Untangling irrigation effects on maize water and heat stress

Preparing to load PDF file. please wait...

0 of 0
100%
Untangling irrigation effects on maize water and heat stress

Transcript Of Untangling irrigation effects on maize water and heat stress

https://doi.org/10.5194/hess-2020-627 Preprint. Discussion started: 20 January 2021 c Author(s) 2021. CC BY 4.0 License.
1 Untangling irrigation effects on maize water and heat stress 2 alleviation using satellite data
3
4 Peng Zhu1*, Jennifer Burney1 5 1School of Global Policy and Strategy, University of California, San Diego, CA USA 6 Correspondence to: Peng Zhu ([email protected])
7
8 Abstract. Irrigation has important implications for sustaining global food production, 9 enabling crop water demand to be met even under dry conditions. Added water also 10 cools crop plants through transpiration; irrigation might thus play an important role in 11 a warmer climate by simultaneously moderating water and high temperature stresses. 12 Here we use satellite-derived evapotranspiration estimates, land surface temperature 13 (LST) measurements, and crop phenological stage information from Nebraska maize 14 to quantify how irrigation relieves both water and temperature stresses. Our study 15 shows that, unlike air temperature metrics, satellite-derived LST detects significant 16 irrigation-induced cooling effect, especially during the grain filling period (GFP) of 17 crop growth. This cooling is likely to extend the maize growing season, especially for 18 GFP, likely due to the stronger temperature sensitivity of phenological development 19 during this stage. The analysis also suggests that irrigation not only reduces water and 20 temperature stress but also weakens the response of yield to these stresses. 21 Specifically, temperature stress is significantly weakened for reproductive processes 22 in irrigated crops. The attribution analysis further suggests that water and high 23 temperature stress alleviation contributes to 65% and 35% of yield benefit, 24 respectively. Our study underlines the relative importance of high temperature stress 25 alleviation in yield improvement and the necessity of simulating crop surface 26 temperature to better quantify heat stress effects in crop yield models. Finally, 27 untangling irrigation effects on both heat and water stress mitigation has important 28 implications for designing agricultural adaptation strategies under climate change.
29
30 Keywords: Irrigation, Evaporative cooling, MODIS LST, High temperature 31 stress, Water stress, Maize
32
1

https://doi.org/10.5194/hess-2020-627 Preprint. Discussion started: 20 January 2021 c Author(s) 2021. CC BY 4.0 License.
33 1. Introduction
34 Irrigation -- a large component of freshwater consumption sourced from water 35 diversion from streams and groundwater (Wallace, 2000, Howell, 2001) -- allows 36 crops to grow in environments that do not receive sufficient rainfall, and buffers 37 agricultural production from climate variability and extremes. Irrigated agriculture 38 plays an outsized role in global crop production and food security: irrigated lands 39 account for 17% of total cropped area, yet they provide 40% of global cereals 40 (Rosegrant et al 2002, Siebert and Dรถll 2010). Meeting the rising food demands of a 41 growing global population will require either increasing crop productivity and/or 42 expansion of cropped areas; both strategies are daunting under projected climate 43 change. Cropland expansion may be in marginal areas that require irrigation even in 44 the present climate (Bruinsma 2009); increasing temperatures will drive higher 45 atmospheric vapor pressure deficits (VPD) and raise crop water demand and crop 46 water losses. This increasing water demand poses a water ceiling for crop growth and 47 might necessitate irrigation application over present rainfed areas to increase or even 48 maintain yields (DeLucia et al., 2019).
49
50 However, the provision of additional irrigation water modifies both the land surface 51 water and energy budgets. Additional water can result in an evaporative cooling 52 effect, which may be beneficial for crop growth indirectly through lowering the 53 frequency of extreme heat stress (Butler et al., 2018). Especially considering the 54 future warmer climate, high temperature stress will be more prevalent (Russo et al., 55 2014) and might result in more severe yield losses than water stress (Zhu et al., 2019) 56 due to reduced photosynthesis, pollen sterility, and accelerated crop senescence in 57 major cereals (Rezaei et al., 2015b; Rattalino Edreira et al., 2011; Ruiz-Vera et al., 58 2018), therefore, a better understanding of irrigation effect on high temperature stress 59 alleviation will be important for agricultural management practices. More broadly, 60 understanding how irrigation can or should contribute to a portfolio of agricultural 61 adaptation strategies thus requires improved understanding of its relative roles in 62 mitigating both water and heat stresses.
63
64 Climate models and meteorological data have been used to investigate how historical 65 expansion of irrigation at global and regional scales has influenced the climate
2

