1. Introduction

1Agricultural productivity has considerably increased thanks to the intensive use of chemical inputs, namely synthetic fertilisers and phytopharmaceutical products (PPP, commonly known as pesticides). Nevertheless, the new crop protection practices had as corollaries issues concerning health (Inserm, 2021), the environment (Mamy et al., 2022) and the resistance developed by the targeted organisms (Hawkins et al., 2019). In order to keep people informed about the evolvement of the practices, data on PPP purchases in France has been published annually at the scale of ZIP codes. This scale could be too broad, especially in rural areas which are mostly concerned by these issues (see map 8 in the Supplemental Materials).

3Changing the scale of spatial data is a complex matter. With the identification of the “Modifiable areal unit problem” by Openshaw et Taylor (1979), two biases should be taken into consideration: i) the scale effect, and ii) the zoning effect. In our case, the aggregation of the information on individual purchases at the ZIP code level smooths out (scale effect) the local disparities. We aim at retrieving this information through the use of data on the precise location of agricultural activities. The spatial distribution of these activities could be uneven among the municipalities sharing the same ZIP code. For instance, we could have two municipalities with the same ZIP code, one specialised in horticulture (municipality A) and the other – in cereals (municipality B). Products specific to cereals bought in the ZIP code should be associated with B rather than with A. Would the crop mix be identical for the two municipalities, then the disaggregation consists simply in a distribution depending on the agricultural areas of each municipality.

4Spatialising at the plot level also allows to limit the zoning effect by ignoring the areas not concerned by pesticides or by applying a “mask” over the territories where pesticides are not authorised. Furthermore, in line with Eagleson et al. (2002), the reaggregation of the information from the parcels toward other scales, more appropriate with respect to environmental issues, renders the data interoperable and more relevant for further analyses. However, this change of scale is based on multiple assumptions concerning pesticides’ use and the localization of crops. We will discuss further in detail the limitations of our methodology as well as those induced by the entry data. It is important to highlight that the spatialisation produces model results and not actual observations. These results are nevertheless an interesting proxy of pesticides’ use and could be valuable for different kinds of analyses, namely in epidemiology, agronomy, ecology and economics when dealing with issues concerning public health, agricultural systems, nonpoint source pollution and environmental damages. In this sense, the model results and their mapping are mostly addressed to scientific personnel and their interpretation in the public debate should be supported by experts’ reading.

5Other spatialisations of the BNV-D data have been conducted recently. Their aim is to assess the exposition to pesticides and the risks associated with it. Rigal et Perrot (2025) spatialise the quantities of active substances at the municipality level by taking into account the agricultural areas from the 2020 census data and apply what they call “standard” rates without information on the actual crops present in the municipality. Galimberti et al. (2025) spatialise the data at the plot level. They do not consider the application rates; they just take into account whether the product is authorised for a given crop. Furthermore, their agricultural parcel layer is incomplete since it only builds on data from the Land parcel identification system (LPIS, see Section 2.2.1). Our approach differs from these two works because: i) our agricultural parcel layer is the most comprehensive one (see Section 2.2.2), ii) we take into account the products’ application rates, and iii) the crops is actually present in the municipalities.

2. Data

6Unlike the data used by Ramalanjaona et al. (2020) and presented by Lungarska et al. (2023), we use one information that is publicly available. We made this choice for two reasons: i) the reproducibility of the results and ii) the possibility to share the results from the spatialisation. The downside is that the results are less precise because of potential discrepancies in the association of parcels and ZIP codes. We further discuss this bias in Section 7 detailing the limits of our approach. Hereafter, we describe the data briefly. We also provide small extractions of the data in the annex. Figure 2 presents how these data are employed throughout the modeling chain.

2.1. Data on phytopharmaceutical products and computation of the application rates

2.1.1. BNV-D “Registre”

8Table 1 summarises the number of substances, CAS identifiers, products and ZIP codes of the BNV-D “Register” per year and for the period as a whole.

Table 1: Total of substances, CAS identifiers, AMM and ZIP codes in the BNV-D “Register”

Year	Substances	CAS	AMM	ZIP codes
2015	492	439	2 844	6 027
2016	495	433	2 787	5 994
2017	506	438	2 798	5 987
2018	507	432	2 775	6 005
2019	493	424	2 616	5 957
2020	490	424	2 603	5 928
Total (unique)	609	515	4 140	6 139

2.1.2. E-Phy

9The E-Phy database is provided publicly by the Anses (Agence nationale de sécurité sanitaire de l’alimentation, de l’environnement et du travail). It is a catalogue of PPP and the terms of their authorised usage in France. This database contains the information concerning PPP and, for instance, their composition, the producer, the starting date of marketing, the year of ban (if any), the crops and the maximal rate of application depending on the targeted pests. Imported product that are approved in other EU Member states are associated with products considered standard (E-Phy table Permis de commerce parallèle). The application rates and units are harmonized and expressed in kg/ha or L/ha when a conversion factor is known. The data as well as its documentation are provided at https://data.gouv.fr and also at https://ephy.anses.fr/​.

