Assessing suitability of major meteorological factors for facility agriculture in mainland China

In China, more than two-thirds of protected cultivation occurs in low-tech facilities with limited ability to withstand adverse weather conditions. However, the specific meteorological factors that hinder facility agriculture production in various locations remain unclear. Here, we evaluated temperature and sunlight for assessing facility agriculture suitability in mainland China across different transplanting dates and ENSO phases (El Niño, La Niña, and Neutral) and to determine the optimal transplanting window. This aids in reducing climatic risks, and enhancing adaptation to changing climates. The results showed that growth cycles starting from March to June provide suitable temperature and sunlight, making them ideal transplanting window for many parts of northern China. However, both El Niño and La Niña significantly increased the high-temperature days and shortened the optimal transplanting window. For growth cycles starting from July to the following January, low temperatures are the primary factor limiting facility agriculture production in northern and western China. In southern China, sparse sunlight is the primary limiting factor year-round, and El Niño exacerbates this, particularly for growth cycles starting from September to November. This combined assessment of major meteorological factors, transplanting dates, ENSO phases, and regions, can assist decision-makers and growers in adapting to the changing climate and minimizing production risks.


Introduction
Facility agriculture is a cornerstone of Chinese agriculture, covering approximately 2.0 million ha, and around 85% of vegetable consumption is sourced from protected cultivation, generating over 70 million jobs in rural areas (Cao et al 2022).Facility agriculture, produced in enclosed or semi-enclosed environments, is less dependent on local climate conditions.However, in China, over two-thirds of protected cultivations are in low-tech facilities with limited ability to regulate and withstand adverse meteorological conditions (Li et al 2017, Zhang et al 2022a).The local climate impacts greenhouse construction costs (Wang et al 2021) and microclimate control costs during production.Notably, the primary expense in greenhouse production stems from the considerable energy consumption for heating, particularly in regions with colder climates at middle and high latitudes (Gruda et al 2019).Implementing facility agriculture that fully considers the local climate can mitigate the effects of abnormal weather conditions and reduce production costs.Hence, it is crucial to conduct climate suitability assessment for facility agriculture.Furthermore, China's booming logistics industry offers the cost-effective options for the longdistance distribution of greenhouse-grown products.
Agroclimatic zoning, or climate suitability assessment, combines long-term climate data with crop requirements to determine the optimal time and regions for crop production.This helps in the rational planning of agricultural to reduce climate-related risks (Wang et al 2023).Previous agroclimatic suitability studies predominantly focused on open-air agriculture, such as the soybean-maize double crop system (Nóia Júnior and Sentelhas 2019) and annatto production (Aparecido et al 2018) in Brazil.Trnka et al (2021) assessed agroclimatic changes in the Czech Republic, and temperature and precipitation are the most crucial climate elements to be considered.In middle-high latitudes, open-air production, such as grain crops, has just one annual growing season due to its long growth cycle and reliance on heat and precipitation.While facility agriculture can generate high production throughout the year with flexible planting dates, agroclimatic suitability varies with these dates.For facility agriculture, environmental factors such as temperature and sunlight take precedence over precipitation (Kim et al 2020, Zhang et al 2022b).Despite these differences, research on the climatic suitability of facility agriculture remains limited.
Various oceanic and atmospheric phenomena, including El Niño Southern Oscillation (ENSO), have been studied as drivers of global climate variability (Birk et al 2010, Frazier et al 2018).These climate variations, in turn, affect crop yield (Royce et al 2011, Guo et al 2021, Perondi et al 2022), posing challenges for agricultural production in China (Shuai et al 2013, 2016, Liu et al 2015).For example, during El Niño phase in Northeast China, rice yield dropped by 2.5% compared to Neutral years (Guo et al 2021).In 2012, a La Niña year associated with record heat and drought in the US, soybean yields declined by approximately 10% (Perondi et al 2022).While many studies have examined the impact of ENSO on climate extremes and open-air agricultural production, the microclimate and control costs of greenhouses are highly influenced by external meteorological conditions, which are in turn affected by ENSO phases.Therefore, it is crucial to investigate the effects of ENSO on facility agriculture with varying transplanting dates, providing valuable insights for agricultural policy formulation.
Climate conditions and their response to ENSO events vary significantly across different regions of China, leading to uncertainty in agricultural production.Existing agroclimatic suitability studies have predominantly focused on open-air cereal crops (Ceglar et al 2019, Nóia Júnior and Sentelhas 2019, Guo et al 2021, Trnka et al 2021).Therefore, there is an urgent need to assess climate suitability for facility agriculture in China.The main objectives of this study are (1) to identify the primary meteorological factors restricting facility agriculture production for each transplanting date in each region; (2) to assess the impact of ENSO phases on the suitability of facility agriculture production with various transplanting dates; and (3) to determine optimal transplanting windows for facility agriculture under different ENSO phases in each region.

