Defining sowing and harvest dates based on the Asian Summer Monsoon

Sowing and harvest dates are a significant source of uncertainty within crop models especially for regions where high-resolution data are unavailable or, as is the case in future climate runs, where no data are available at all. Global datasets are not always able to distinguish when wheat is grown in tropical and sub-tropical regions, they are also often coarse in resolution. South Asia is one such region where large spatial variation means higher resolution datasets are needed, together 5 with greater clarity for the timing of the main wheat growing season. Agriculture in South Asia is closely associated with the dominating climatological phenomena, the Asian Summer Monsoon (ASM). Rice and wheat are two highly important crops for the region, rice being mainly cultivated in the wet season during the summer monsoon months and wheat during the dry winter. We present a method for estimating the crop sowing and harvest dates, for rice and wheat, using the ASM onset 10 and retreat. The aim of this method is to provide a more accurate alternative to the global datasets of cropping calendars than are currently available and generate input for climate impact assessments. We first demonstrate that there is skill in the model prediction of monsoon onset and retreat for two downscaled General Circulation Models (GCMs) by comparing modelled precipitation with observations. We then calculate and apply sowing and harvest rules for rice and wheat for each 15 simulation to climatological estimates of the monsoon onset and retreat for a present day period. We show that this method reproduces the present day sowing and harvest dates for most parts of India. Application of the method to two future simulations demonstrates that the estimated sowing and harvest dates are successfully modified to ensure that the growing season remains consistent with the internal model climate. The study therefore provides a useful way of modelling potential 20 growing season adaptations to changes in future climate. 1 Earth Syst. Dynam. Discuss., https://doi.org/10.5194/esd-2017-88 Manuscript under review for journal Earth Syst. Dynam. Discussion started: 1 November 2017 c © Author(s) 2017. CC BY 4.0 License.


Introduction
Field studies dominate the modelling literature on crops and agriculture.Many crop models are developed and applied at the site scale using site specific observations to drive models and optimize outputs.The growing awareness of climate change and the likely impact this will have on food pro-ensure the soil is in a suitable condition for wheat sowing after the rice harvest.Erenstein and Laxmi (2008) describe the zero-tillage approach which allows for a reduced turn-around time between the harvest of rice and sowing of wheat.Potential avenues by which the uncertainty from sowing and harvest dates can be reduced in inputs to crop simulations include: -The use of higher resolution regional data sets of recorded sowing and harvest dates for crop 100 calendars rather than existing global data sets.
-The use of new methods for estimating crop calendars in the absence of higher resolution regional data sets.

Motivation
The correct representation of the crop duration within crop models are crucial for the interpretation 105 of the important outputs from the model.For example if the datasets used for sowing and harvest dates are inaccurate, the simulations could grow crops during the wrong season, thereby affecting the reliability of the simulated water use and crop yield. Figure 1 compares observed sowing and harvest dates from a high resolution regional dataset from the Government of India, Ministry of Agriculture from Bodh et al. (2015) with a coarser scale global dataset over India from Sacks et al. (2010).Fig. 1 110 shows that the main differences are for spring wheat (plot a and b) with Sacks et al. (2010) providing sowing windows for spring wheat between 120 and 200 days earlier than the Bodh et al. (2015) data.This large difference is caused by the misclassification of spring wheat grown in winter as winter wheat in the Sacks et al. (2010) data.This is discussed by Sacks et al. (2010) as a potential limitation when using the data for tropical and subtropical regions.Spring wheat is the more common type of 115 wheat grown in the South Asia region (Hodson and White, 2007) because minimum temperatures there are not low enough to allow vernalization to take place, which is needed for winter varieties of wheat (Sacks et al., 2010;Yan et al., 2015).

