Study site

The study was conducted in the Central Bohemian region (valleys of the central parts of the Czech Republic; Fig. 1). This fruit-vegetable-producing region is characterized by the warmest and driest climatic conditions, where drought stress is often a limiting factor for crops; however, agricultural advantages of the fruiting region ensure the longest growing season and the longest frost-free period with the most productive soil conditions. The transplanting dates for fresh-market tomato cultivars grown under open field conditions correspond to the stable transition of the average daily air temperature above 15°C (resulting in over 110-day season). During 1961–2020, the mean temperature of the tomato-growing season ranged from 16.5 to 18.0°C, and the total precipitation varied from 289 to 325 mm. The average maximum and minimum temperatures were 22.1 and 11.3°C, respectively (Potopová et al., Reference Potopová, Štěpánek, Zahradníček, Farda, Türkott and Soukup2018, Reference Potopová, Zahradníček, Štěpánek, Türkott, Farda and Soukup2017a). Growers usually delay the planting date of thermophilic vegetables past 15 May to minimize the risk of frost damage (only 10% risk). In cooperation with vegetable farms, a field trial was carried out at two experimental sites (Hanka Mochov, 189 m a.s.l. and Praha-Suchdol, 287 m a.s.l) during 2014, 2015, 2016, 2017, 2018 and 2020 tomato-growing seasons, where input data for the model were collected.

Fig. 1. Map of field locations and their geographical positions in the Czech Republic.

Field experiments

Crop management adopted in this study (cultivation, weeding, irrigation, fertilization, standard protection against diseases and pests) was based on the practices of local farmers. The soil was ploughed to a depth of 25 cm in the autumn and cultivated by spring. Nitrogen fertilization was scheduled throughout the season to deliver 200 kg/ha each season. Weeds were controlled by hand; pests and diseases were completely controlled by organic treatments. Drip irrigation (drip tube in line emitters, 2 litres/h, spaced 0.5 m) was applied according to the soil moisture once or twice per week with a dose of 15 and 20 mm. Irrigation was scheduled according to the changing needs of the plant; the young developmental stage coincided with a time with high drought frequency (mid-May and June) and the middle phase of fruit growth was a time of high water consumption (July, ~140 m²/ha per tonne yields) (Potopová et al., Reference Potopová, Štěpánek, Farda, Türkott, Zahradníček and Soukup2016, Reference Potopová, Zahradníček, Štěpánek, Türkott, Farda and Soukup2017a). During the late reproductive stages, the tomato yields were less sensitive to droughts due to the ability of tomato plants to extract water from deeper layers during soil moisture stress (70 m²/ha per tonne yields).

All the data on climate, soil, crop growth, management and yields collected in the experiments were entered in the standard DSSAT files (*.TMX, *.TMA, *.TMT, *.WTH, *.SOL) needed for execution of the CROPGRO-Tomato model. Experimental data sets and managing crop as well as weather and soil data for model evaluation were used. Measured and simulated growth and development of the fresh-market Thomas F1 indeterminate tomato cultivar grown under open field conditions at two locations with different soil and climate conditions were evaluated. The general cultivar information and experimental data on phenology and yield components have previously been described (Potopová et al., Reference Potopová, Tűrkott, Hiřmanová, Rožnovský and Litschmann2017b). Thomas F1 is early season tomato with a large red fruit (57–67 mm; weight of fruit ~120 g) which quickly ripens. Fruit does not crack even in the adverse weather conditions and is widely preferred by growers. Thomas F1 referred to as LSL (long shelf-life) with suppressed aroma formation throughout ripening. After preparing the pots, the tomato seeds were sown in the greenhouse by mid-February for each experimental year and approximately 45–55-day-old seedlings were transplanted by hand into the main field, corresponding with agrotechnical requirements, by mid-May. The spacing maintained in the main field was 1.00 and 0.50 m between rows and plants within rows, respectively. The sampled plants were collected every 14 days for analysing basic physiological parameters: LAI and RGR (Relative Growth Rate). Phenological experimental data are crucial information for calibration of a crop model and plays a central role in the portioning of assimilates. Therefore, phenology was observed weekly according to the BBCH scale (Feller et al., Reference Feller, Bleiholder, Buhr, Hack, Hess, Klose, Meier, Stauss, Boom, van den and Weber1995). Easily distinguishable visual phenological stages of the indeterminate tomato cultivars were recorded (Deligios et al., Reference Deligios, Cossu, Murgia, Sirigu, Urracci, Pazzona, Pala and Ledda2016). Site heterogeneity can cause some plants within the stand to develop at different rates, resulting in a significant time lag between individual plant's ontogenesis. Thereby, weekly phenology (%) was calculated as a ratio of plants recorded within each phenological stage and the total number of plants in the field. LAI was determined by an infrared image analysis (infrared photographs with 8 Mpx resolution). The images were processed with the analytical tool in Adobe Photoshop. The dry biomass of the sample plant was taken after drying the plant in an oven at 105°C. Parameters affecting leaf growth, dry biomass production, and dry biomass of leaves, stem and generative organs from transplanting to harvest were calibrated against the measured data.