Study area
Our study area was mainland China (figure 1), a region known for its diverse topography, varying distances from the sea, and susceptibility to various meteorological disasters.Mainland China exhibits a wide range of climates with cold, mid-temperate, warm temperate, northern subtropical, middle subtropical, southern subtropical, and edge tropical zones, denoted by Roman numbers I-VII in figure 1.Additionally, there are humid, semi-humid, semiarid, and arid regions in China from east to west, indicated by letters A-D in figure 1.More detailed information about each ecological geographic region in figure 1 can be found in table S1 (Wu et al 2003, Zhou et al 2021).Facility agriculture is prevalent throughout mainland China, with a higher concentration in the eastern low-altitude areas.

Data sources
We obtained data of ecological geographic regions in figure 1 from the Resource and Environment Science and Data Center of China (www.resdc.cn/data.aspx?DATAID=125).Daily data of maximum and minimum temperatures and sunshine hours from 1990 to 2019 were collected from 690 weather stations and meticulously quality-controlled by the China Meteorological Administration (http://data.cma.cn/).These weather data were used to assess the primary meteorological factors affecting facility agriculture in different regions.
Production cost data for facility agriculture in each province were sourced from the China Statistical Yearbook of Cost and Income Data of National Agricultural Products for the years 2012-2019.In the yearbook, production costs are categorized as direct and indirect costs.Indirect costs primarily consist of fixed assets depreciation, closely tied to greenhouse construction costs, which are influenced by regional thermal conditions (Chaudhary andMirza 2012, Razin et al 2021).Direct costs include expenses related to fertilizer usage, irrigation, seed costs, among others.

Criteria for restricting meteorological elements during growth cycles
We assessed suitability for 24 transplanting dates (the 1st and 16th day of each month) due to the greenhouse-grown crops can produce throughout the year with flexible planting dates.Some crops, such as tomato, pepper, and strawberry, are infinite grown varieties, have long growth cycles (e.g.120-270 d) (Khoshnevisan et al 2013, Qiu et al 2013, 2015), while others, such as melons and green beans, have short growth cycles (around 90 d) (Orgaz et al 2005).The detail lengths of growth cycle of different crops were shown in table S2 in the supplementary material.In this study, we considered a 180 day growth cycle for crops like tomatoes, providing 720 possible growth cycles over 30 years  and 24 transplanting dates per year.The first growth cycle started from January 1, 1990 and the transplanting date of the last growth cycle was on December 16, 2019.We also included suitability results for 90 day and 270 day growth cycles in the supplementary material (figures S3-S10) to assist with planting different crops.
We considered temperature and sunlight as the primary meteorological factors for facility agriculture production.Criteria, as outlined in table 1, were applied to determine whether specific days were restricted by heat, low temperature, or sparse sunlight.Details on how these criteria were established can be found in table S3.To determine the suitability of each growth cycle, we used the average number of restricting days for all growth cycles at all stations from 1990 to 2019 as the criteria, as indicated in table 1.Following the approach outlined by Nóia Júnior and Sentelhas (2019), we considered a transplanting date as restricted by a meteorological element if over 60% of the corresponding growth cycles were deemed unsuitable for that element.