2.1.3. Annual agricultural statistics (SAA)

10We also use annual data on the area devoted to different crops per region (equivalent to the EU NUTS 2 level). This information comes from the French agricultural ministry’s statistical unit via the web portal Agreste. Thanks to these data, we can weight the E-Phy application rates per group of crops (see the description provided in Section 3.2). Table 2 details the matching between the groups of crops used in the spatialised version of the BNV-D (the BNVDs), the crops in E-Phy, and those in the SAA. The latter data and its description are provided at the Agreste website: https://agreste.agriculture.gouv.fr/​agreste-web/​disaron/​SAA-SeriesLongues/​detail/​.

Table 2: Matching between crops as presented in the BNVDs, E-Phy, and SAA

Crop BNVDs	Crop E-Phy	Crop SAA
BETT	Industrial and fodder beetroot, seed-bearing beetroot	Industrial beetroot, table beet/beetroot
BLE	Wheat, cereals, straw cereals	Wheat, triticale
CEREALES	Oat, cereals, straw cereals, buckwheat, rye	Other cereals, oat, rye and mixed cereals
COLZA	Oleaginous cruciferae (brassicaceae)	Rapeseed and turnip rape
COQUE	Chestnut, fruit production, nuts, hazel, walnut	Chestnut, nuts, walnut
DIVERS	Trees and shrubs
FEV	Proteaginous seeds	Field beans and beans
FIBRE	Hemp, flax, seed-bearing fiber plants	Hemp (paper), flax (textile), fiber plants (including seeds)
FOURR	Cereals, cabbages, maize, fodder seeds, seed-bearing fodder crops	Fodder cabbage, fodder maize, artificial and temporary grasslands
GEL	Fallow and intermediate crops	Fallow
INDUS	Sugar cane, sweet herbs, hop, infusion plants, aromatic and medicinal plants, perfume plants, tobacco, seed-bearing aromatic and medicinal plants, flowers and garden plants	Sugar cane, hop, aromatic plants, medicinal and perfume plants (excluding seeds), tobacco
LEGU	Artichoke, asparagus, ornamental bulbs, carrot, celery, chicory, cabbage, cucumber, watercress, cucurbits,flowers, vegetables crops, ornamental crops, spinach, strawberry bush, raspberry bush, beans and peas, lettuce, root vegetables and tropical tubers, melon, turnip, onion, house plants, leek, bell pepper, rose tree, salsify, tomats, eggplant	Artichoke, asparagus, bulbs (onion, tubers, rhizome), melons, carrot, celery, chicory, cabbage, cucumber, watercress, spinach, flowers and ornamental plants, strawberry, raspberry, beans, lettuce, fresh vegetables, turnip, shallot, pea, leek, bell pepper, chilli, okra, salsify and scorzonera, tomatoes
LFOURR	legumes, seed-bearing legumes, grassland	Artificial grassland (alfalfa, red clover, ...)
LGRAIN	Beans, leguminous plants (garden), peas	Dry beans (including seeds), lentils (including seeds), dry peas (including seeds)
MAIS	Cereals, maize, seed-bearing maize, sorghum	maize, sorghum
OLEA	Flax	Flax
OLIVE	Fruit crops, stone fruit, olive tree	olives
ORGE	Cereals, straw cereals, barley	Barley and bere
PDT	Potato	Potatoes
POIS	Protein peas, peas	Protein peas
PP	Grassland	Natural grasslands or 6 years after sowing
PROTEA	Protein seeds	Sweet lupine
PT	Grasses, grassland	Temporary grassland
RIZ	Cereals, rice	Rice
SOJA	Groundnut, soya	Soya
STH	Grasses, grassland	Constantly grassed plots, poorly productive pastures
TOUR	Sunflower	Sunflower
VERGERS	Citruses, Almond tree, Pineapple, Avocado tree, Banana tree, star fruit, blackcurrant, cherry tree, soursop, fruit crops, tropical crops, common fig, passion fruit, stone fruits, common guava, persimmon, kiwi fruit, lychee, mango tree, papaya, peach tree, apricot tree, small fruits, apple tree, plum tree	Apricots, kiwi, citruses, almonds, pineapple, avocado, plantain banana, blackcurrant and blueberries, cherry, soursop, apples, figs, stone fruits, tropical and sub-tropical fruits, lychee, longan, rambutan, mango, maracuja, passion fruit, peach, small fruits apple (including for cider), plum
VIGNES	Vineyard	Vineyard

