CropLife Europe Crop Development Database: An open‐source, pan‐European, harmonized crop development database for use in regulatory pesticide exposure modeling and risk assessment

There is a regulatory need for crop development dates to assess current default values used within chemical exposure assessments as well as to justify refinements within risk assessments. However, a readily available pan‐European crop phenology database covering key FOrum for the Co‐ordination of pesticide fate models and their USe (FOCUS) crops and scenarios to meet this need is not currently available. Therefore, we describe the development of a harmonized, pan‐European, CropLife Europe Crop Development Database (C2D2), that is fully aligned with this regulatory requirement utilizing efficacy trials data generated for regulatory submissions when registering plant protection products under Regulation (EU)1107/2009. Evaluation of C2D2 against an independent data set showed good agreement for equivalent time periods, crop growth stages, and geographical regions. We illustrate how this database can be used to evaluate existing default crop development dates mandated by regulatory agencies for use within exposure assessments. Despite the large data set compiled and the geographical coverage of C2D2, not all FOCUSsw/gw scenarios have sufficient data to facilitate comparison, with less significant scenarios, like FOCUSgw Porto, being underrepresented. For those scenarios with sufficient data, clear differences between C2D2 and crop development dates assumed in the FOCUS modeling framework (using the AppDate tool) are often indicated over many growth stages, suggesting that amendment of the existing representation of crop development within the risk assessment process may be required. C2D2 is freely available under a Creative Commons license to facilitate innovation in exposure science to allow for more accurate and realistic risk assessment leading to enhanced crop and environmental protection. Integr Environ Assess Manag 2024;20:1060–1074. © 2023 CropLife Europe (Corteva Agriscience) and The Authors. Integrated Environmental Assessment and Management published by Wiley Periodicals LLC on behalf of Society of Environmental Toxicology & Chemistry (SETAC).


INTRODUCTION
Determining appropriate pesticide application windows is an important component of pesticide exposure modeling conducted for regulatory risk assessments under the Plant Protection Products (PPP) Regulation ([EU]1107([EU] /2009)).This is the case for all environmental compartments where nontarget organisms may be exposed to residues of PPPs applied to crops.Defining these windows requires an understanding of crop phenology, being the physiological development stages of crop growth from planting to harvest, the factors that drive these, and how these directly control crop productivity by mediating carbon, nutrient, water, and energy fluxes (Richardson et al., 2013).
Increasingly, regulators are constraining these application windows using estimates of the timing of crop growth stages provided by AppDate software (Klein, 2012).This is reflected in the Central Zone environmental fate and behavior guidance (v1.1.Anonymous, 2018), where, for product submissions, the use of AppDate for application timings in line with the product good agricultural practice (GAP) is mandated, subject to "expert judgment based on 'common sense' and information presented in the efficacy section."Similarly, the repair action of the FOrum for the Co-ordination of pesticide fate models and their USe (FOCUS) surface water (SW) models (European Food Safety Authority, 2020) has integrated the AppDate crop development dates into the SWASH software tool to remove subjectivity from the selection of product application windows to increase confidence in simulated exposure estimates.Users may override these default AppDate dates with justification.
AppDate software calculates consistent application dates for use in FOCUS SW and groundwater (GW) modeling with the purpose of aligning the product GAP being assessed with the fixed model crops described by the FOCUS models.The dates for major crop development stages captured in AppDate are based largely on FOCUS, which parametrized crops using local expert opinion (FOCUS, 2000), with linear interpolation used for Biologische Bundesanstalt, Bundessortenamt, und CHemische Industrie (BBCH) growth stages between each defined growth stage.Given that exposure assessments are a function of the application dates that this software generates, it is important to understand how realistic and representative they are.
More recently, an assessment of the requirements for the development of exposure scenarios for nontarget terrestrial organisms (Adriaanse et al., 2022) highlighted the following: "Another data gap involves a clear insight in crop development stages for various crops in different member states (MS) as a function of day-of-year."While the European Phenology Network was established two decades ago (Vliet et al., 2003) and there have been important developments in this area in recent years, for example, the pan-European Phenology project (Templ et al., 2018), which created a database of largely noncrop plant phenology data sets, there is currently no comprehensive, readily available, pan-European crop phenology data set available to undertake an assessment of existing default crop development dates or for use in regulatory risk assessment refinement.
The crop protection industry holds large data sets of BBCH crop growth stages within their efficacy and residue trial data sets generated to agreed standards for use in PPP regulatory submissions that span the required crop types as well as the agronomic and pedoclimatic diversity of Europe.A CropLife Europe (CLE) project was initiated to source suitable trials data sets from member companies to create a harmonized crop development database that allows for the assessment of recommended application dates and, where required, the justification of more realistic and locationspecific dates in regulatory risk assessment refinements.
While the project, when complete, will have compiled a harmonized freely available CropLife Europe Crop Development Database (C2D2) that will facilitate the comparison of recommended application windows for a wide range of rotational and permanent crops, only the dominant combinable crops are discussed within this publication.In particular, winter and spring cereals are used as example crops to illustrate key characteristics of C2D2 and to detail the kinds of comparison with AppDate that it enables.