4
Earth Syst.Dynam.Discuss., https://doi.org/10.5194/esd-2017-88Manuscript under review for journal Earth Syst.Dynam.Discussion started: 1 November 2017 c Author(s) 2017.CC BY 4.0 License. Figure 2 shows the averaged rice (green rectangles) and wheat (orange rectangles) growing season durations for Sacks et al. (2010) (diagonal hatching) and the Bodh et al. (2015) dataset (per-120 pendicular hatching-labeled MinAg) over-laid on the present day South Asia averaged precipitation climatology and estimates of the monsoon onset and retreat.This illustrates the differences between the Bodh et al. (2015) and Sacks et al. (2010) datasets showing that in Sacks et al. (2010), the main growing period for both rice and wheat appears to be during the monsoon.While rice is usually grown during the monsoon it is not typical that wheat should be grown during this period for this -To test the method in current and future climates.
We therefore present the methodology in Sect. 2. We show the proposed method is viable and show it works in Sect.3. Discussion of the results and conclusions are provided in Sect. 4 and Sect. 5 140 respectively.

Methodology
The methodology is summarized in the flow chart in Fig. 3.The model datasets, described in detail in Sect.A of the Appendix, include General Circulation Models (GCMs) and a Regional Climate Model (RCM).GCMs provide spatially consistent boundary data to an RCM, which generates 25km 145 regional fields (see Fig. 3 blue boxes).RCMs are based on the same physical equations as GCMs and therefore represent the entire climate system including the carbon and water cycle.Their higher resolution allows a better representation of the regional-scale processes adding detail to fields like precipitation (Mathison et al., 2015).The individual RCM simulations (also called HNRCMS -see Appendix Sect.A) used in this analysis are referred to using their global driving data abbreviations; 150 HadCM3, ECHAM5 and ERAint as described in Appendix Sect. A. Precipitation fields are used to generate a precipitation climatology which are used to calculate monsoon statistics (See Sect.2.2) from which sowing and harvest dates are estimated; shown by the pink rectangles (see Sect. 2.3).
These estimated sowing and harvest dates are referred to as relative monsoon sowing and harvest dates (see Fig. 3).Observations are used throughout the process to ensure the method is viable and produces sensible results, these are described in Sect.2.1 and shown by the green boxes.

Observations
In order to demonstrate the viability of the methodology outlined in Fig. 3 we compare the simulated precipitation with observations from the Asian Precipitation-Highly Resolved Observational Data Integration Towards the Evaluation of Water Resources (APHRODITE - Yatagai et al., 2012) dataset.

160
APHRODITE is a daily, 0.25 • resolution land only gridded dataset that is also used in Mathison et al. (2015) to show that the RCMs in this analysis capture the general hydrology of the region.The midpoints of these observed ranges are calculated and compared against the midpoints of the model pentads for onset and retreat in days of the year.As a post-processing step the differences are then masked using the (ICRISAT, 2015) crop areas, so that only the areas where rice or wheat are grown are considered.

Estimating monsoon onset and retreat
There are a wide variety of metrics for estimating the monsoon onset and retreat, some use a combination of meteorological variables such as 850hPa wind and precipitation (Martin et al., 2000), others such as Sperber et al. (2013)   where P is the unsmoothed pentad precipitation climatology and P min and P max are the annual minimum and maximum at each gridbox respectively.The monsoon onset is then defined as the pentad in which the NPPI exceeds 0.618 for the first time and withdrawal as the last time the NPPI drops below this threshold in the year.The NPPI only reaches a value of 1.0 once in the annual cycle 185 which corresponds to the monsoon peak.In this analysis we use the NPPI metric to calculate the pentad of the monsoon onset, retreat, peak and duration for the APHRODITE observations and the three HNRCM simulations.

Calculating sowing and harvest dates from monsoon characteristics
We use estimates of the monsoon onset and retreat together with present day rules on sowing and har-190 vest for rice and wheat to calculate the sowing and harvest dates relative to the monsoon (See Fig. 3).
This method allows any crop model that uses, for example a driving dataset similar to APHRODITE or the HNRCMs, to derive sowing and harvest dates that are consistent with the monsoon of the driving data (see Fig. 3).Thus growing the crop at the appropriate time of the year i.e rice is kept during the monsoon period and wheat is sown and harvested during the dry season.the usefulness of the method where observations were not available, which is more the point of this method.
The RelM onsoon croprule is then applied to the monsoon onset and retreat field to provide an estimate of sowing and harvest dates for rice and wheat based on the monsoon.We refer to these estimates of sowing and harvest dates as 'monsoon derived crop dates' for brevity.(2015)) and therefore the analysis for the present day in Sect.3.3 focuses on these areas.On the basis that most of the South Asia region is dominated by the ASM, the RelM onsoon croprule , though tuned using India observations, can be applied to the whole South Asia region in order to estimate sowing and harvest dates for larger areas with a rice-wheat rotation (see Sect 3.4).3 Results

245
We compare the model monsoon to the monsoon calculated from precipitation observations to demonstrate that the model is able to reproduce the monsoon (See Sect.3.1) and therefore the methodology summarized in Fig. 3 and Sect. 2 is viable.In Sect.3.2 we compare the simulated monsoon with the observed sowing and harvest dates in order to calculate the monsoon derived sowing and harvest dates and compare these new simulated sowing and harvest dates with the observations.