https://doi.org/10.5194/hess-2020-627 Preprint. Discussion started: 20 January 2021 c Author(s) 2021. CC BY 4.0 License.
66 system, including surface cooling and precipitation variation (Kang and Eltahir, 2019; 67 Thiery et al., 2017; Bonfils and Lobell, 2007; Sacks et al., 2009). However, many 68 crop models still use air temperature rather than canopy temperature to estimate heat 69 stress; this may overestimate heat stress effect in irrigated cropland (Siebert et al., 70 2017), since canopy temperature can deviate significantly from air temperature 71 depending on the crop moisture conditions (Siebert et al., 2014). Recently, a 72 comparison of crop model simulated canopy temperature suggests that most crop 73 models lack a sufficient ability to reproduce the field-measured canopy temperature, 74 even for models with a good performance in grain yield simulation (Webber et al., 75 2017).
76
77 Alternatively, satellite-derived land surface temperature (LST) has been used to 78 directly quantify regional scale surface warming or cooling effects resulting from 79 surface energy budget changes due to changes in land cover and land management 80 (Loarie et al., 2011; Tomlinson et al., 2012; Peng et al., 2014). Importantly, yield 81 prediction model comparisons suggest that replacing air temperature with MODIS 82 LST can improve yield predictions because LST accounts for both evaporative 83 cooling and water stress (Li et al., 2019). Satellite data also provide the observational 84 evidence to constrain model performance or directly retrieve crop growth status 85 information. For example, satellite derived soil moisture had been used to characterize 86 irrigation pattern and improve irrigation amount estimation (Felfelani et al., 2018; 87 Lawston et al., 2017; Jalilvand et al., 2019; Zaussinger et al., 2019). Therefore, 88 integrating satellite products have the potential to improve our understanding of how 89 irrigation and climate change impact crop yield and thus provide guides for farmers to 90 make the optimal decisions.
91
92 In this study, we focus on Nebraska, the third largest maize producer in the United 93 States. Multi-year mean climate data shows that conditions are drier in western areas 94 and warmer in southern areas (Figure 1a and b). Importantly, Nebraska features a 95 mixture of irrigated and rainfed maize that facilitates comparison (more than half 96 (56%) of the Nebraska maize cropland is irrigated with more irrigated maize in the 97 western area (Figure 1c), according to the United States Department of Agriculture 98 (USDA, 2018a)). County yield data from the USDA shows that interannual 99 fluctuations in rainfed maize yield are much larger than for irrigated maize (Figure
3

https://doi.org/10.5194/hess-2020-627 Preprint. Discussion started: 20 January 2021 c Author(s) 2021. CC BY 4.0 License.
100 1b). Although irrigated yields are higher, rainfed maize yields have grown faster than 101 irrigated (3.9% per year versus 1.0% per year) over the study period (2003-2016) 102 (Figure 1b), one of the possible reasons is that breeding technology progress has 103 improved the drought tolerance of maize hybrids (Messina et al., 2010).
104
105 As noted above, irrigation potentially benefits crop yields by moderating both water 106 and high temperature stress. Here we use satellite-derived LST and satellite-derived 107 water stress metrics to statistically tease apart the contributions of irrigation to water 108 and heat stress alleviation, separately. We: (1) evaluate the difference in temperature 109 and moisture conditions over irrigated and rainfed maize croplands; (2) explore how 110 irrigation mitigates water and high temperature stresses using panel statistical models; 111 (3) quantify the relative contributions of irrigation-induced water and high 112 temperature stress alleviation to yield improvements; and (4) explore whether current 113 crop models can reproduce the observed irrigation benefits on maize growth status.
114 2. Materials and Methods
115 We first describe the data used, followed by a brief description of statistical 116 methodology.
117 2.1 Satellite products to identify irrigated and non-irrigated maize areas
118 We used the United States Department of Agricultureโ€™s Cropland Data Layer (CDL) 119 to identify maize croplands for each year in the study period 2003-2016 (USDA, 120 2018b). The irrigation distribution map across Nebraska was obtained from a previous 121 study that used Landsat-derived plant greenness and moisture information to create a 122 continuous annual irrigation map across U.S. Northern High Plains (Deines et al., 123 2017). The irrigation map showed a very high accuracy (92 to 100%) when validated 124 with randomly generated test points and also highly correlated with county statistics 125 (R2 = 0.88โ€“0.96) (Deines et al., 2017). Both the CDL and irrigation map are at 30m 126 resolution. We first projected them to MODIS sinusoidal projection and then 127 aggregated them to 1km resolution to align with MODIS ET and LST products. Then, 128 pixels containing more than 60% maize and an irrigation fraction >60% were labeled 129 as irrigated maize while pixels with >60% maize and <10% irrigation fraction were 130 labeled as rainfed maize croplands. As always, threshold selection involves a tradeoff 131 between mixing samples and retaining as many samples as possible. Our choices of
4