DATA AND METHODS
The broad approach for the project was comprised of the following stages: (i) establish robust data-sharing agreements and collect data from member companies; (ii) harmonize the data received into a single coherent database for further analysis where any changes are made to facilitate this process, include change fields and codes that document these; (iii) assign standardized spatial identifiers that facilitate the use of the database for a wide range of risk refinement options in future; and (iv) statistically summarize the distribution of Jdays of the year for each crop development stage for each FO-CUSsw/gw crop and scenario (see https://esdac.jrc.ec.europa.eu/projects/focus-dg-sante for further information) to facilitate comparison with AppDate.An overarching principle of the harmonization process was to identify issues and correct these using pragmatic approaches to preserve records in C2D2 but attribute these changes such that a clear record of these is maintained along with the retained data.

Data supply
Seven member companies of CLE were approached with respect to supplying crop efficacy and residue trial data collected as part of the data package required for the registration of pesticides under the PPP Regulation (EU)1107/ 2009.Following a review of the data held by each company, it was concluded that crop efficacy trial data could be supplied, while crop residue trial data were either not available digitally or generally not stored in an appropriate format to allow efficient extraction.Each member company uploaded crop efficacy trial data into a confidential data room specific to their company, provided by CLE.These data were supplied in a format to ensure that no product-specific information was contained within the data set.The data were downloaded from the CLE data room to a secure local data room and processed by an independent consultancy in line with the data-sharing agreement signed by all companies.The harmonized C2D2 produced from these individual databases contains no attributes that would allow identification of an individual company's data contribution.Likewise, trial site coordinates were not included, to maintain the confidentiality of the trial sites and their owners in accordance with the General Data Protection Regulation.On completion of the project, the original data supplied and all intermediate derived data sets will be destroyed, and a certificate of destruction will be issued by the consultancy to each company that contributed data.This process ensures compliance with data protection as well as antitrust and anticompetition legislation.