250
We then show results from applying the method in Sect.3.3.As a demonstration, we also apply the method to two future periods in Sect.3.4.The climatology shown in Fig. 2 shows that on average the observed rice and wheat sowing and harvest dates from MinAg align well with the monsoon onset and retreat in the simulations.Observed rice sowing dates generally compare well with the monsoon onset in the model as shown in Fig. 5 and Fig. 6.   for India.Table 1 shows the that on average across India rice sowing occurs between 3 and 20-days prior to the averaged modelled monsoon onset (3rd block, Table 1).In general the differences between rice harvest and monsoon retreat are larger but still consistent across the region (see Fig.  The monsoon derived sowing and harvest dates are calculated from applying the RelM onsoon croprule for each model (See Table 1) to the simulated monsoon onset and retreat fields (see Fig. 3).Here we compare these with the gridded observations to see how well the method performs for the present day.

Comparison of model monsoon onset and retreat with precipitation observations
295

Analysis of future monsoon onset and retreat
As a demonstration of the method summarised in Fig. 3, the HELIX SWLs (described in Sec.In fact all of monsoon onset, peak, retreat and duration show a large degree of variability as shown in Fig. 9 where each statistic has been averaged for South Asia.Each point in Fig. 9   10 for each of the simulations are calculated using the method described in Fig. 3).This shows that 330 the proposed method provides an estimate of sowing and harvest dates that ensures the crops can continue to be grown, in the simulation, when the climate is most appropriate rather than being fixed to the present day observed values.(green) calculated using the method described in Fig. 3 for HadCM3 (perpendicular hatching) and ECHAM5 (diagonal hatching).The monsoon onsets for each simulation are shown using blue vertical lines and retreat pink vertical lines (ECHAM5-dash dot lines, HadCM3-solid

Discussion
Recent climate impact studies such as AgMIP (Rosenzweig et al., 2013(Rosenzweig et al., , 2014))) and ISIMIP (Warszawski et al., 2013(Warszawski et al., , 2014) ) have highlighted the importance of reliable input data for models.Section 1.1 highlights the scale of the uncertainties present when solely using a global sowing and harvest dataset to simulate region specific cropping patterns.We have therefore proposed a new method for generating sowing and harvest dates for South Asia based on the ASM.In general the method reproduces observed sowing and harvest dates for much of India, these results are discussed further in 340 Sect.4.1.This method will also be useful in other monsoon regions where data are scarce, unreliable or unavailable such as in future climate simulations.The future results are discussed further in Sect 4.2.

Present day analysis
In general the method described by Fig. 3 works well across most of India for the present day, sowing given by the observations (see Fig. 6).However there are regions where the estimated sowing and harvest dates do not compare as well against present day observations.Rice sowing is generally closely associated with ASM onset across most of central India, however in the south of India there is a small region where the differences between the observations of sowing dates and the monsoon are 350 larger than everywhere else (see Fig. 5).In Sect.3.2 this region is shown to have different monsoon characteristics to the rest of India.This part of India includes the state of Tamil Nadu, this state is located on the lee side of the Western Ghats and therefore does not receive the large amounts of ASM rainfall that is more commonly associated with this part of the world.Tamil Nadu receives up to 50 percent of its annual rainfall during October-December via the less stable North Eastern (NE)