Table 2 has been translated by the authors. The original version in French is given at the end of the supplementary material section (table 21). The crops’ denomination from the E-Phy and SAA data sets correspond to precise definition of the crops, and their translation here is only indicative. Furthermore, some entries were redundant and were removed for convenience in the English version.

2.2. Data on land use

11In order to distribute the sales, we use fine scale data on the areas susceptible to phytopharmaceutical treatments. We are mostly interested in agricultural plots but also in non-agricultural zones subject to weeding.

2.2.1. Land parcel identification system

12The French data from the Land parcel identification system (LPIS, in French RPG) summarises the information provided by farmers in the context of the Common agricultural policy (CAP) of the EU. Since 2015, the data is publicly available thanks to the IGN (French national institute for geographic and forestry information) at the agricultural plot level, considered homogeneous with respect to the primary crop declared. LPIS covers most arable crops very well but is less exhaustive when it comes to vineyards, orchards or marketed gardening (see table 19). For a full access to documentation and data, check the Géoservices’ portal: https://geoservices.ign.fr/​rpg.

2.2.2. RPG complété

13The RPG complété layer was proposed by Cantelaube & Lardot (2021) and Lardot et al. (2021) in order to complete the gaps in the LPIS. The RPG complété is constructed on the basis of the cadastral parcels. A land use is then attributed to the parcel depending on different sources of information combined through a statistical model. Among the data used, we can cite the BD Topo database proposed by IGN and the OSO (Land use, Occupation des sol) layer derived from satellite imagery (Thierion et al. 2022). The documentation of the RPG complété is available here: https://hal.inrae.fr/​hal-03818008.

14For instance, the RPG complété data is available for the year 2020 here: https://entrepot.recherche.data.gouv.fr/​dataverse/​rpg_complete_2020.

15Figure 1 presents the vineyards area [VIGNES], vegetables [LEGU], orchards [VERGERS], fodder crops [FOURR], and olive groves [OLIVE] at the French metropolitan scale and the contribution of RPG complété in terms of percentage. These are the groups of crops for which more than 10% of the area is made available by the RPG complété. In the case of vineyards, for instance, the total area is 782 thousand hectares (see table 19 in the Supplemental Materials) and 25% of the area is covered in the RPG complété while the rest is mapped in the LPIS. The RPG complété is of particular importance for vineyards in regions where farms are exclusively specialised in viticulture and do not declare their parcels to LPIS. This is namely the case for the Marne (51) département (NUTS3 level) as shown in the figure 9 and the map 7. For instance, 4 of the municipalities in the ZIP code 51190 are divided into two groups with the majority of their agricultural areas devoted either to vineyards or to arable crops. The RPG complété is the source of information for more of 90% of the area in municipalities specialised in viticulture.

2.2.3. Non-agricultural zones (ZNA)

16The non-agricultural zones are derived from the IGN’s BD Topo database. They include the built-up areas, public and family gardens, roads (and railways), cemeteries, airports, sports equipment and other activity zones. The description of the BD Topo database is available at the Géoservices’ portal: https://geoservices.ign.fr/​bdtopo.

3. Spatialising at the plot level

17The spatialisation is organised into four stages: i) the construction of the land use layer; ii) the calculation of application rates of reference per land use category; iii) calculation of distributional weights per hectare; iv) spatialisation of the sales’ data; and v) aggregation depending on the target scale. These different stages are presented below in the case of municipalities. The method is presented in figure 2. Table 3 provides information on the volume of data treated and produced.