Data processing
All data processing for the harmonization and statistical analysis to facilitate the comparison with AppDate was conducted using R statistical software (R Core Team, 2022).For all harmonization processes, a series of quality fields and codes were introduced such that any and all changes made were documented fully.The final C2D2 database comprises harmonized raw, intermediate, and final data along with the quality fields.
Processing spatial data.The area of interest (AOI) that C2D2 was intended to cover included (i) the 27 MS of the European Union (EU); (ii) the three European Economic Area (EEA) countries of Iceland, Liechtenstein, and Norway; and (iii) Switzerland and the United Kingdom.A key step in the data harmonization process was the selection of sites that were located within the AOI, which was conducted using ArcGIS Pro (Environmental Systems Research Institute, 2022).This step identified several sites that had incorrect coordinates as a result of a number of factors, including, among others, latitude and longitude coordinates switched around, longitude values that were reported as being on the wrong side of the Greenwich prime meridian, and the reported latitude and/or longitude values that were X-and Y-coordinates in other projections.Trial sites where the spatial data issue was clear and the revised location could be verified with other location descriptions provided (e.g., site address, county or region attribute) were corrected.Sites that could not be geolocated, and as such could not be assigned to a FOCUSsw/gw scenario, were included in the final harmonized database but not included in any comparative analyses.
The precision of latitude and longitude coordinate reporting for the trials sites was variable, ranging from zero decimal places (>10 km accuracy) to 10 decimal places (<10 m accuracy).This varying spatial accuracy may introduce uncertainty into spatial attribution assigned to each site as part of the harmonization process (e.g., the FO-CUSsw/gw scenario derived from a 1 km resolution raster data set).The FOCUS crop type and scenario are primary attributes by which the crop development data are summarized for use in comparative analyses such that uncertainty in their classification may introduce additional noise into the results of the analysis.The following spatial attributes were assigned to each trial site: (i) administrative zone IDs for 2021 including NUTS0 through NUTS3 IDs (EUROSTATS, 2022), (ii) FOCUSsw/gw scenario name (FOCUS, 2015(FOCUS, , 2021)) The FOCUSgw scenario extents were sourced from the European Food Safety Authority (EFSA) spatial data set v1.1 (Heiderer, 2012).This data set does not cover the full extent of the AOI and was expanded to do so using the same Worldclim v1.4 data set (Hijmans et al., 2005) and scenario rainfall and temperature ranges (FOCUS, 2014) outlined in Table S2.While a more recent version of the Worldclim data (v2.0) is available (Fick & Hijmans, 2017), the same data sets used previously were utilized to ensure consistency with existing FOCUS zone definitions, given that one of the key uses for C2D2 is envisaged to be use within the existing FOCUS models.The FOCUSsw spatial data sets available, derived using pedoclimatic factors, do not cover all arable land within the AOI, given that they define a series of realistic worst-case scenarios.As such, a FOCUSsw zone raster data set akin to that of the FOCUSgw raster data set was created using the same Worldclim v1.4 data sets and climatic factors only approach.Given the importance of temperature in driving crop phenology, rainfall and temperature combinations that were not covered by the standard FOCUSsw scenario definitions were assigned to existing scenarios within the same temperature band (see Table S3).The resulting FOCUSsw and gw scenarios maps are illustrated in Figure S1.
Processing crop attributes.Crop names were first harmonized to standardized European and Mediterranean Plant Protection Organization (EPPO) crop codes, which were subsequently assigned to FOCUSsw/gw crop groups (e.g., winter wheat and winter barley are both assigned to the winter cereals crop group).Crops that did not have a winter or spring designation in their name or EPPO code (e.g., wheat [EPPO Code: 3SWHC] rather than winter wheat [TRZAW]) were assigned a seasonal crop grouping using the month in which they were planted (see Table S1) based on expert knowledge.This was a pragmatic attempt to utilize as much of the available data as possible (<7% of wheat data were affected) but has the potential to introduce some uncertainty into the results as there is likely to be variation in the months in which crops are drilled at the scale of the AOI under consideration.Comparison of these planting dates for winter wheat with the JRC AGRI4CAST crop calendar (JRC, 2015) indicates that they are consistent with the range of winter wheat cropping dates presented (see Table S4).For trials where the drill date was absent, this could not be done, resulting in these data being included in the final database but excluded from the statistical analyses.Some crops occur in multiple FOCUS crop groups (e.g., winter barley as well as winter cereals); this results in two records being output into the C2D2 for each input record.
In all supplied databases, the crop development stages followed the BBCH nomenclature.The manner in which the BBCH crop development stages were reported by different organizations varied both between organizations and through time, typically including the dominant BBCH, the range of BBCHs (minimum and/or maximum), and/or both.The order in which these values was reported also varied and had to be decoded and harmonized as part of this process.It is suspected that the dominant BBCH in some cases was an average of the maximum and minimum observed BBCHs as the resulting value was not a recognized growth stage (<0.03% of winter wheat records).In this case, the maximum BBCH value was selected for use in the statistical analyses as the difference in days is typically small (for winter wheat, the average difference is six days and the median difference is two days).Records where the maximum BBCH value was not available or was also not a recognized growth stage were excluded from the statistical analyses.
Three possible date and associated BBCH crop development stage pairing types are reported within a crop efficacy trial: (i) the planting date, (ii) the date(s) on which the test product was applied, and (iii) the date(s) on which the efficacy was assessed.For each date, the corresponding Julian day (Jday) of the year was calculated for use with the corresponding BBCH growth stage recorded.The dates within the database were found to be a mixture of United States (month, day, year) and European (day, month, year) date formats, with no attribute indicating which format the dates were stored in for each record.For US format dates where the day was >12, this format issue was simple to identify and correct.For US format dates where the day was <12, this was less straightforward and were identified by outlier analysis (i.e., the Jday derived for the associated growth stage would produce an outlier within the distribution of available values; e.g., US format date 8/3/2000 [3 August 2000; Jday 215] with an associated BBCH of 80 would define an outlier when interpreted as 8 March 2000 [Jday 67] where growth stages are typically <BBCH 20).This approach would not be able to identify small differences in Jdays (e.g., US format date 4/3/2000 [3 April 2000; Jday 93] would possibly not produce an outlier when interpreted as 4 March 2000 [Jday 63]).As such, some of the variation within C2D2 may be a result of US format dates that have not been identified and corrected.In some cases, the year of the supplied date was inconsistent with all other dates for that record, leading to incorrect or outlier Jdays (e.g., the planting date for an autumn-sown crop was in the same year and consequently after the application and/or evaluation dates, resulting in negative Jdays).These anomalous years were corrected to be consistent with the application and evaluation dates.
The pairs of Jday and BBCH growth stage for each FOCUSsw/gw scenario and crop group were utilized within the statistical analysis for AppDate comparison.Any records that did not produce a complete pairing of Jday and BBCH crop stage were excluded from the statistical analyses.
Processing soil attributes.The climatic attributes of some FOCUSsw/gw scenarios are the same; for example, the D2, D3, D4, and R1 scenarios form one such group, while the D6 and R4 scenarios form another (see Tables 1, S2, and S3 for SW and GW, respectively).The different scenarios with similar climate describe differences in soil texture, structure, and organic matter as well as the influence of topographic and shallow GW factors.The crop development dates within the FOCUS scenarios with the same climatic ranges vary (e.g., for winter cereals, BBCH 00 is from 12 September, 15 October, 02 November, and 11 November for D4, D2, R1, and D3, respectively) and, as such, assignment of trial sites to specific scenarios required an interpretation of the topsoil description at each site.This description varied considerably, including no information, basic information on broad textural class (e.g., sandy), dominant textural class (e.g., silt loam but not necessarily reported on a consistent basis using a standardized soil classification system like the US Department of Agriculture [USDA] soil taxonomy) to detailed information describing the soil particle size distribution (PSD) and organic matter status.Detailed information was likewise not captured on a consistent basis.
For example, at some trial sites, the PSD had been normalized, summing to 100%, while in others, it had not.A series of assumptions and rulesets were developed to harmonize the topsoil data: for example, the difference in the PSD total and 100% was a function of unreported stone content; using the USDA classification system on stoniness (Ditzler et al., 2017, tab. 3-3), soils with a difference of <15% (i.e., <15% stones) were considered acceptable and the PSD normalized to 100% and then classified using the USDA soil taxonomy with the "Soil Texture" (v1.5.1)R package (Moeys, 2018).Only mineral soils were included in this assessment and peaty soils as well as artificial soils were excluded.Trial sites where no or insufficient soils data were available to assign the site to an individual FOCUS scenario were retained in C2D2 but excluded from the statistical analyses.The resulting trial site textural class, in some cases, spanned more than one FOCUS scenario, resulting in two records being output into the C2D2 for each input record, one for each FOCUS scenario.For scenarios that have a dedicated climatic zone, all trials regardless of the soil type and how closely it matches the FOCUS scenario soil that are located in that climatic zone were considered as being representative of that zone.