https://doi.org/10.5194/hess-2020-627 Preprint. Discussion started: 20 January 2021 c Author(s) 2021. CC BY 4.0 License.

132 <10% as the threshold for rainfed maize and 60% to define irrigated maize 133 represented the best optimization in our sample, as we found that more stringent 134 threshold had a very small effect on LST differences between irrigated and rainfed 135 maize at county level but resulted in significant data omission (more details in 136 supplementary Figure 1-2).
137

138 2.2 Maize phenology information 139 Maize growth stage information derived in a previous study was used to assess the

140 influence of irrigation on maize growth during different growth stages (Zhu et al.,

141 2018). Stage information including emergence date, silking date, and maturity date,

142 was derived with MODIS WDRVI (Wide Dynamic Range Vegetation Index, 8-day

143 and 250m resolution) based on a hybrid method combining shape model fitting (SMF)

144 and threshold-based analysis. Then we defined vegetative period (VP) as period from

145 emergence date to silking date, grain filling period (GFP) as period from silking date

146 to maturity date and growing season (GS) as period from emergence date to maturity

147 date. Details can be found in our previous studies (Zhu et al., 2018). WDRVI was

148 used due to its higher sensitivity to changes at high biomass than other vegetation

149 indices (Gitelson et al., 2004) and was estimated with the following equation:

150 ๐‘๐ท๐‘‰๐ผ = (๐œŒ๐‘๐ผ๐‘… โˆ’ ๐œŒ๐‘Ÿ๐‘’๐‘‘)/(๐œŒ๐‘๐ผ๐‘… + ๐œŒ๐‘Ÿ๐‘’๐‘‘)

(1)

151 WDRVI=100 โˆ— [(๐›ผ๏ผ1)+(๐›ผ+1)ร—๐‘๐ท๐‘‰๐ผ]

(2)

[(๐›ผ+1)+(๐›ผ๏ผ1)ร—๐‘๐ท๐‘‰๐ผ]

152 where ๐œŒ๐‘Ÿ๐‘’๐‘‘ and ๐œŒ๐‘๐ผ๐‘… were the MODIS surface reflectance in the red and NIR bands, 153 respectively. To minimize the effects of aerosols, we used the 8-day composite

154 products in MOD09Q1 and MYD09Q1 and quality-filtered the reflectance data using

155 the band quality control flags. Only data passing the highest quality control were

156 retained (Zhu et al., 2018). The scaling factor, ฮฑ=0.1, was adopted based on a

157 previous study to degrade the fraction of the NIR reflectance at moderate-to-high

158 green vegetation and best linearly capture the maize green leaf area index (LAI)

159 (Guindin-Garcia et al., 2012).

160 2.3 Temperature exposure during maize growth 161 We used daily 1-km spatial resolution MODIS Aqua LST (MYD11A1) data to 162 characterize the crop surface temperature; since its overpassing time is at 1:30 and 163 13:30, it is closer to the times of daily minimum and maximum temperature than the
5

https://doi.org/10.5194/hess-2020-627 Preprint. Discussion started: 20 January 2021 c Author(s) 2021. CC BY 4.0 License.