Table 3: Volume of the treated and produced data

Subject	Indicator	Value
BNV-D	Number of sales (couple ZIP code and AMM)	950 725
	Number of distinct ZIP codes	5 928
	Number of distinct AMM	2 603
	Number of distinct active substances	490
	Total QSA (in tons)	62 991
E-Phy	Number of distinct AMM	15 206
	Number of distinct AMM with a spatialisable application rate	13 014
Application rates	Number of distinct application rates (AMM, land use, region)	231 563
	Number of distinct AMM with at least one application rate	2 432
Land use layer	Number of polygons	15 906 167
	Total area (in 1 000 ha)	34 012
	ZNA area (in 1 000 ha)	3 974
	Agricultural area covered (in 1 000 ha)	29 689
	Other agricultural area (in 1 000 ha)	349
Weights	Number of weights (distinct ZIP code, AMM, land use, and region)	51 230 322
BNVDs	Number of sales (couple ZIP code and AMM)	916 406
	Number of distinct ZIP codes	5 515
	Number of distinct municipalities	34 547
	Number of distinct AMM	2 404
	Number of distinct active substances	434
	Total QSA (in tons)	61 581	0.97762	2.24%
			Ratio between QSA BNVDs/BNV-D	% loss

3.1. Construction of the land use layer

Table 4: Groups of crops for the LPIS (RPG) and the RPG complété as represented in the BNVDs

BNVDs’ crop group	Label (translated)	LPIS (RPG) crop code	RPG complété crop code (translated)
BETT	Non fodder beetroot	BTN	Beetroot
BLE	Wheat	BTH, BDT, BDP, TTH, EPE, BDH, BTP, TTP	Wheat
CEREALES	Other cereals	CHH, CPA, CGF, CAG, CHA, CGS, CPZ, CHS, CHT, AVH, CPH, CPT, SGH, CPS, SRS, AVP, SGP, CGP, MCR, CGH	Other cereals, cereals
COLZA	Rapeseed	CZP, CZH	Rapeseed
COQUE	Nuts	CTG, NOS, NOX, PIS, CAB	Chestnut, nuts, walnut
FEV	Field bean	FVL, FVT
FIBRE	Fiber crops	LIF, CHV	Hemp, fiber crops
FOURR	Grasses and fodder crops	CAF, BVF, RDF, GFP, FAG, PAT, FLO, CHF, CPL, PP6, PP5, FSG, DTY, XFE, GAI, PH6, NVF, PH5, FET	Fescue, fodder crops
GEL	Fallow	J6P, JNO, J6S, J5M	Fallow
INDUS	Industrial crops	BAR, EST, VNB, YLA, MTH, ROM, PMD, MAV, CIB, SRI, MOT, FNO, MLI, VNV, CRF, CMM, ANE, CHR, HBL, VNL, PPP, SGE, ANG, TOT, FNU, CUR, OSE, PAR, BRH, CAV, PPA, PSN, MRJ, TAB, PSL, CUM, CRD, MLP, CML, BAS, PSY, VAL, PPF, THY, LAV, ANI	Industrial crops (including under cover), verbena, hop, lavender, aromatic plants, ginkgo biloba, rosemary, sage
LEGU	Vegetables or flowers	POR, GER, CES, TOM, PVP, FLA, CAR, PAS, AIL, CEL, CRA, PMV, LBF, CCT, MAC, CSS, DOL, CMB, HSA, HPC, HAR, PPO, MLO, CRN, CCN, ART, FLP, TOP, AUB, PAN, CHU, EPI, MRG, PSE, BUR, CRS, NVT, ROQ, BLT, SFI, FRA, RDI, OIG, POT, LSA, PAQ, RUT, VER	Flowers (including under cover or indoors), strawberries, community garden, vegetables (including under cover, indoors or in field) nursery (plants and flowers), tomatoes
LFOURR	Fodder legumes	GES, ME5, FFO, LOT, VED, PFP, LH6, LU5, SE5, MH5, ME6, TR5, MC5, LUD, FF6, SA6, LFH, MC6, LUZ, JO5, JOD, LP5, SAD, MEL, LFP, TRD, VE6, JOS, LEF, JO6, LP6, SAI, VES, SED, TRE, ML5, SE6, FF5, MIN, MED, SER, SA5, MLD, TR6, MLG, LU6, PFH, ML6, LH5, MH6, VE5	Legumes, leguminous plants, alfalfa, sainfoin
LGRAIN	Grain legumes	PCH, LEC	Grain legumes
MAIS	Maize	MOH, MIE, MLT, MCT, CGO, SOG, MID, MIS	maize, miscanthus
OLEA	Other oleaginous	NVH, OEH, MOL, OHN, LIP, OEI, NVE, OAG, OHR, OPN, LIH, OPR	Other oleagionous
OLIVE	Olive tree	OLI	Olive tree (including nursery for olive trees)
ORGE	Barley	ORP, ORH	Barley
PDT	Potatoes	PTC, PTF	Potatoes
POIS	Protein peas	PHI, PPR, PPT
PP	Permanent grassland	PPH, PRL	Grassland, (including grassland used by horses, cattle or in between crops in long crop rotations, and orchards)
PROTEA	Other protein crops	MPT, LDP, FEV, LDT, PAG, MPC, LDH	Protein crops
PT	Temporary grassland	RGA, PTR	Temporary grassland, ryegrass
RIZ	Rice	RIZ
SOJA	Soya	SOJ, ARA	Soya
STH	Pastures and heathland	SPH, SPL, BOP	woodland pastures, heathland, pastures
TOUR	Sunflower	TRN	Sunflower
VERGERS	Orchards	BCA, ANA, VRG, AVO, VGD, BEI, PFR, AGR, BCI, BCP, CAC, BCF, PWT, BEF, PRU, BER, CBT, PVT, BCR, BEA, BEP, PEP	Citruses, almond, nursery (orchard), small red berries, orchards
VIGNES	Vineyards	VRT, RVI, VRC	Vineyard (table grapes, wine grape, restructuring)
DIVERS	Various		Shrub, berry, tree nursery (including wood and fir), heather
SEMENCES	Seeds		Seeds