355
Monsoon.The NE monsoon is therefore more important for water resources for this part of India than the ASM which accounts for approximately 30 percent of the annual rainfall for this region (Dhar et al., 1982).These differing monsoon characteristics mean different agricultural practices are required to cultivate rice in this part of the country.This is illustrated by Fig. 11 (left plot) which shows that the southern region of India with differing monsoon characteristics irrigates rice more 360 intensively than other parts of India.In the Tamil Nadu region, rivers are usually dry except during the monsoon months and the flat gradients mean there are few locations for building reservoirs, therefore approximately one third of the paddy rice crop is irrigated from a large network of water tanks (Anbumozhi et al., 2001).The Southern states of India have the highest density of irrigation tanks with large numbers also found in Andhra Pradesh and Karnataka, these are also regions shown 365 to have a high irrigation intensity in Fig. 11.Rice harvest is typically not as closely associated with the monsoon onset as rice sowing, which usually requires the monsoon to be fully established before planting.
The widespread irrigation of wheat shown in Fig. 11 (right plot) has less of an impact on the estimates of wheat sowing/harvest dates because this crop is less closely linked to the monsoon onset than rice.Therefore the regional differences between the MinAg observations and the monsoon derived sowing and harvest dates for wheat are not as large as some of those for rice (see Sect. 3.3).
Given that the method has provided reasonable estimates of sowing and harvest dates for most of India, it would be useful and interesting to extend this method to improve it for the South of India.The proposed method successfully adjusts the sowing and harvest dates where the monsoon begins earlier in the future simulations and therefore provides a good estimate of sowing and harvest dates for the two future periods considered.This is a key benefit of using this method as it simulates the 395 decision a farmer might take to sow before the usual observed date if the monsoon arrived early.This method therefore provides the capability for climate simulations to replicate the type of adaptation response that would happen in the real world.This method would also be useful for other regions that have a crop calendar that is similarly defined such as the SSA; this is a multiple cropping region with sowing and harvest dates closely associated with the main rainy season (Waha et al., 2013).

Figure 1 .
Figure 1.The difference in the end points of the sowing (plot a and c) and harvest (plot b and d) windows between the global dataset from Sacks et al. (2010) and the regional dataset from Bodh et al. (2015) for wheat (plot a and b) and rice (plot c and d) for 1990-2007 climatologies.

Figure 2 .-
Figure 2. The one and a half year precipitation climatology for the 1990-2007 period averaged for South Asia for each simulation (ERAint-cyan line, ECHAM5-blue line, HadCM3-red line) and APHRODITE observations (black line) using a 5-day smoothed rolling mean.Also shown are the growing seasons also averaged for 1990-2007 for South Asia for wheat (orange) and rice (green) from two datasets; Sacks et al. (2010) (diagonal hatching -labeled sacks) and Bodh et al. (2015) (perpendicular hatching-labeled minag) and the monsoon onset (blue vertical lines) and retreat (pink vertical lines) from each of the simulations (APHRODITE-dotted, ERAintdashed, HadCM3-solid,ECHAM5-dash dot).

Figure 3 .
Figure 3.A flow chart summarizing the methodology.The blue rectangles represent datasets that are used within the methodology, green rectangles represent observations and pink rectangles represent any calculations parts of the methodology. 170 and the Normalised Pentad Precipitation Index (NPPI) (Lucas-Picher et al., 2011) only use precipitation.The NPPI metric provides a climatological estimate of the 175 monsoon onset, retreat, peak and duration and is calculated using Eq.(1).The NPPI metric uses the climatology of precipitation to estimate the monsoon statistics for a climatological period because the data are too noisy to calculate the monsoon statistics per year.The NPPI metric has been suc-Earth Syst.Dynam.Discuss., https://doi.org/10.5194/esd-2017-88Manuscript under review for journal Earth Syst.Dynam.Discussion started: 1 November 2017 c Author(s) 2017.CC BY 4.0 License.cessfully applied previously by Lucas-Picher et al. (2011) to analyse the monsoon of models of a similar resolution to the simulations used here (See Fig 3).
Fig.3).Equation2shows how the monsoon statistics are used along with the the sowing and harvest dates of each of the crops to calculate a croprule for each crop and stage.Collectively the crop rules given in Eq. 2 are referred to as RelM onsoon croprule .We use an area average rather than a gridbox by gridbox difference to define the RelM onsoon croprule because this provides a simple rule that 205