Classification of ENSO phases and corresponding growth cycles
We classified ENSO phases as El Niño (EN), La Niña (LN), or Neutral (NE) based on the Oceanic Niño Index (ONI) from NOAA (www.cpc.ncep.noaa.gov/), a widely used indicator.ONI is calculated as a three-month running spatial average of sea surface temperature anomalies in the Niño 3.4 region  S4.
To investigate the impact of ENSO phases on the climatic suitability of growth cycles of each transplanting date.We categorized the 720 potential growth cycles between 1990 and 2019 into three groups: El Niño (EN), La Niña (LN), and Neutral (NE).We classified a growth cycle as EN-type if it occurred entirely within an El Niño period, and similarly for LN.Due to the delayed impact of ENSO on climate (Zhou et al 2021, Xing et al 2022), NE growth cycles were defined as those occurring entirely within a Neutral period and starting at least 6 months after the last ENSO event.The numbers of growth cycles for NE-, EN-and LN-type were 124, 90, and 68, respectively.The distribution of transplanting dates of these growth cycles is detailed in figure S1.

Data analysis 2.5.1. Effects of ENSO phases on the suitability of each transplanting date
We counted the number of restricting days for each meteorological element across the 720 possible growth cycles from 1990 to 2019 at each weather station.These counts were then grouped according to the corresponding growth cycle type (EN, LN and NE) for each transplanting date and station.To analyze the impact of ENSO phases, we conducted group t-tests to determine whether the numbers of unsuitable days in EN/LN growth cycles were significantly different from those in NE growth cycles.This allowed us to assess the effect of ENSO phases on each transplanting date and station.

Determination of the optimal transplanting window
We determined the suitability of each transplanting date for every station under different ENSO phases and meteorological elements by the number of multiyear average restricting days in the corresponding growth cycles.We used criteria specified in table 1 for each meteorological element.The multi-year average number of restricting days for each ecological geographic region was calculated as the average across all stations in it.Transplanting dates with all three meteorological elements met the criteria were considered as suitable transplanting window.

Main restricted meteorological elements for each transplanting date
We assessed the primary restricting meteorological elements during growth cycles for each transplanting date, considering temperature and sunlight.Figure 2 illustrates that for transplanting dates from March to early June, many places of northern China have suitable temperature and sunlight conditions, other places with restrictions primarily due to high temperatures.In contrast, most of southern China faces limitations from both high temperatures and sparse sunlight.For transplanting dates from July to the following January, cold to warm temperate zones and plateau areas face restrictions primarily due to low temperature.In north-to-south transition regions, facility agriculture for transplanting dates from August to early January is constrained by both low temperatures and sparse sunlight.Southern China is primary constrained by sparse sunlight, while subtropic western higher altitude areas have suitable conditions for facility agriculture production in terms of both temperature and sunlight.
We utilized provincial multi-year average production cost data for facility agriculture to validate the suitability assessment for low temperatures.Figure 3 displays correlation coefficients between the number of years without low-temperature restrictions during 1990-2019 and provincial multi-year average production costs.There was a negative correlation for each transplanting date, indicating that provinces not constrained by low-temperatures have lower production costs.These negative correlations are primarily driven by fixed assets depreciation rather than direct cost, transplanting dates from July to the following February show significantly negative correlations (p < 0.05) for fixed assets depreciation.As illustrated in figure 2, almost the entire China did not face low-temperature restrictions for transplanting dates from March to June.This aligns with the result of no significant correlation for these transplanting dates in figure 3. The time span of production cost data is 2012-2019, we provided the correlation coefficients between the number of years without low-temperature restrictions during 2012-2019 and provincial multi-year average production costs, that have the same time span (figure S2).And the correlation results are consistent with full time span (1990-2019).