Evaluation against independent data
Given the operational requirements for efficacy trials and the extent of data cleaning undertaken within the harmonization process, an evaluation of the compiled C2D2 against an independent data set was undertaken.The PEP725 database (Templ et al., 2018) contains natural plant as well as crop phenology data up until 2015 and is the best available data set for this purpose.Data for winter wheat, winter cereals (wheat, barley, and rye), and maize from Germany, representing winter-and spring-sown crops, were considered.The analysis was restricted to Germany as data from this country comprise the majority of available data (e.g., >92% for winter wheat) within PEP725 for the AOI.The PEP725 data for Germany were supplied by the German Weather Service (Deutscher Wetterdienst [DWD], 2022), which has additional data available for the period 2016-2022.As such, the evaluation was undertaken using the DWD data, which relate to undeclared field-level observations within a 5 km radius of the defined weather station.As such, assigning the sites to a FOCUS scenario was not done as this assignment would not be reliable; instead, a comparison for all sites in Germany within both data sets was undertaken for both crops.The DWD data capture the earliest dates on which a crop reaches a defined growth stage (for winter wheat: 0, 10, 31, 51, 75, 87, 99) and is the equivalent of a maximum BBCH.While the primary purpose of this evaluation was to compare the C2D2 and DWD (in PEP725) data sets, including an appropriate AppDate comparator would be useful.The FOCUSsw R1 and FOCUSgw Hamburg scenarios were considered appropriate as they are regarded by the national regulator (Bundesamt für Verbraucherschutz und Lebensmittelsicherheit [BVL], 2022) to be most representative of German conditions (BVL, 2022).