M
onsoonDerivedCropDate = M onsoonStatistic − RelM onsoon croprule (3) where the M onsoonStatistic can be monsoon onset or retreat and the RelM onsoon croprule is 215 one of the four crop rules given in Eq. 2 The spatial variability of the monsoon derived sowing and harvest dates is accounted for by the monsoon onset and retreat in the climatology used to calculate the RelM onsoon croprule .The monsoon derived sowing and harvest dates for both the APHRODITE and HNRCM simulations are provided and compared against MinAg observed sowing and harvest dates in Sect.3.3.The calcu-220 lation of the RelM onsoon croprule is based on observations for India (from MinAg and ICRISAT using monsoon derived estimates of sowing and harvest dates for two future periodsThe method summarised in Fig.3is applied to two future periods using the ECHAM5 and HadCM3 RCM simulations (described in Sect.A of the Appendix).Global mean temperatures are used (within the High-End cLimate Impacts and eXtremes project -HELIX) to define the future climate in terms 230 of specific warming levels (SWLs), i.e considering a 2 • C, 4 • C and 6 • C world.The simulations used here are for the period 1965 to 2100 and therefore only the 2 • C threshold for global mean temperature is actually passed during these simulations.For HadCM3 this occurs in 2047 and for ECHAM5, 2055.Therefore the two future periods used in this analysis are 2040-2057 and 2080-2097.The 2040-2057 period is chosen because it includes the year that the global mean temperature 235 exceeds 2 • C in the two simulations and the 2080-2097 period is chosen because it is furthest into the 10 Earth Syst.Dynam.Discuss., https://doi.org/10.5194/esd-2017-88Manuscript under review for journal Earth Syst.Dynam.Discussion started: 1 November 2017 c Author(s) 2017.CC BY 4.0 License.future in these simulations and therefore likely to show the greatest warming.The length of the two future analyses periods has been chosen for consistency with the ERAint RCM simulation which is only available for the period 1990-2007.Although the threshold of 2 • C is exceeded globally it is important to note that the relationship between the projected global mean change in temperature and 240 the regional climate change in temperature for South Asia is complicated.Heat and moisture and how they vary across the globe are not evenly distributed with land warming faster than the ocean(Christensen et al., 2013), therefore the actual temperature change experienced in South Asia may be higher than the global mean change.

Figure 4
Figure4shows plots of the onset (left column) and the retreat (right column) of the South Asian Summer Monsoon as defined using the NPPI described in Sect.2.2.The NPPI index for the clima-255

Figure 4 .
Figure 4. Plots of the 1990-2007 monsoon statistics; monsoon onset (left column) and retreat (right column).The APHRODITE precipitation observations (a and b) are shown and the three model simulations; ERAint (c and d), HadCM3 (e and f) and echam5 (g and h) calculated using the NPPI metric.White areas are the regions where the model precipitation exceeds the threshold indicating the start of the monsoon at the initial pentad, this does not imply early monsoon but more likely a model bias in the precipitation at this location.

Figure 5 .
Figure 5. Plots of the difference between the midpoint of the monsoon onset in the model and the midpoint of the observed rice sowing period for 1990-2007.

Figure 6
Figure 6 shows the comparison between the rice sowing MinAg observations compared with the

Figure 6 .
Figure 6.The comparison of the model monsoon onset in terms of the days of the year (to within the pentad)and the range of days of the year for the observed sowing date for rice.This is shown in terms of hit (blue) and overlap (yellow) or if there was no overlap this is shown as a miss (red) for the whole of India and are discussed in more detail in Sect. 4. Fig 2 highlights that the the average sowing and harvest dates for rice and wheat are closely aligned with the monsoon precipitation from all three RCM simulations.2903.3 Monsoon derived estimates of sow/harvest dates for rice and wheat

Figure 7
Figure 7 shows the monsoon derived estimates of rice sowing dates (left column) and compared with MinAg observations (right column).Fig. D.1 shows the same plots for rice harvest, with plots for wheat shown in Fig.D.2 and Fig. D.3 for sowing and harvest respectively.In general the monsoon derived estimates of sowing and harvest dates compare well with observations across much of the region for both crops.300