Model input data sets

The CROPGRO-Tomato model predicts tomato growth, LAI, yield and other components in response to the soil types, weather, crop management practices and crop cultivar (Jones et al., Reference Jones, Hoogenboom, Porter, Boote, Batchelor, Hunt, Wilkens, Singh, Gijsman and Ritchie2003; Hoogenboom et al., Reference Hoogenboom, Porter, Boote, Shelia, Wilkens, Singh, White, Asseng, Lizaso, Moreno, Pavan, Ogoshi, Hunt, Tsuji and Jones2019). The algorithm used in the model is a set of differential equations representing rates of growth or development as functions of the soil-plant-atmosphere dynamics. The soil-plant-atmosphere modules include competition for light and water among the soil, plants and atmosphere. The CROPGRO-Tomato model incorporated with DSSAT was calibrated and evaluation using the field-measured data of Thomas F1 tomato variety Mochov and Praha-Suchdol sites. These simulations are conducted at a daily step. At the end of each day, the plant and soil water, nitrogen, phosphorus and carbon balances are updated, as well as the crop's vegetative and reproductive development stage. To run CROPGRO-Tomato model, we used the following four basic data set groups: (1) crop species and cultivar characteristics, (2) meteorological daily data (rainfall, global solar radiation, maximum and minimum air temperatures), (3) soil conditions and (4) cultivation technology (term of transplantation, term and dose of irrigation, fertilization and harvest).

Crop management data

Crop management data included Thomas F1 cultivar growth characteristics such as planting date, emergence date, transplanting date, plant population, plant height, leaf area measurements, anthesis date, maturity date, harvest date and yield obtained during GS of 2014, 2015, 2016, 2017, 2018 and 2020 at two experimental sites (Table 1). The seedlings were planted at 5 cm depth in the greenhouse-grown plants. The field preparation began in early-May before the transplanting of the tomato seedlings. By mid-May, Thomas F1 transplants were hardened off before transplanting to the field. The final harvest occurred from mid-September to early-October each year.

Table 1. Selected crop management practices including the calibration (Mochov 2016) and evaluation (Praha-Suchdol 2017) periods

MO, Mochov; SU, Praha-Suchdol; dap, days after transplanting.

Climate data and detection of compound weather events

The DSSAT-CSM required the minimum data set of daily maximum (Tmax, °C) and minimum temperature (Tmin, °C), incoming solar radiation (R _G, MJ/m⁻²/d) and precipitation (P, mm) to simulate crop growth and development. If the incoming solar radiation was not recorded directly, it was converted accurately for photosynthesis and potential transpiration using the Priestley–Taylor equation (Priestley and Taylor, Reference Priestley and Taylor1972). The study connected daily weather data recorded at Poděbrady and Praha-Ruzyně climatological stations (1961–2020) from the Czech Hydrometeorological Institute and daily weather variables recorded by meteorological sensors in crop canopy at field farm levels for the period 2014–2020. The R_G was calculated by Ångström-Prescott formula (Angstrom, Reference Angstrom1924) based on the fraction of daily total atmospheric transmittance of the extra-terrestrial solar radiation (R _A), a fraction of actual (n) and potential sunshine duration (N) during the day:

(1)$$R_{\rm G} = R_{\rm A} \times ( A + B \times ( n/N) ) $$

where A and B are empirical coefficients determined for the particular site (e.g. for Mochov site: A = 0.21 and B = 0.54).

The DSSAT-CSM weather module was used to arrange and incorporate all the weather data in the standard format of weatherman database. According to this data set, the CROPGRO further calculated daily potential evapotranspiration by Priestley and Taylor (Reference Priestley and Taylor1972) method.

