246 1 The skew and kurtosis of the datasets. The very high kurtosis indicates that
247 the data is not normal. Thus the Bootstrap method is used to find confidence
248 intervals...... 13
249 2 A summary of the estimated price of water conserved by using PET bottles
250 as floating covers...... 14
12 Simon, May. 20, 2014 TABLE 1: The skew and kurtosis of the datasets. The very high kurtosis indicates that the data is not normal. Thus the Bootstrap method is used to find confidence intervals.
Variable Skew Kurtosis Control 0.59 11.57 Empty Bottles 3.63 31.76 Sand-filled Bottles 0.64 12.40
13 Simon, May. 20, 2014 TABLE 2: A summary of the estimated price of water conserved by using PET bottles as floating covers.
Value Optimistic Estimate Pessimistic Price of Water [USD/m3] 1.16 0.50 0.035 Price of PET Waste [USD/kg] 0.03 0.58 0.75 Bottle Weight [g] 20 20 28 Cross-section [m2] 0.028 0.02 0.012 Avg. Yearly Evaporation [mm/year] 2250 2250 2250 Evaporation Reduction [%] 58 40 18 Bottle lifetime [years] 20 10 1 Bottle Price [USD] 0.001 0.012 0.015 Filled Bottle Price [USD] 0.006 0.017 0.020 Filled Coverage Price [USD/m2] 0.28 0.82 0.98 Cost of Saved Water [USD/m3] 0.01 0.09 2.42
14 Simon, May. 20, 2014 251 List of Figures
252 1 The evaporation ponds from the experimental setup...... 16
253 2 The total water that has evaporated from each pan over the course of the
254 experiment. Due to the leak in one of the ‘Empty’ treatments, those data
255 from day 39 and day 46 to day 52 were excluded from the final hypothesis
256 test. The slight difference ...... 17
257 3 Above are the histograms of the evaporation rates for the treatments and the
258 control. The long tails indicate that the data collected is not normal. . . . . 18
259 4 The graphical results of a multi compare test with a p-value of 0.1. Both
260 treatments are significantly different from the control, but are not significantly
261 different from each other. The addition of the soil does not significantly affect
262 the evaporation reduction of the intervention even though there appears to be
263 a small difference in Figure 2...... 19
264 5 A histogram of resampled with replacement. It shows that 90% of resampled
265 days reduce evaporation by at least 1.1 mm/day ...... 20
266 6 An evaporation pond from the experiment with the empty bottle treatment. 21
15 Simon, May. 20, 2014 FIG. 1: The evaporation ponds from the experimental setup.
16 Simon, May. 20, 2014 600 Control Empty Bottle 500 Soil Filled Bottle
400
300
200
100 Total water evaporation [mm]
0 0 20 40 60 80 Days into experiment FIG. 2: The total water that has evaporated from each pan over the course of the experiment. Due to the leak in one of the ‘Empty’ treatments, those data from day 39 and day 46 to day 52 were excluded from the final hypothesis test. The slight difference
17 Simon, May. 20, 2014 Control evaporation rate histogram Treatment evaporation rate histogram 90 90
80 80
70 70
60 60
50 50
40 40
30 30
20 20 Number of occurrences Number of occurrences
10 10
0 0 −5 0 5 −5 0 5 Evaporation rate (mm/day) Evaporation rate (mm/day) FIG. 3: Above are the histograms of the evaporation rates for the treatments and the control. The long tails indicate that the data collected is not normal.
18 Simon, May. 20, 2014 Multiple Comparison (p=0.1)
Control
Empty
Sand−filled
2 3 4 5 6 7 8 Mean Daily Evaporation Rate [mm/day] FIG. 4: The graphical results of a multi compare test with a p-value of 0.1. Both treatments are significantly different from the control, but are not significantly different from each other. The addition of the soil does not significantly affect the evaporation reduction of the intervention even though there appears to be a small difference in Figure 2.
19 Simon, May. 20, 2014 x 104 Resampled Distribution 4.5
4
3.5
3
2.5
2
1.5 Number of Results 1
0.5
0 −0.2 0 0.2 0.4 0.6 Mean−shift for Reduced Evaporation FIG. 5: A histogram of resampled with replacement. It shows that 90% of resampled days reduce evaporation by at least 1.1 mm/day
20 Simon, May. 20, 2014 FIG. 6: An evaporation pond from the experiment with the empty bottle treatment.
21 Simon, May. 20, 2014