19The crops’ designation in table 4 directly from the data source, sometimes with multiple names referring to a single crop. As for table 2, we translated the source material, with table 22 the original version of it. Likewise, the translation given here is purely indicative, and should not be trusted to recreate the groups described hereafter.

3.1.1. Assigning municipality’s code and ZIP code to parcels

20Each parcel is associated with a unique municipality and its code. When a parcel overlaps two or more municipalities, we assign to it the municipality where lays the most of its area. Then, following the rules set in Ramalanjaona et al. (2020), we attribute a unique ZIP code considered majoritarian. This somewhat simplifies the problem with the initial BNV-D scale. Indeed, multiple ZIP codes can be present in the same municipality, as it is often the case for big agglomerations. On the other hand, multiple municipalities can be associated with the same ZIP code which is rather the case for rural communes. We can observe that some ZIP codes of the BNV-D do not have a match in the land use layer. This can be the result from the elimination of the ZIP codes that are not majoritarian but also professional ZIP codes (CEDEX). In order to spatialise these “orphan” sales, we redistribute the quantities associated to the other ZIP codes by means of associated municipalities (see figure 3).

21Table 5 presents some indicators of the compatibility between the different database used for the spatialisation of the BNV-D. We can observe that the match between AMMs from E-Phy and the BNV-D is better for more recent years. This can be the result from the efforts in improving the quality of the two databases as well as the fact that we are using E-Phy’s version from 2023. As for the ZIP codes, the match between the BNV-D and the land use layer remains rather stable around 97%. The impact of this indicators on the actual spatialisation are detailed in Section “Results from the spatialisation” further below.

Table 5: Indicators on the compatibility of the data used

Indicator	2015	2016	2017	2018	2019	2020
Total number of AMMs	2 844	2 787	2 798	2 775	2 616	2 603
Number of AMMs in E-Phy	2 681	2 645	2 686	2 685	2 558	2 565
% of the total	94.3%	94.9%	96.0%	96.8%	97.8%	98.5%
Number of substances	492	495	506	507	493	490
Number of ZIP codes other than ‘00000’ and ultramarine	5 974	5 970	5 964	5 939	5 906	5 906
Number of ZIP codes known in the land use layer	5 776	5 777	5 776	5 762	5 720	5 722
% of the total	96.7%	96.8%	96.8%	97.0%	96.9%	96.9%

3.2. Calculation of application rates of reference

22The calculation of the application rates of reference is made on the basis of the data sources described before. The first stage consists in retaining a unique application rate for each couple “product – crop” from E-Phy. The initial data propose an application rate for each product and each combination of “crop – the part of the plant treated – pest” associated. The unique application rate is the median one. Some of the rates and their units are not spatial (e.g., g/L). In these cases, we apply a conversion factor in accordance with Ramalanjaona et al. (2020).

23Note: in tables 6 to 10, the usage and crop columns come respectively from the E-Phy and LPIS database, thus, they are referring to precise definitions and are not translated. Readers will find a short and approximative translation below each relevant line of the table.

24Table 6 presents an example for two products for which we suppose there are no other uses.