Statistical analyses and AppDate comparison
Not all FOCUS crops are parametrized within each FOCUS scenario (e.g., see tab.4.2.1-1 in FOCUS, 2015).The statistical analysis and subsequent AppDate comparison were conducted for recognized FOCUS scenario and crop combinations.The crop development Jday (or range of Jdays in the case of FOCUSsw) for recognized BBCH growth stages for each crop and scenario was extracted from AppDate v3.06.Box and whisker plots summarizing the range of Jday values captured within the C2D2 for each recognized BBCH growth stage for each FOCUS scenario and crop group were produced using the R library ggplot2 (Wickham, 2016).These plots followed the standard box and whisker plot typology, with the box bounding the interquartile range (IQR) and the median value.The upper and lower extent of the whiskers are defined by the upper quartile (Q75) + 1.5 × IQR and the lower quartile −1.5 × IQR, respectively.Values outside of these upper and lower whisker values are considered potential outliers and plotted as individual points.In addition, the mean value and associated 90% confidence interval were plotted.The Jday (or mid-point and range of Jdays in the case of FOCUSsw) from AppDate v3.06 was included for visual comparison.

RESULTS AND DISCUSSION
The data supplied for inclusion into C2D2 span two decades (1999-2020), comprise >250 crops (combinable, top fruit, soft fruit, pasture, grass, ornamentals, vegetables, and herbs among others), and cover 25 EU MS, three EEA countries, and Switzerland and the United Kingdom.While Luxembourg and Malta are part of the AOI, no trial data for these countries were supplied.
While variation occurs for each combinable crop, typically, a small number of countries (France, Germany, United Kingdom, Poland, and Czechia), contribute the majority of the data, largely from the central regulatory zone (∼59%-60%) as illustrated for winter and spring cereals in Table 2.The C2D2 data span two decades and show a strong temporal bias, with the majority (79.2% winter and 80.2% of spring cereals) of the data generated in the last 10 years (see Figure S2).Most FOCUS scenarios have a good representation, although the less significant scenarios are typically underrepresented (e.g., Jokioinen for winter cereals or Porto and R2 for spring cereals); see Table 1.

Evaluation against independent data
The comparison between C2D2, DWD, and AppDate for winter wheat is illustrated in Figure 1.The DWD data set summary is similar to that from C2D2, sharing a number of key features in common: (i) the presence of outliers for all growth stages; (ii) the presence of spring dates for leaf emergence (BBCH 10); (iii) the IQRs for all BBCH stages in common are tightly defined and highly comparable; and (v) BBCH stages 0, 10, 75, and 87 are overestimated by AppDate but comparable for stages 31, 51, and 99.
The comparison between C2D2, DWD, and AppDate for winter cereals is illustrated in Figure S3.The DWD data set summary is similar to that from C2D2, sharing a number of key features in common: (i) the presence of outliers for all growth stages; (ii) the presence of spring dates for drilling (BBCH 0) and leaf emergence (BBCH 10); (iii) the IQRs for all BBCH stages in common are tightly defined and highly  (https://www.aemet.es/es/serviciosclimaticos/datosclimatologicos/fenologia), and Latvian Nature and History Calendar (Kalvane et al., 2021; https://zenodo.org/record/3982086) data sets.