Effects of ENSO phases on the suitability of different transplanting dates
Each of the 24 transplanting dates (the 1st and 16th day of each month) correspond to one growth cycle each year, the number of each type of growth cycles during 1990-2019 is actually the number of years affected by each type of ENSO phase.Based on figure S1, we tallied the number of LN-, EN-, and NE-type growth cycles for each transplanting date spanning 30 years.Figure 4 indicates that the number of NEtype growth cycles for every transplanting date is consistently no less than three.However, the number of EN-type and LN-type growth cycles is no less than three only for transplanting dates from May 1st to November 1st and from July 1st to December 1st, respectively.Consequently, we can infer that ENSO phases predominantly impact these specific transplanting dates.Subsequently, we will focus on exploring the effects of ENSO phases on these selected dates.
Figure 5 highlights stations with significant differences in the number of restricting days during El Niño growth cycles compared to Neutral growth cycles.For transplanting dates from May to June, sparse sunshine days significantly decreased (green circles) in some northeastern and western highland subduction regions, while high-temperature days significantly increased (red triangles) in some temperate semi-arid and arid regions of northern China.For transplanting dates from mid-June to mid-July, low-temperature days significantly decreased (purple crosses) in the eastern subtropical humid region.Additionally, sparse sunshine days significantly increased (yellow circles) for El Niño growth cycles starting from September to November in most southeastern China.Figure 6 displays stations with significant differences in the number of restricting days between La Niña and Neutral growth cycles.For transplanting dates from July to August, La Niña significantly decreased growth cycle low-temperature days (purple crosses) in the northern subtropical humid and western plateau regions, while significantly increased high-temperature days in some mid to warm temperate zones (red triangles).For transplanting dates from October to December, during La Niña  phases, low-temperature days (brown crosses) in the growth cycles significantly increased in parts of southeast China, while sparse sunshine days significantly decreased (green circles) in northern China and increased (yellow circles) in southwest higher altitude areas.

The optimal transplanting windows for facility agriculture under different ENSO phases
For each meteorological element, a recommended transplanting date is determined by the multi-year average restricting days in the corresponding growth cycles being less than the criteria shown in table 1 (less than 55, 41, 61 d for sparse sunlight, high temperature, and low temperature, respectively).Suitable transplanting windows vary across regions and ENSO phases (figure 7).The ecological geographic regions in figure 7, from left to right, encompass highland, cool temperate to tropical zones.Highland (HI, HII) and cool temperate zones (I) are not constrained by sparse sunshine (figure 7(a)) and high temperature (figure 7(b)) throughout the year but are restricted by low temperature (figure 7(c)).Compared to the Neutral phase, La Niña phase extends the suitable transplanting window in highland zones for transplanting dates from April to mid-May, where both temperature and sunlight are favorable.
In the mid-and warm temperate zones (II, III), sparse sunlight is not a constraint throughout the year (figure 7(a)).There are numerous high-temperature days for transplanting dates from February to July (figure 7(b)), which is the only period not restricted by low temperature (figure 7(c)).However, compared to the Neutral phase, both El Niño and La Niña phases lead to a shortened suitable transplanting window due to increased high-temperature days.In the subtropical to tropical zones (IV-VII), low temperature is not a constraint throughout the year (figure 7(c)), but these regions face extensive limitations due to sparse sunshine (figure 7(a)) and high temperature from mid-January to August.In the Yunnan-Guizhou plateau (VA5), the suitable transplanting window for facility agriculture production extends from December to the next February when all the three elements are suitable.During La Niña phases, there is a slight extension of this window, primarily due to reduced high-temperature constraints.