Table 6: Example of application rates from E-Phy for two products

Product	Usage	Crop	Rate (L/ha)
Product A	Vigne*Trt Part.Aer.*Oïdium(s) Vineyard*treatment of aerial part of the plant*oidium	Vigne Vineyard	2
	Vigne*Trt Part.Aer.*Rougeot parasitaire Vineyard*treatment of aerial part of the plant*rotbrenner		1.5
	Vigne*Trt Part.Aer.*Mildiou(s) Vineyard*treatment of aerial part of the plant*downy mildew		2
	Vigne*Trt Part.Aer.*Excoriose Vineyard*treatment of aerial part of the plant*Eutypa dieback		1.5
	Vigne*Trt Part.Aer.*Black rot Vineyard*treatment of aerial part of the plant*black rot		1.5
Product B	Artichaut*Trt Part.Aer.*Chenilles phytophages Artichoke*treatment of aerial part of the plant*Phytophagous caterpillars	Artichaut Artichoke	0.075
Product B	Asperge*Trt Part.Aer.*Chenilles phytophages Asparagus*treatment of aerial part of the plant*Phytophagous caterpillars	Asperge Asparagus	0.125
Product B	Betterave industrielle et fourragère*Trt Part.Aer.*Coléoptères phytophages Industrial and fodder beetroot*treatment of aerial part of the plant*Phytophagous beetle	Betterave Beetroot	0.05

25At the end of phase 1, we have the result presented in table 7.

Table 7: Median rate for two products on the basis of E-Phy data

Product	Crop	Median rate(L/ha)
Product A	Vigne Vineyard	1.5
Product B	Artichaut Artichoke	0.075
Product B	Asperge Asparagus	0.125
Product B	Betterave Beetroot	0.05

26The second phase involves the information from the annual agricultural statistics. For each region, we keep the area devoted to each of the crops surveyed. This way, we compute an average rate considered “of reference” for the crop groups of the BNVDs by weighting the median rates of crops by their area in the region.

27Following the preceding example, let us assume the land use ratios for 2 regions as those presented in table 8.

Table 8: Proportion of different crops in the regional land use for the crop group Légumes of the BNVDs

	Crop	Proportion	Crop group BNVDs
	Artichaut Artichoke	30%	Légumes Vegetables
	Asperge Asparagus	70%
	Artichaut Artichoke	50%	Légumes Vegetables
	Asperge Asparagus	50%

28Thus the rate of reference for the product B is, depending on the region, 0.11 or 0.1 (table 9).

Table 9: Calculation of the rate of reference per region

Product	Crop	Crop group BNVDs	Region	Rate of reference
Product B	Artichaut Artichoke	Légumes Vegetables	Region 1	0.3 * 0.075 + 0.7 * 0.125 = 0.11
	Asperge Asparagus
	Artichaut Artichoke	Légumes Vegetables	Region 2	0.5 * 0.075 + 0.5 * 0.125 = 0.1
	Asperge Asparagus

29The last phase of the calculation of the application rates of reference concerns the products which can be used for “general treatments”. Indeed, besides the usages on specific crops, some products can be employed as “general treatments” as stated in E-Phy. We take this into account by averaging the rate labelled “general treatments” and the rate calculated in the preceding steps.

30Let us assume that product B has the following usages (table 10).

Table 10: Accounting for the “general treatments” case in the calculation of the rates of reference

Product	Usage	Crop	Rate (L/ha)
Product B	Artichaut*Trt Part.Aer.*Chenilles phytophages Artichoke*treatment of aerial part of the plant*Phytophagous caterpillars	Artichaut Artichoke	0.075
Product B	Asperge*Trt Part.Aer.*Chenilles phytophages Asparagus*treatment of aerial part of the plant*Phytophagous caterpillars	Asperge Asparagus	0.125
Product B	Betterave industrielle et fourragère*Trt Part.Aer.*Coléoptères phytophages Industrial and fodder beetroot*treatment of aerial part of the plant*Phytophagous beetle	Betterave Beetroot	0.05
Product B	Traitements généraux*Trt Sol*Nématodes General treatment*soil treatment*nematode	Tous All	0.06

31The last line in table 10 shows that besides the application rates specific to each crop, the product can be applied at a generic rate when aiming nematodes. We take this information into account by computing a second rate of reference for this product as shown in table 11.

Table 11: Calculation of the rates of reference by averaging with the one for “general treatments”

Product	Region	Rate of reference	Rage general treatment	Rate of refences No 2
Product B	Region 1	0.11	0.06	(0.11 + 0.06) / 2 = 0.085
	Region 2	0.1	0.06	(0.1 + 0.06) / 2 = 0.08

33Table 12 presents the distribution of the coefficients of variation (the ratio between the standard deviation and the mean value) for the application rates of refence among regions and per AMM and land use. We focus specifically on the coefficients greater than 0.1. Indeed, among the 14 873 coefficients computed for the different combination of AMM and land use, only 388 (or 2.6 %) are greater than 0.1. The most important variations between regions are observed for the crop groups composed of very diverse plants (e.g., VERGERS, LEGU[mes]). The majority of the rates do not differ much between regions.