AppDate comparison
An example comparison for the FOCUSgw Hamburg scenario winter and spring cereals is illustrated in Figure 2. The range of Jdays (y-axis) for each BBCH growth stage (x-axis) within C2D2 is summarized using box and whisker plots, which may be compared to the AppDate values.The plot for winter cereals (Figure 2A) was split into two data ranges for BBCH stages below 30 using a Jday of 200, enabling the visualization of spring and winter Jday ranges for winter cereals for BBCH stages below 30.For stages 0 through 20, the split in the data for visualization purposes reflects a definition issue, as what constitutes a winter or spring cereal is quite flexible from an agronomic perspective.If the conditions are conducive, a farmer may sow a spring cereal cultivar from early December.Similarly, a farmer may crop a winter cereal cultivar as late as March.Efficacy trials are conducted at real-world sites under real-world conditions and, as such, these kinds of agronomic peculiarities are captured within C2D2.In addition, some of the records were assigned a season based on cropping date as the crop type (e.g., wheat rather than winter wheat) was not season-specific.These definition-and season-specific situations introduce variation into C2D2 and the statistical analysis that considers winter and spring FOCUS crops with more rigid definitions.
For stages 21 through 30, this also reflects a number of interacting agronomic factors, specifically when the crop was drilled in the autumn.For crops drilled early and/or where the autumn is warm, these growth stages may occur in the autumn (Jdays 275-365), while in cooler autumns or when crops are drilled late, these may occur the following spring (Jdays 10-125).This reflects real-world conditions under central European conditions rather than data issues within C2D2.The AppDate crop development stages for scenarios like Hamburg are applied within the FOCUS models in a rigid sense, with stages 0 through 20 always occurring in the autumn and stages 21 through 30 always occurring in the spring.The C2D2 data analysis demonstrates how restrictive applying this assumption in all years Similarly, trials are conducted to meet a regulatory need and on occasion may be conducted at less typical times where seasonal or operational constraints have been encountered, for example, a first crop fails to establish sufficiently and is redrilled, resulting in later crop development dates than might be typical for that particular year.The extent to which the data within individual trials in C2D2 conform to typical agronomic practice in any given year is unknown; however, the practices and crop development dates captured within the database are believed to be representative of agronomic practice on the whole, especially given the size, geographic, and temporal coverage of the data set that has been compiled.
The Jdays in AppDate for BBCH 0 through 20 and 33 through 60 are in good agreement with those from C2D2.In contrast, the Jdays in AppDate for BBCH 21 through 32, and 65 through 97, systematically overestimate those derived from C2D2, which would result in later application timings.The C2D2 results for spring cereals (Figure 2B) indicate that BBCH 0 through 20 are also influenced by winter cropping, which influences the distribution and produces outliers.The Jdays in AppDate for BBCH 0 through 65 systematically underestimate those derived from C2D2 while overestimating BBCHs 80 through 97.Underestimation by AppDate would result in earlier application timings that may affect simulated exposure; for example, later applications in the autumn for winter cereals and earlier applications for spring cereals may result in applications that increasingly coincide with the drainflow period.Similar sorts of patterns are indicated for the example comparisons for FOCUsw D5 scenario winter and spring cereals (Figure 3).2018).Cultivar development in maize offers new varieties with better cold tolerance, allowing potentially slightly earlier establishment as well as being early/very early developing to aid harvest, soil management, and future crop establishment while trying to maintain similar or better yield from lower heat unit input.The graphical analyses for all combinable crops and FOCUS scenarios are provided in Figure S5 through Figure S20.A tabular summary of all data in the boxplots is also included in the MSExcel spreadsheet labeled Part B provided in the Supporting Information Materials.
A high-level qualitative assessment of the data available in C2D2 is provided in Table 3 for GW and Table 4 for SW.The results indicate that the extensive GW scenarios, like Châteaudun, Hamburg, and Kremsmünster, and SW scenarios like D3, D4, and D5, tend to be well represented in C2D2, producing a comprehensive set of Jday distributions for the majority of BBCH growth stages that enables a robust comparison of the recommendations in AppDate.This is a function of the increased probability that efficacy trials will be conducted within these scenario climates, given their prevalence.The results suggest that crop-scenario combinations with more than ∼2000 records contain sufficient   data to conduct a comprehensive comparative assessment.
In contrast, there are many FOCUS crops (e.g., FOCUSgw Peas or FOCUSsw spring oilseed rape), where there are insufficient data available in C2D2 to undertake an assessment of existing default values.Crop-scenario combinations with less than ∼500 records contain insufficient data to conduct even a rudimentary comparative assessment.Across all the combinable FOCUS crops and scenarios for both GW and SW, there are sufficient data in C2D2 to facilitate robust comparisons for the majority of crop-scenario combinations.While gaps for key combinations remain (see Tables 3 and 4), it is likely that generalized observations can be drawn from the crop-scenario combinations for which there are data to fill these.