Figure 7 .
Figure 7.The monsoon derived rice sowing dates (left) and the difference between the MinAg observations and the monsoon derived rice sowing dates (right) for the period 1990-2007.
2.4) are 310 used to select two future periods: 2040-2057 and 2080-2097.Considering only these future periods, spatially HadCM3 and ECHAM5 show quite different future climates.HadCM3 shows a similar onset to the present day for 2040-2057 (see Fig. 8 (a) and (c)) but later onset compared with the 16 Earth Syst.Dynam.Discuss., https://doi.org/10.5194/esd-2017-88Manuscript under review for journal Earth Syst.Dynam.Discussion started: 1 November 2017 c Author(s) 2017.CC BY 4.0 License.present day for 2080-2097 (see Fig. E.1 (a) and (c)).ECHAM5 shows an earlier onset compared with the present day for the 2040-2057 period (see Fig. 8 (b) and (d)) but much later for the 2080-2097 315 period (see Fig. E.1 (b) and (d)).This suggests high variability in monsoon onset in these simulations.

Figure 8 .
Figure 8.The difference between the monsoon statistics for the 2040-2057 future period and the present day 1990-2007 for HadCM3 (left) and ECHAM5 (right).

Figure 9 .
Figure 9. Monsoon statistics; onset (a), retreat (b), peak (c) and duration (d) averaged for South Asia for twelve 17-year timeslices between 1970-2097 to provide a timeseries of values for the region to assess the variability of the monsoon

Figure 10 .
Figure 10.The one and a half year precipitation climatology for the period 2040-2057 (a) and the 2080-2097 (b) averaged for the whole of South Asia for each simulation (HadCM3-red line, ECHAM5-blue line) using a 5day smoothed rolling mean.Also shown are the monsoon derived growing seasons for wheat (orange) and rice

345
with the monsoon derived estimates of sowing and harvest dates falling within the range of days for 19 Earth Syst.Dynam.Discuss., https://doi.org/10.5194/esd-2017-88Manuscript under review for journal Earth Syst.Dynam.Discussion started: 1 November 2017 c Author(s) 2017.CC BY 4.0 License.

Figure 11 .
Figure 11.The average irrigation fraction for rice (a) and wheat (b) calculated from the ICRISAT observations of irrigation area and area planted

Figure 12 .
Figure 12.The annual timeseries of total monsoon precipitation, smoothed -using 5yr averaging, averaged for the whole of South Asia for all simulations; APHRODITE-solid black line, ERAint-solid cyan line, ECHAM5blue dashed line and HadCM3-red dotted line.Assuming that crops continue to be grown in accordance with the monsoon, Sect.3.4 shows that the method described in Sect. 2 provides a good estimate of sowing and harvest dates for the two future periods shown.Spatial plots of the sowing and harvest dates for the two future periods (not shown) are similar to those in Sect.3.3 for the present day with the south of the Indian peninsula continuing to show different monsoon characteristics (see Sect. 4.1) to the rest of India in the future, resulting in later estimated sowing and harvest dates for this region.

Figure C. 2 .
Figure C.2.The difference between the midpoint of the monsoon retreat in the model and the midpoint of the observed wheat sowing period for 1990-2007.

Figure C. 3 .
Figure C.3.The difference between the midpoint of the monsoon onset in the model and the midpoint of the observed wheat harvest period for 1990-2007.

Table 1 .
Table of RelM onsooncroprule for each dataset, crop and stage.The RelM onsooncroprule is the value subtracted from the monsoon onset/retreat in order to calculate a new sowing/harvest date based on the monsoon onset/retreat.In each case the new estimate of the sowing and harvest dates is calculated by subtracting the RelM onsooncroprule from the M onstat where M onstat is Monsoon onset or Monsoon retreat from a Earth Syst.Dynam.Discuss., https://doi.org/10.5194/esd-2017-88Manuscript under review for journal Earth Syst.Dynam.Discussion started: 1 November 2017 c Author(s) 2017.CC BY 4.0 License.

Table 2 .
Analysis of the differences between the midpoints of the MinAg data and Monsoon onset/retreat for rice/wheat sowing and harvest dates: The table shows the minimum, maximum, mean and standard deviation (SD) averaged across South Asia where wheat or rice are planted.Earth Syst.Dynam.Discuss., https://doi.org/10.5194/esd-2017-88Manuscript under review for journal Earth Syst.Dynam.Discussion started: 1 November 2017 c Author(s) 2017.CC BY 4.0 License.