To better understand how such impactful events may affect the filed tomato productivity, the CEs methodology was investigated (Potopová et al., Reference Potopová, Lhotka, Možný and Musiolková2021). Here, the combination of variables that lead to an extreme impact on tomato growth was referred to as a compound event. To quantify the coupling effect of temperature and rainfall anomalies on tomato growth during GS, we applied a quantile-based analysis to Tmax–P coupled data sets for the period 1961–2020. We combined the 85th percentile for daily maximum temperature (Tmax₈₅) and 35th percentile for daily precipitation (P ₃₅) and Tmax₃₅ and P ₈₅, which represented compound hot-dry events (Tmax₈₅–P ₈₅) and cool-wet, respectively. We combined the 85th percentile for daily Tmax and P, which represented compound hot-wet events (Tmax₈₅–P ₈₅). The heat stress (TS30, °C) was calculated as the Tmax excesses above the 85% quantile that met the hot day criterion. This pattern is important for tomato production because it drives canopy growth patterns.

Soil data

Specific soil parameters required for the model input, such as a lower limit, drained upper limit and saturation, drainage coefficient, and runoff curve number, were estimated from measurement of the soil profile. The soil data from the experimental sites were collected from the field based on the soil layer depths (10 cm each up to 90 cm) by using the soil standard sampling materials and kept the soil samples in air-tight zip bags and in soil cores for laboratory analysis. The physical and chemical analyses of sampled soil were performed in the soil analysis laboratory. The soil profiles for Mochov and Praha-Suchdol sites were characterized as Haplic Chernozems and Sandy Loamy Cambisol, respectively. The soil input data sets in the SBuild module included per cent of clay, silt, sand and organic carbon, pH, cation exchange capacity, slope, albedo, colour, drainage, drained upper limit (DUL), total soil nitrogen, lower limit (LL), saturated water content (SAT), hydraulic conductivity, bulk density, root growth factor (SRGF) and soil fertility factor (SLPF) (Jones et al., Reference Jones, Hoogenboom, Porter, Boote, Batchelor, Hunt, Wilkens, Singh, Gijsman and Ritchie2003). The measured soil data from both Mochov and Praha-Suchdol sites are presented in Table 2.

Table 2. Selected soil parameters for CROPGRO-Tomato model from the experimental sites

MO, Mochov; SU, Praha-Suchdol; C_org, organic carbon; CEC, cation exchange capacity; Total N, total nitrogen; LL, lower limit of water availability to plant; DUL, drained upper limit or field capacity.

Model calibration and evaluation

Crop management data set for tomato (Thomas F1) from Mochov-2016 experimental site was used to perform the calibration of the DSSAT-CROPGRO-Tomato model and evaluation of the model was done in Praha-Suchdol experimental site in 2017. Experimental information such as planting date, plant population per square meter, planning depth, row spacing and harvesting date was used as crop management practices. At the same time, data on plant phenological stages such as emergence, anthesis, LAI and AGB based on days after planting, pod formation, physiological maturity and harvest maturity were also used in the calibration process. The AGB values were expressed in dry matter for the calibration and evaluation of the model. Since the Thomas F1 variety of tomato was not included in DSSAT cultivar database, it was added as a new cultivar into the database and its parameters were populated based on the field experimental data set. We applied the newly calibrated values of ecotype file and cultivar coefficients for LAI and AGB. The simulated dates and values of the LAI and AGB were compared with the observed dates. The simulated dates and values varied in different years due to the differences in planting dates, photothermal duration, precipitation and weather-related parameters during the tomato-growing seasons. Performance statistics indicators used in this study were standard error of estimate (S _e), mean absolute error (MAE) and root mean square error (RMSE), which were calculated using Eqns (2)–(4), respectively. A lower RMSE value indicates fewer differences between the simulated and observed values.

(2)$$S_{\rm e} = \sqrt {\displaystyle{{\mathop \sum \nolimits_{i = 1}^n {( {Y_i-{\hat{Y}}_i} ) }^2} \over {n-( {k + 1} ) }}} $$

(3)$${\rm MAE} = \displaystyle{{\mathop \sum \nolimits_{i = 1}^n ( {Y_i-{\hat{Y}}_i} ) } \over n}$$

(4)$${\rm RMSE} = \sqrt {\displaystyle{{\mathop \sum \nolimits_{i = 1}^n {( {{\hat{Y}}_i-Y_i} ) }^2} \over n}} $$

where Y _i = observed value, $\hat{Y}_i$ = simulated value, $\bar{Y}_i$ = average of simulated value, $\bar{Y}$ = average of observed value, N = number of observations and n–(k + 1) = degrees of freedom.