Health Damage Assessment Modeling on Agricultural Water Scarcity based on Regression Analysis of Statistical Data National Inst. of AIST, Japan Tokyo City Univ., Japan Kogakuin Univ., Japan Masaharu MOTOSHITA Norihiro ITSUBO Atsushi INABA
2 The way of water use The use of withdrawal water in the world (1995) Food demand Population growth Domestic 354 billion m 3 (10%) Agricultural 2503 billiion m 3 (70%) Industrial 715 billion m 3 (20%) Agricultural water scarcity is one of the most serious issues
3 Flow of the modeling: 1 st step The lack of agricultural water Agricultural water use Crop production Annual crop production Dietary energy consumption Undernourishment
4 Water demand for crops production Same volume water Different yields in weights Gerbens-Leenes et al.(2004): Demand of water for fundamental growth will be independent from crop types Dietary energy production Net volume of water Same dietary energy of different crops Almost constant
5 Crop productivity estimates Precipitation: Q p Irrigation: Q irr Net volumes of water for crop production Evaporation: Q e = c 2 * (Q p + Q irr ) Infiltration: Q inf = c 1 * (Q p + Q irr ) Q p + Q irr - Q e Q inf = (1 - c 1 - c 2 ) * (Q p + Q irr ) Assumed to be constant inside countries Gross volume of supplied water Dietary energy production Net volume of water Almost constant Crop productivity [kcal/m 3 ] = Total crop production [kcal] Gross volume of supplied water [m 3 ]
6 Flow of the modeling: 2 nd step The lack of agricultural water Crop production Annual crop production Dietary energy consumption Dietary energy consumption Undernourishment
7 Simplified model of crop consumption Annual crop production: P Export rate: Er(%) Consumed inside the country P * (1 - Er/100) Direct consumption: DC DC = P * (1 - Er/100) Exported and consumed in other countries P * Er/100 Indirect consumption (country i): IC i IC i = P * (Er/100) * M i M i = Imported crops in country i Total imported crops in the world
8 Flow of the modeling: 3 rd step The lack of agricultural water Crop production Dietary energy consumption Dietary energy consumption Undernourishment Health damages
9 Previous modeling (reference) Previous approach for damage prediction Damages predicted by Human Development Index Life expectancy index Education index GDP index Further improvement? Source: Pfister et al. (2009), ES&T
10 Related factors on undernourishment Undernourishment Prevention Sufficient nutrient condition Average inside each country Dietary energy consumption Gini coefficient of energy consumption Gaps inside each country Treatment Medical treatment level and opportunity Health Expenditure per capita Non-linear multiple regression analysis
11 Results of multiple-regression analysis Explanatory variables t-value P-value Average daily dietary energy consumption X 1 : [kcal/capita/day] Gini coefficient of dietary energy consumption X 2 : [%] Health expenditure per capita X 3 : [US$/capita (PPP)] 3.56 0.00 4.60 0.00 5.23 0.00 Undernourishment damages [*10-6 DALYs] R 2 = 0.69 (n=147) = 63.9*Exp( -0.00138*X 1 ) + 1.03*Ln(X 2 ) - 1.32*Exp(-0.0166X 3 ) 9.22 Partial differentiation coefficient -0.0885 * Exp(-0.00138*X 1 )
Predicted damages by this model Predicted damages [10-6 DALYs/person] Prediction errors of the model [10-6 DALY/person] 250 AFRO AMRO EMRO 200 EURO SEARO WPRO ±25% 150 ±50% Optimal estimation 100 50 0 [10-6 DALY/person] 0 50 100 150 200 250 Reported damages [10-6 DALYs/person] Reported damages by WHO Ranges of error almost within 50% 12
Prediction errors ratios in this study 13 Improvement in damage prediction 4 3 2 African region American region Eastern mediterranean region European region South east asian region Western pacific region Error ratio = (Dp-Dr) / Dr Dp: predicted damages Dr: reported damages Improved area 1 0-2 -1 0 1 2 3 4-1 -2 Prediction error ratios in the model of Pfister et al. (2009) Improved area Error sum of squares decreased from 167 (previous) to 52 (this model) Improved in around 53% countries (particularly overestimated countries)
14 Calculation of damage factors Damage factor [DALY/m 3 ] inside country = Water Stress Index * Crop productivity [kcal/m 3 ] * Direct or Indirect Consumption [kcal/capita/day] * Damage differentiation factor [DALY/(kcal/capita/day)] Water Stress Index: weighting factor of water scarcity (Pfister et al. (2009)) 1 WSI = 1+e -6.4 WTA* (1/0.01-1) WTA = Total water use Total water availability
15 Distribution of damage factors World ave. damage factors : 0.14 [10-7 DALY/m 3 ]
16 Conclusions Health damages of undernourishment could be modeled Prediction errors could be improved by the consideration of three factors World ave. damage factor : 0.14 [10-7 DALY/m 3 ] the damages caused by the emission of 0.1 [kg-co 2 ] For reflecting more realistic situation Differences of several conditions inside each country Individual trade relationship between countries Purchase power in cases of insufficient food provision
17 Thank you for kind attention! Contact: m-motoshita@aist.go.jp
18 Procedures of regression analysis Data collection (country scale) Non-Linear single regression analyses Non-Linear multiple regression analysis Exclusion of countries with outlier data (Smirnov-Grubbs test) and no available data For decision of optimal line shape ( Linear, Cumulative power, Exponential, Logarithm, Logit (only for Gini coefficient)) Extraction of statistically significant variables based on t-value, P-value and adjusted R-square by backward step Undernourishment damage prediction model
19 Crop productivity of water Sensitivity to insufficient food provision
20 Allocations in trade relations Export: X [kcal] Export: Y [kcal] Import: X A [kcal] Import: Y A [kcal] Import: X B [kcal] Import: Y B [kcal] Import ratio: M A =(X A +Y A )/(X+Y) Import ratio: M B =(X B +Y B )/(X+Y) Disregard what kind of crops are imported from which country
Damages caused by insufficient food provision [DALY / (kcal/day)] Sensitive factor on damage factors 12 10 African region American region Eastern mediterranean region European region South east asian region Western pacific region 8 6 4 2 0 0 1 10 100 1000 10000 100000 1000000 Crop productivity [kcal/m3] Crop productivity will highly affect on the scale of damage factors 21