164 MODIS Terra LST (Wan et al., 2008) and is therefore better for characterizing crop 165 surface temperature stress (Johnson 2016; Li et al., 2019). For quality control, pixels 166 with an LST error >3 degree were filtered out based on the corresponding MODIS 167 LST quality assurance layers. Missing values (less than 3%) were interpolated with 168 robust spline function (Teuling et al., 2010). Aqua LST data are available after July 169 2002; we thus restricted our study to the period 2003-2016. For comparison, we also 170 obtained minimum and maximum daily surface air temperature (Tmin and Tmax) at 171 1-km resolution from Daymet version 3 (Thornton et al., 2018). For both MODIS 172 LST and air temperature, we calculated integrated crop heat exposure -- the growing 173 degree days (GDD) and extreme degree days (EDD) -- with the following equations:

๏ƒฅ ๏ƒฌ0, when T ๏€ผ 8๏‚ฐC

๏ƒผ

174

GDD30 ๏€ฝ N DD , DD ๏€ฝ ๏ƒฏ๏ƒญT ๏€ญ 8, when 8๏‚ฐC ๏‚ฃ T ๏€ผ 30๏‚ฐC๏ƒฏ๏ƒฝ (3)

8

t

t

t ๏€ฝ1

๏ƒฏ๏ƒฎ22, when T ๏‚ณ 30๏‚ฐC

๏ƒฏ ๏ƒพ

๏ƒฅ 175

EDD๏‚ฅ

๏€ฝ

N

DD ,

DD

๏ƒฌ0, ๏€ฝ๏ƒญ

when T ๏€ผ 30๏‚ฐC

๏ƒผ ๏ƒฝ

(4)

30

t

t ๏€ฝ1

t ๏ƒฎT ๏€ญ 30, when T ๏‚ณ 30๏‚ฐC๏ƒพ

176 Here temperature (T) could be either air temperature or LST and had been 177 interpolated from daily to hourly values with sine function (Tack et al., 2017). ๐‘ก

178 represents the hourly time step, N is the total number of hours in a specified growing

179 period (either the entire growing season, or a specific phenological growth phase, as

180 defined below).
181

182 2.4 Maize Water Stress 183 Water stress during maize growth was characterized by the ratio of evapotranspiration 184 (ET) to potential evapotranspiration (PET), as used in previous study (Mu et al., 2013). 185 MODIS product (MYD16A2) provided both ET and PET from 2003 to 2016 and 186 showed good performance for natural vegetation (Mu et al., 2011), however, our 187 comparison using flux tower observed ET at an irrigated maize site at Nebraska 188 suggested that ET at the irrigated maize was significantly underestimated by MODIS 189 ET (Supplementary Figure 3). Therefore, we used another ET product (SSEBop ET) 190 to replace MODIS ET. SSEBop ET was also estimated with MODIS products (Senay 191 et al., 2013), like LST, vegetation index, and albedo as input variables, but used a 192 revised algorithm including predefined boundary conditions for hot and cold reference 193 pixels (Senay et al., 2013) and showed better performance than MODIS ET (Velpuri
6