C2D2 and crop modeling
Vogel (2022), assessing changes in plant phenology for sites in southern Europe with at least 30 years of observations, noted the following for crops: • maize, redcurrants, vines, and potatoes showed the highest advances in leaf unfolding; • potatoes, olives, and vines showed the highest advances in flowering; and • oats, maize, barley and wheat showed the highest advances in fruiting.
These changes were attributed to a changing climate.Similarly, Rezaei et al. ( 2018) found a 14%-18% decline in the temperature sum required from emergence to flowering for modern cultivars of winter wheat compared with the cultivars grown in the 1950s and the 1960s when analyzing a long-term  data set of phenological observations across western Germany.They conclude that using a "single-cultivar concept" in climate change impact assessments results in an overestimation of winter wheat sensitivity to increasing temperature, requiring the consideration of changing cultivars to draw appropriate conclusions.The data compiled within C2D2 were derived predominantly between 2010 and 2020 and as such are highly relevant for use in simulated exposure assessment as they capture crop phenology under prevailing climatic conditions for modern crop cultivars, whereas the crop development data in FOCUS and AppDate reflect expert opinion from the 1990s.Menzel et al. (2020), when assessing long-term phenological data , noted that while farmers were responding to a range of factors, they had clearly adapted their practices in response to warming springs and autumns, which was reflected in their crop phenology data set in the two most recent decades.Jägermeyr and Frieler (2018) demonstrated that accounting for observed spatial variations in growing seasons increased the fluctuations in reported national maize and wheat yields that can be explained by process-based crop yield model simulations.These studies highlight the need for the adoption of dynamic crop models instead of the static crop models currently included within the FOCUSsw/gw models.This would allow for more realism within the modeling frameworks with farmer behavior and crop phenology varying in each year.Reflecting such crop and agronomic developments in the models will allow the risk assessments to assess product GAPs in early spring in those years where this would occur.C2D2 has the potential to assist with the parametrization and calibration of more dynamic crop models in FOCUS with varying levels of complexity.At the simplest level, the key crop development dates for each FOCUS scenario could be revised; at an intermediate level, the growing degree days between the key crop development stages could be determined (PEARL can already use heat sums to dynamically grow crops); and at an advanced level, the data could be used to calibrate more sophisticated crop models, like WOFOST (Ceglar et al., 2018).Formal changes to the FOCUS models would require the approval of EFSA and implementation in the models via the FOCUS model Version Control Group.In the interim, FOCUS model users can use C2D2 to justify overriding the FOCUS default values captured in AppDate.Similarly, C2D2 could be used to justify the selection of dates for other ecological risk assessment purposes, for example, to define periods of crop flowering used in pollinator risk assessments.Supplementing C2D2 with crop phenology data from other data sets (PEP725, DWD, TEMPO, AEMet, Kalvane et al., 2021) should be considered for a future version of the database.

CONCLUSION
Exposure assessments conducted as part of regulatory risk assessments for the registration of PPPs are increasingly utilizing default or standardized crop development dates in order to remove subjectivity in deciding when the crop development stages mandated on the product GAP occur.While there have been important developments in the crop phenology area in recent years (Templ et al., 2018), there is currently no comprehensive, readily available, pan-European crop phenology data set available to undertake an assessment of existing default crop development dates or for use in regulatory risk assessment refinement.A CLEfunded project has facilitated the collation and harmonization of crop protection product company efficacy trial data sets to produce a crop development database, C2D2, which meets these regulatory needs for many FOCUSsw/gw crops and scenarios.Despite the large data set compiled and the geographical coverage of C2D2, not all FOCUSsw/ gw scenarios have sufficient data to facilitate comparison, where, for those scenarios with sufficient data, significant differences between the CLE and AppDate crop development dates are often indicated over many growth stages, suggesting that amendment of the existing representation of crop development within the risk assessment process may be required.The C2D2 database holds significant opportunities for future development of crop models to improve estimates and timing of exposure, not only in the context of exposure modeling but also, for example, to help predict accurate timing of applications with respect to on-and offfield weeds and nontarget arthropods, especially bees.C2D2 is freely available under a Creative Commons license to facilitate innovation in exposure science to allow for more accurate and realistic risk assessment leading to enhanced crop and environmental protection.

FIGURE 1
FIGURE 1 Example plots comparing the AppDate BBCH growth stages (blue dots = recognized growth stages) for Hamburg winter wheat in Germany with (A) the evaluated data collated within C2D2, (B) the DWD data set, and (C) C2D2 and DWD for BBCH crop growth stages in common.For BBCH stages up to 23, the data are split into spring and winter crop ranges using a Jday value of 200.The gray band represents the 90% confidence interval of the mean (cross).In Panel (C), the green diamond represents the corresponding AppDate value.BBCH, Biologische Bundesanstalt, Bundessortenamt und CHemische Industrie; C2D2, CropLife Crop Development Database; DWD, Deutscher Wetterdienst

FIGURE 2
FIGURE 2 Example plots comparing the AppDate BBCH growth stages (blue dots = recognized growth stages) and the evaluated data collated within the C2D2 data set for the FOCUSgw Hamburg scenario for the FOCUS crops (A) winter and (B) spring cereals.For winter cereals, BBCH stages up to 29 the data are split into spring (yellow boxes) and winter (green boxes) crop ranges using a Jday value of 200.The gray band shows the 90% confidence interval of the mean (cross).The number of samples within each growth stage considered within the plotted distribution is provided in red just above the x-axis.Box and whisker plots are not produced for BBCH stages with <10 records available.BBCH, Biologische Bundesanstalt, Bundessortenamt und CHemische Industrie; C2D2, CropLife Crop Development Database; FOCUS, FOrum for the Co-ordination of pesticide fate models and their USe