Table 12: Distribution of the coefficients of variation for the rates in 2020 per AMM and land use between regions, focus on coefficients greater than 0.1

Land use	Min	Q1	Median	Q3	Max	Count
VERGERS	0.101	0.120	0.201	0.351	2.021	121
LEGU	0.102	0.123	0.133	0.228	1.291	61
INDUS	0.104	0.104	0.137	0.586	1.087	25
LFOURR	0.113	0.113	0.113	0.173	0.347	19
FOURR	0.107	0.150	0.207	0.279	0.363	18
VIGNES	0.104	0.132	0.146	0.309	0.338	15
MAIS	0.107	0.122	0.198	0.282	0.363	14
OLIVE	0.108	0.108	0.139	0.141	0.245	13
CEREALES	0.124	0.150	0.305	0.343	0.548	12
BETT	0.124	0.141	0.199	0.476	1.001	10
BLE	0.145	0.198	0.347	0.358	0.636	10
LGRAIN	0.136	0.143	0.172	0.337	0.337	10
ORGE	0.145	0.198	0.347	0.358	0.636	10
RIZ	0.116	0.195	0.234	0.239	1.902	10
COQUE	0.109	0.120	0.124	0.162	0.321	8
PDT	0.124	0.124	0.124	0.152	0.201	5
FIBRE	0.151	0.161	0.164	0.164	0.164	4
OLEA	0.146	0.146	0.146	0.149	0.159	4
POIS	0.136	0.136	0.136	0.142	0.160	4
COLZA	0.146	0.146	0.146	0.146	0.146	3
FEV	0.136	0.136	0.136	0.136	0.136	3
PROTEA	0.153	0.153	0.153	0.153	0.153	3
SOJA	0.146	0.146	0.146	0.146	0.146	3
TOUR	0.146	0.146	0.146	0.146	0.146	3

3.3. Calculation of the distributional weights

34Once the application rate of reference is defined, expressed in kg/ha or in L/ha depending on the packaging of the PPP, we can calculate the distributional weight of the product for the ZIP code. Thus, we first multiply the application rate by the crop group’s area in the ZIP code, let us call this quantity Q1. We then sum these quantities (expressed in kg or L) and obtain the total quantity of a product potentially used in the ZIP code, QT. The distributional weight per ZIP code, per product and per crop group is then the ratio Q1/(QT*area). This calculation produces a result per hectare.

35Let us consider the example of a given product that can be used on three different land use classes, vegetables (LEGU), pulses (LGRAIN), and potatoes (PDT). The calculation of the distributional weights is done in the following manner.

36As explained before, first we have to estimate the quantity of the product potentially used in the ZIP code (table 13).

Table 13: Calculation of the potentially used quantities per product and ZIP code totals

Crop group	Rate of reference (kg/ha or l/ha)	Area (ha)	Potential quantity (kg or l)
LEGU	65	20	1 300
LGRAIN	80	1	80
PDT	42	3	126
		TOTAL	1 506

37Then, we calculate the weights by dividing each potential quantity by the total weighted by the areas (table 14).

Table 14: Calculation of the distributional weights

Crop group	Distributional weight (per ha)
LEGU	1 300 / (1 506 * 20) = 0.043
LGRAIN	80 / (1 506 * 1) = 0.053
PDT	126/ (1 506 * 3) = 0.029

3.4. Spatialisation of the results

38In order to spatialise the BNV-D sales, we multiply them, for each ZIP code, by the distributional weight and by the area of the associated crop group. This method assumes that all parcels from the same ZIP code and sharing the same land use receive the same quantities of products (per ha).

39Let us resume our example by assuming that at the ZIP code level the sales of the product are of 500 kg (table 15).

Table 15: Distribution of the pesticides sold among the crop groups of the ZIP code

Quantity sold (kg or l)	Crop group	Area (ha)	Distributional weight (per ha)	Spatialised quantity (kg or l)
500	LEGU	20	0.043	500 * 20 * 0.043 = 430
	LGRAIN	1	0.053	500 * 1 * 0.053 = 26.5
	PDT	3	0.028	500 * 3 * 0.029 = 43.5

40We recover the quantity sold but distributed among the potentially treated areas of the ZIP code.