https://doi.org/10.5194/hess-2020-627 Preprint. Discussion started: 20 January 2021 c Author(s) 2021. CC BY 4.0 License.
194 et al., 2013), which was confirmed when we compared it with flux tower observed ET 195 at an irrigated maize site (Supplementary Figure 4). The comparison of MODIS PET 196 and flux tower estimated PET shows MODIS PET has satisfactory performance 197 (Supplementary Figure 5). Since MODIS PET from MYD16A2 has a spatial 198 resolution of 500 m with 8-day temporal resolution, while SSEBop ET has 1km 199 spatial resolution with daily time step, we reconciled the two datasets to 1km spatial 200 resolution and 8-day temporal resolution. Then ET, PET and ET/PET were averaged 201 over time to get mean ET, PET and ET/PET during VP, GFP and GS with satellite 202 derived phenology to characterize water status during maize growth.
203 2.5 Crop model simulation results 204 We compared the results of our statistical analysis with four gridded crop models. 205 Simulation results from pAPSIM, pDSSAT, LPJ-GUESS, CLM-crop for both rainfed 206 and irrigated maize across Nebraska were obtained from Agricultural Model 207 Intercomparison and Improvement Project (AgMIP) (Rosenzweig et al., 2013) and 208 Inter-Sectoral Impact Model Intercomparison Project 1 (ISIMIP1) (Warszawski et al., 209 2014). The four models were driven by the same climate forcing dataset (AgMERRA) 210 and run at a spatial resolution of 0.5 arc-degree longitude and latitude. All simulations 211 were conducted for purely rainfed and near-perfectly irrigated conditions. These 212 models simulated maize yield, total biomass, ET and growing stage information 213 (planting date, flowering date and maturity date). Planting date occurs on the first day 214 following the prescribed sowing date in which soil temperature is at least 2โ€‰degrees 215 above the 8โ€‰ยฐC base temperature. Harvest occurs once the specified heat units are 216 reached. Heat units to maturity were calibrated from the prescribed crop calendar data 217 (Elliott et al., 2015). Crop model simulation was evaluated by calculating the Pearson 218 correlation between simulated yields in the baseline simulations and detrended 219 historical yields for each country from the Food and Agriculture Organization. 220 Management scenario โ€˜harmnonโ€™ was selected, meaning the simulation using 221 harmonized fertilizer inputs and assumptions on growing seasons. More details on the 222 simulation protocol can be found in Elliott et al. (2015) and Mueller et al. (2019). We 223 used this model comparison project outputs to shed light on how well crop models 224 had simulated the irrigation benefits we identified in different phases of crop growth.
7

https://doi.org/10.5194/hess-2020-627 Preprint. Discussion started: 20 January 2021 c Author(s) 2021. CC BY 4.0 License.

225 2.6 Method 226 We used standard panel statistical analysis techniques to identify the impacts of

227 irrigation on maize productivity via heat stress reduction and water stress reduction

228 pathways.

229

230 Comparison of LST, ET, PET, ET/PET, GDD and EDD between irrigated and rainfed

231 maize areas was performed within each county to minimize the effects of other

232 spatially-varying factors, like background temperature and management practices, on

233 surface temperature and evapotranspiration. These biophysical variables averaged

234 over each county were then integrated over vegetative period (VP, from emergence

235 date to silking date), grain filling period (GFP, from silking date to maturity date) and

236 whole growing season (GS, from emergence date to maturity date) so we could

237 evaluate whether and how irrigation had differentially influenced maize growth

238 during early VP and late GFP.

239

240 We further examined how irrigation had changed the sensitivity of maize yield and its

241 components to temperature variation. As done in our previous study (Zhu et al., 2019),

242 we decomposed the total yield variation into three components: biomass growth rate

243 (BGR), growing season length (GSL) and harvest index (HI) based on the following

244 equation:

245 ๐‘Œ๐‘–๐‘’๐‘™๐‘‘ = ๐ป๐ผ โˆ™ ๐ด๐บ๐ต = ๐ป๐ผ โˆ™ ๐ต๐บ๐‘… โˆ™ ๐บ๐‘†๐ฟ

(5)

246 Aboveground biomass (AGB) was retrieved through a regression model:

247 AGB= 16.4 โˆ™IWDRVI0.8

(6)

248 which was built in the previous study through regressing field measured maize AGB

249 against MODIS derived integrated WDRVI (IWDRVI) (Zhu et al., 2019). Then HI

250 could be estimated as Yield/AGB and BGR could be estimated as AGB/GSL. Such

251 decomposition allowed us to examine how different crop growth physiological

252 processes responded to external forcing: HI characterizes dry matter partitioning

253 between source organ and sink organ and is mainly related with processes

254 determining grain size and grain weight; BGR is related with physiological processes

255 of daily carbon assimilation rate through photosynthesis and GSL is related with crop

256 phenological development. The uncertainties related with AGB estimation was

257 quantified through resampling as we did in previous studies (Zhu et al., 2019).

258
8

https://doi.org/10.5194/hess-2020-627 Preprint. Discussion started: 20 January 2021 c Author(s) 2021. CC BY 4.0 License.