FIGURE 3
FIGURE 3 Example plots comparing the AppDate BBCH growth stages (blue dots = midpoint of recognized growth stages window, the green band) and the evaluated data collated within the C2D2 data set for the FOCUSsw D5 scenario for the FOCUS crops (A) winter and (B) spring cereals.For winter cereals, BBCH stages up to 29 the data are split into spring (yellow boxes) and winter (green boxes) crop ranges using a Jday value of 200.The gray band is the 90% confidence interval of the mean (cross).The number of samples within each growth stage considered within the plotted distribution is provided in red just above the x-axis.Box and whisker plots are not produced for BBCH stages with <10 records available.BBCH, Biologische Bundesanstalt, Bundessortenamt und CHemische Industrie; C2D2, CropLife Crop Development Database; FOCUS, FOrum for the Co-ordination of pesticide fate models and their USe

FIGURE 4
FIGURE 4 Example plots comparing the AppDate BBCH growth stages (blue dots = midpoint of recognized growth stages window, the green band) and the evaluated data collated within the C2D2 data set for maize for the FOCUS (A) SW D3 and (B) GW Hamburg scenarios.The gray band is the 90% confidence interval of the mean (cross).The number of samples within each growth stage considered within the plotted distribution is provided in red just above the x-axis.Box and whisker plots are not produced for BBCH stages with <10 records available.BBCH, Biologische Bundesanstalt, Bundessortenamt und CHemische Industrie; C2D2, CropLife Crop Development Database; FOCUS, FOrum for the Co-ordination of pesticide fate models and their USe; GW, ground water; SW, surface water Summary of the countries and regulatory zones within which the FOCUSsw/gw WC and SC trials were located Integr Environ Assess Manag 2024:1060-1074 © 2023 CropLife Europe (Corteva Agriscience) and The Authors DOI: 10.1002/ieam.4870TABLE 2 C2D2: CROP DEVELOPMENT DATABASE FOR RISK ASSESSMENT-Integr Environ Assess Manag 20, 2024 comparable; and (v) BBCH stages 31 and 87 are overestimated by AppDate but comparable for stages 0, 10, 51, 61, 65, and 99.The comparison between C2D2, DWD, and AppDate for maize is illustrated in

TABLE 3
Summary of the number of records in C2D2 for each FOCUSgw crop type and scenario underpinning the BBCH distributions of Julian days Qualitative assessment of the amount and BBCH growth stage coverage of the data available in C2D2 to undertake a comparison with AppDate.Comprehensive-The majority of recognized growth stages have data; sample numbers underpinning the distributions described by the box and whisker plots are good for the majority of BBCH growth stages.Moderate-Many of the recognized growth stages have data; sample numbers are reasonable for most BBCH growth stages.Poor-Few of the recognized growth stages have data; sample numbers are low for most BBCH growth stages.Where scenario-crop combinations have attributes that straddle the boundary between two of the classes outlined above, a dual class is assigned, with the first assigned class being the more dominant Abbreviations: BBCH, Biologische Bundesanstalt, Bundessortenamt und CHemische Industrie; C2D2, CropLife Crop Development Database; FOCUS, FOrum for the Co-ordination of pesticide fate models and their USe; ND, no data; NP, FOCUS crop not present in this FOCUSgw scenario; OSR, oilseed rape.Summary of the number of records in C2D2 for each FOCUSsw crop type and scenario underpinning the BBCH distributions of Julian days Qualitative assessment of the amount and BBCH growth stage coverage of the data available in C2D2 to undertake a comparison with AppDate.Comprehensive-The majority of recognized growth stages have data; sample numbers underpinning the distributions described by the box and whisker plots are good for the majority of BBCH growth stages.Moderate-Many of the recognized growth stages have data; sample numbers are reasonable for most BBCH growth stages.Poor-Few of the recognized growth stages have data; sample numbers are low for most BBCH growth stages.Where scenario-crop combinations have attributes that straddle the boundary between two of the classes outlined above, a dual class is assigned, with the first assigned class being the more dominant.Abbreviations:BBCH, Biologische Bundesanstalt, Bundessortenamt und CHemische Industrie; C2D2, CropLife Crop Development Database; FOCUS, FOrum for the Co-ordination of pesticide fate models and their USe; ND, no data; NP, FOCUS crop not present in this FOCUSgw scenario; OSR, oilseed rape.