Sensitivity analysis of criteria used to define suitable growth cycle
The sensitivity analysis of the assessment results to uncertainties in the criteria used to define suitable growth cycles (table 1) is depicted in figure 8.The criteria for a day being unsuitable for greenhouse production were varied by ±1, 2, 3 (h or • C).The average change in the number of unsuitable days in growth cycles for each transplanting date was assessed (figures 8(a)-(c)).The sensitivity of unsuitable days in growth cycles to low-temperature criteria exhibits the lowest variability, with average change ranging  The criteria for determining a suitable growth cycle involve the total number of unsuitable days within it.For sensitivity analysis, the variation in this criteria was set to ±3, 5, 7 d, and then the average change in the proportion of suitable growth cycles was assessed (figures 8(d)-(f)).The average change in the proportion of suitable growth cycles ranged from 2.9% to 7.2% for sparse sunlight, 1.4% to 3.6% for low temperature, and 1.9% to 5.0% for high temperature.Sensitivities vary considerably for different transplanting dates, depending on the distribution of unsuitable days throughout the year.2).El Nino exacerbates the restriction of sparse sunlight (figure 5).Southern China has consistently low solar radiation due to its low altitude, extensive cloud cover, and high water vapor content (Zeng et al 2018).Therefore, greenhouse facilities in northern and western China should prioritize better thermal insulation and heating equipment, while plastic greenhouses in southern China benefit from supplemental lighting.Transition regions between north and south face constraints from both low temperature and sparse sunlight, making them less suitable for facility agriculture due to potentially high microclimate control costs.

Discussion and conclusion
Selecting the appropriate crops and varieties based on meteorological conditions during the growth cycle is crucial for successful production.Different crops with varying environmental tolerances, such as cold tolerance, heat tolerance, shade preference, etc.Even with the same transplanting date, climate suitability may be different when the growth cycle length is different.For instance, in warm temperate semi-humid regions (IIIB1/B2) the climate is more suitable for a 90 day growth cycle compared to a 180 day cycle for transplanting dates in March and September (figure S3).Therefore, it is recommended to choose crops with shorter growth cycles, such as melons and green beans, in these regions.

Effects of ENSO phase on major meteorological factors for facility agriculture production
ENSO phases have varying effects on meteorological elements in different regions and times, impacting facility agriculture production.Consistent with Liu et al (2014), our results showed that in northern China both El Niño and La Niña significantly increased high-temperature days during growth cycles from May to August, while El Niño and La Niña decreased sparse sunlight days in May to August and October to November, respectively.Winter-spring precipitation in southern China tends to be higher during El Niño years (Chen et al 2014, Yan et al 2020), leading to more sparse sunshine days.This aligns with our findings that in most southeastern China, El Niño significantly increased sparse sunlight days during growth cycles from September to November.In conclusion, previous studies primarily examined the impact of ENSO on precipitation and temperature for field crops in specific areas of China.This study innovatively assesses the nation-wide facility agriculture under various transplanting dates, considering temperature and sunlight.Our results highlight the transplanting dates most affected by ENSO phase.The effects of ENSO on facility agriculture vary by location, transplanting date, and phase.These findings offer valuable insights for decision-makers and growers to adapt to a changing climate.

Limitations of the research
This study highlights the impact of ENSO on the suitability of facility agriculture during different ENSO phases.However, it is important to consider the time lag between ENSO episodes and their effects in some regions in future studies (Shuai et al 2013, Xing et al 2022).Strengthening the analysis of delayed ENSO effects can enable predictions of climate suitability for facility agriculture some months before transplanting.
We assessed the suitability of low temperatures by examining costs of fixed assets depreciation in facility agriculture.High temperatures and sparse sunlight mainly contribute to increased greenhouse microclimate control costs.However, due to a lack of relevant records, production costs are not used to validate the suitability assessment for high temperature and sparse sunlight.
In this study, we assessed the climate suitability of each transplanting date.However, selecting the appropriate transplanting date and crop species involves considering factors beyond climate, including market demand, economic gain, and labor force.Any policy formulations based on our results should also account for these additional factors to ensure overall production success.