259 Temperature sensitivity of irrigated or rainfed yield (๐‘†๐‘‡๐‘Œ๐‘–๐‘’๐‘™๐‘‘) was estimated using a 260 panel data model (Eq. (7)) with growing season mean LST and ET/PET as the 261 explanatory variables: 262 ๐‘™๐‘œ๐‘”โก(๐‘Œ๐‘–๐‘’๐‘™๐‘‘๐‘–,๐‘ก) = ๐›พ1๐‘ก + ๐›พ2๐ฟ๐‘†๐‘‡๐‘–,๐‘ก + ๐›พ3 ๐‘ƒ๐ธ๐ธ๐‘‡๐‘‡๐‘–,๐‘ก + ๐ถ๐‘œ๐‘ข๐‘›๐‘ก๐‘ฆ๐‘– + ๐œ€๐‘–,๐‘ก (7) 263 ๐‘Œ๐‘–๐‘’๐‘™๐‘‘๐‘–,๐‘ก is maize yield (t/ha) in county i and year t. It was a function of overall yield 264 trends (๐›พ1๐‘กโก) that had fairly steadily increased over the study period (Figure 1b), local 265 crop temperature stress (๐ฟ๐‘†๐‘‡๐‘–,๐‘ก ), and local crop water stress (๐‘ƒ๐ธ๐ธ๐‘‡๐‘‡๐‘–,๐‘ก ). The ๐ถ๐‘œ๐‘ข๐‘›๐‘ก๐‘ฆ๐‘– 266 terms provided an independent intercept for each county (fixed effect), and thus 267 accounted for time-invariant county-level differences that contributed to variations in

๏‚ถ ln(Yield) 268 yield, like the soil quality. ๐œ€๐‘–,๐‘ก is an idiosyncratic error term. ๐›พ2 or ๏‚ถLST defines 269 the temperature sensitivity of yield. The temperature sensitivity of BGR (๐‘†๐‘‡๐ต๐บ๐‘…), HI 270 (๐‘†๐‘‡๐ป๐ผ) and GSL (๐‘†๐‘‡๐บ๐‘†๐ฟ) could be estimated with Eq (7) in a similar way through using 271 BGR, HI and GSL as the dependent variable. Here the dependent variable Yield 272 (BGR, GSL and HI) was logged, so the estimated temperature sensitivity represented 273 the percentage change of Yield (BGR, GSL and HI) with 1ยฐC temperature increase.
274
275 To quantify the relative contribution of water and high temperature stress alleviation 276 to yield benefit, the yield difference between irrigated and non-irrigated maize 277 (irrigation yield-rainfed yield, โˆ†๐‘Œ๐‘–๐‘’๐‘™๐‘‘) was regressed over the quadratic function of 278 growing season EDD and ET/PET differences between irrigated and rainfed maize: 279 โˆ†๐‘Œ๐‘–๐‘’๐‘™๐‘‘๐‘–,๐‘ก = ๐›พ1โˆ† ๐‘ƒ๐ธ๐ธ๐‘‡๐‘‡๐‘–,๐‘ก + ๐›พ2โˆ† ๐‘ƒ๐ธ๐ธ๐‘‡๐‘‡๐‘–2,๐‘ก + ๐›พ3โˆ†๐ธ๐ท๐ท๐‘–,๐‘ก+โก๐›พ4โˆ†๐ธ๐ท๐ท๐‘–2,๐‘ก + ๐ถ๐‘œ๐‘ข๐‘›๐‘ก๐‘ฆ๐‘– + ๐œ€๐‘–,๐‘ก (8) 280 The yield improvement explained by heat and water stress alleviation was estimated

๏ƒฅ ๏ƒฅ ๏ƒฅ ๏ƒฅ ๏ง ๏„ ET ๏€ซ ๏ง ๏„ ET 2 ๏€ซ ๏ง ๏„EDD ๏€ซ ๏ง ๏„EDD2

1

PET i,t 2

PET i,t 3

i,t

4

i,t

281 as

๏ƒฅ ๏„Yieldi,t

. The relative

282 contribution of water and high temperature stress alleviation was estimated as

๏ƒฅ ๏ƒฅ ET

ET 2

๏ง1 ๏„

+๏ง 2 ๏„

PET i,t

PET i,t

๏ƒฅ ๏ƒฅ ๏ƒฅ ๏ƒฅ ๏ง ๏„ ET ๏€ซ ๏ง ๏„ ET 2 ๏€ซ ๏ง ๏„EDD ๏€ซ ๏ง ๏„EDD2

283

1

PET i,t 2

PET i,t 3

i,t

4

i,t

and

9

https://doi.org/10.5194/hess-2020-627 Preprint. Discussion started: 20 January 2021 c Author(s) 2021. CC BY 4.0 License.

๏ง 3 ๏ƒฅ ๏„EDDi,t +๏ง 4 ๏ƒฅ ๏„EDDi2,t

๏ƒฅ ๏ƒฅ ๏ƒฅ ๏ƒฅ ๏ง ๏„ ET ๏€ซ ๏ง ๏„ ET 2 ๏€ซ ๏ง ๏„EDD ๏€ซ ๏ง ๏„EDD2

284

1

PET i,t 2

PET i,t 3

i,t

4

i,t
, respectively. Given

285 the potential collinearity between โˆ† ๐ธ๐‘‡ and โˆ†EDD, we also calculated the Variance
๐‘ƒ๐ธ๐‘‡

286 inflation factor (VIF) to diagnose the severity of collinearity. The daytime LST

287 difference ( ๏„LST ) was also tested to characterize heat stress alleviation with the 288 following equation: 289 โˆ†๐‘Œ๐‘–๐‘’๐‘™๐‘‘๐‘–,๐‘ก = ๐›พ1โˆ† ๐‘ƒ๐ธ๐ธ๐‘‡๐‘‡๐‘–,๐‘ก + ๐›พ2โˆ† ๐‘ƒ๐ธ๐ธ๐‘‡๐‘‡๐‘–2,๐‘ก + ๐›พ3โˆ†๐ฟ๐‘†๐‘‡๐‘–,๐‘ก + ๐›พ4โˆ†๐ฟ๐‘†๐‘‡๐‘–2,๐‘ก + ๐ถ๐‘œ๐‘ข๐‘›๐‘ก๐‘ฆ๐‘– + ๐œ€๐‘–,๐‘ก (9) 290 Then, the relative contribution of water and high temperature stress alleviation was

๏ƒฅ ๏ƒฅ ET

ET 2

๏ง1 ๏„

+๏ง 2 ๏„

PET i,t

PET i,t

๏ƒฅ ๏ƒฅ ๏ƒฅ ๏ƒฅ ๏ง ๏„ ET ๏€ซ ๏ง ๏„ ET 2 ๏€ซ ๏ง ๏„LST ๏€ซ ๏ง ๏„LST 2

291 estimated as

1

PET i,t 2

PET i,t 3

i,t

4

i,t and

๏ง

3

๏ƒฅ

๏„L

STi

,

t

+๏ง

4

๏ƒฅ

๏„L

STi

2 ,t

๏ƒฅ ๏ƒฅ ๏ƒฅ ๏ƒฅ ๏ง ๏„ ET ๏€ซ ๏ง ๏„ ET 2 ๏€ซ ๏ง ๏„LST ๏€ซ ๏ง ๏„LST 2

292

1

PET i,t 2

PET i,t 3

i,t

4

i,t
, respectively.

293 3. Results

294 As expected, irrigation improved maize yield and the yield benefit showed a distinct 295 spatial variation when we compared areas we identified as irrigated versus rainfed 296 maize. The yield benefit of irrigation was much higher in the western area of the state 297 (Figure 2a), because the drier environment in western area widened the yield gap 298 between irrigated and rainfed cropland in an average year. The satellite derived 299 vegetation index WDRVI reflected these differences, with higher values in areas we 300 identified as irrigated maize, especially around maize silking (Figure 2b). Importantly, 301 this suggested that, in conjunction with ground-based information calibrated crop 302 phenology, irrigated and rainfed cropland were distinguishable with time series 303 satellite data where rainfall does not meet crop water demand.
304
305 When county-level LST data were averaged over 2003-2016, the daytime LST in 306 irrigated maize was 1.5โ„ƒ cooler than rainfed maize, while nighttime LST showed a 307 very slight difference (0.2 โ„ƒ ) (Figure 3a,b). When the LST differences were
10
IrrigationIrrigatedWaterDiscussionLst