Conclusion
In summary, our conclusions are as follows: (1) In many parts of northern and western China, temperature and sunlight are suitable for facility agriculture production during growth cycles starting from March to June, and high temperatures often being the primary restriction in some other places.During El Niño and La Niña phases, there is an increase in high-temperature days, resulting in a shorter optimal transplanting window.For growth cycles starting from July to the following January, low temperatures are the primary limiting meteorological factors.
(2) Southern China faces limited sunlight for facility agriculture production year-round, with El Niño exacerbating sparse sunshine days in growth cycles starting after September.(3) In highland zones (HIIA/B1), the suitable transplanting window is from March to June, and it elongates during La Niña phases.In the Yunnan-Guizhou plateau (VA5), transplanting from December to next February is recommend, with a slight elongation during La Niña phases.

Figure 1 .
Figure 1.Study area and meteorological stations.(I Cool Temperate Zone; II Mid-Temperate Zone; III Warm Temperate Zone; IV Northern Subtropical Zone; V Middle Subtropical Zone; VI Southern Subtropical Zone; VII Edge Tropical Zone; H Highland; A Humid Region; B Semi-Humid Region; C Semi-Arid Region; D Arid Region.).

Figure 2 .
Figure 2. Primary restricting meteorological elements for facility agriculture production at each transplanting date (the first and sixteenth day of each month).Green suitable dot means not restricted by low temperature, heat and sparse sunlight; red unsuitable dot means restricted by all the three meteorological elements.

Figure 3 .
Figure 3. Correlation coefficients between the number of years without low-temperature restrictions during 1990-2019 on each transplanting date and production cost (fixed assets depreciation and direct cost).

Figure 4 .
Figure 4.The number of growth cycles of each type during 1990-2019 for each transplanting date.Green, red and blue lines denote Neutral (NE), El Niño (EN) and La Niña (LN) growth cycle.

Figure 5 .
Figure 5. Effects of El Niño on the number of restricting days in growth cycles.The stations highlighted are those with significantly different restricting days during El Niño and Neutral growth cycles.

Figure 6 .
Figure 6.Effects of La Niña on the number of restricting days in growth cycles.The displayed stations are those with significantly different restricting days during La Niña and Neutral growth cycles.

Figure 7 .
Figure 7.The optimal transplanting windows for production of facility agriculture by considering sparse sunshine (a), high temperature (b) and low temperature (c) at each ecological geographic region.(All the three elements are suitable is called optimal transplanting window.Light blue, red, and dark blue lines denote suitable transplanting dates for Neutral, El Niño, and La Niña phases, respectively.).

Figure 8 .
Figure8.Sensitivity analysis for number of unsuitable days in growth cycles to criteria of determining whether a day is suitable, (a)-(c) denotes sparse sunshine, low temperature and high temperature, respectively.Sensitivity analysis for proportion of suitable growth cycles to criteria used to determine whether growth cycle is suitable, (d)-(f) denotes sparse sunshine, low temperature and high temperature, separately.
from 4.5 to 14.8 d (figure 8(b)).For sparse sunlight (figure 8(a)) and high temperature (figure 8(c)) criteria, the average change in the number of unsuitable days in growth cycles ranges from 7.0 to 25.2 and 6.5-19.8d, respectively.Sensitivity is notably higher for transplanting dates from March to May regarding high-temperature criteria.In general, there is not a significant change in unsuitable days for growth cycles of 180 d.
4.1.1.Possible suggestions for production of facility agriculture The most common planting rotation pattern in Chinese facility agriculture starts transplanting as early as September, with the picking stage extending until the following early-July (Qiu et al 2011, Huang et al 2020, Gong et al 2021).It is crucial to focus on assessing the suitability of transplanting dates from August to January of the next year.The growth cycles of transplanting during August to January of next year are primarily limited by low temperatures in northern and western China and sparse sunlight in southern China (figure

Table 1 .
Criteria used to classify whether days and growth cycles are suitable.