3.5. Aggregating at the scale of French municipalities

41Once we have the distributional weights per hectare, we can use them to derive the quantities of BNV-D sales per municipality by:

1. aggregating the areas per municipality, ZIP code, region, and land use;

2. joining to the information the distributional weights (through the ZIP code, the region and the land use);

3. joining the BNV-D data (through the AMM, the unit, and the ZlP code, corrected if necessary, as explained before).

42The quantities (Qty) for the different products per municipality and per land use are then obtained by multiplying the distributional weights and the BNV-D quantities following the formula in equation (1) (the region is not cited among the indices but is taken into account).

Table 16: Calculation of the spatialised quantities per municipality

ZIP code	Region	Municipality	Land use	AMM	Unit	Area (ha)	Weight	Qty product	Qty spatialised
27530	28	27193	BLE	2100138	L/ha	128.09	0.002404	20	6.16
27530	28	27230	BLE	2100138	L/ha	139.60	0.002404	20	6.71
27530	28	27193	ORGE	2100138	L/ha	85.81	0.002748	20	4.72
27530	28	27230	ORGE	2100138	L/ha	43.91	0.002748	20	2.41
								Total	20.00

43This is presented in table 16. For instance, the ZIP code 27530 and the AMM 2100138 are associated with two municipalities (27193 and 27230) and two potentially treated land uses: BLE (wheat) and ORGE (barley). The distributional weight is unique per ZIP code and land use translating the heterogeneity of application rates for the different crops.

44These calculations can be performed for whatever geographical unit once it has been intersected with the land use layer and that the identifier of the targeted unit is attributed. For an even more precise estimation, the parcels can be split between targeted geographical units. Among the geographical divisions of interest and different from administrative units, we can cite those related to nonpoint source pollution of water bodies. We should stress that our current methodology does not account for true farming practices and even less for their territorial specificities. Such information can be integrated in the procedure by adjusting the crops treated and/or the application rates. There are currently developments in progress allowing the insertion of local experts’ data in the BNV-D spatialisation process (Descout et al., 2023) in line with the work of Martin et al. (2023).

4. Results of the spatialisation

45All BNV-D sales have not been spatialised for the reasons presented in table 17. The most important reason is the loss of the ZIP code information, replaced by ‘00000’ in the source data. This is default ZIP code attributed if the distributors omit it in their sales declarations. In terms of quantities, these cases represent between 3 and 4% of the total QSA. The year 2020 is exceptional here with only 1%. Furthermore, we do not spatialise the products used only for storage (between 0.1 and 0.3% of the QSA). Another problematic case is the one of ZIP codes that we do not associate with any of the land use layer’s plots and that we could not redistributed following the rules detailed before.

46In some cases, the product is absent from E-Phy or the units in the BNV-D do not match those for application rates/distributional weights (E-Phy). Moreover, for some combination of products and ZIP codes, there is no approved usage given the land use. Finally, the presented method and the data available allows us to spatialise between 94 and 98% of the QSA depending on the years (the total of the lines ‘ZIP codes distributed” and “Spatialised” in table 17) with the best score for 2020 mainly because of the low proportion of sales with ZIP code ‘00000’.

Table 17: Results of the spatialisation in terms of quantity of active substance (kg)

	2015		2016		2017		2018		2019		2020
Status	QSA (in kg)	in %	QSA (in kg)	in %	QSA (in kg)	in %	QSA (in kg)	in %	QSA (in kg)	in %	QSA (in kg)	in %
QSA not available	-	0.00	-	0.00	-	0.00	-	0.00	-	0.00	-	0.00
Units ephy and BNV-D incompatible for the ZIP code	2 437	0.00	1 331	0.00	40 637	0.06	76 544	0.09	2 161	0.00	1 458	0.00
No possibility to redistribute the ZIP code	25 105	0.04	22 775	0.03	29 838	0.04	27 690	0.03	30 970	0.06	24 946	0.04
AMM for storage	119 115	0.18	97 635	0.14	92 787	0.14	128 614	0.16	118 046	0.22	187 219	0.30
ZIP code unknown in the land use layer	216 879	0.33	71 072	0.10	67 342	0.10	108 628	0.13	137 344	0.26	70 379	0.11
No receiving land use	367 898	0.56	379 220	0.54	278 486	0.41	338 154	0.41	233 590	0.44	569 515	0.90
AMM not in ephy	439 249	0.67	307 119	0.44	286 909	0.42	40 431	0.05	32 911	0.06	95 283	0.15
ZIP code 00000	2 434 437	3.73	2 348 316	3.34	2 224 419	3.24	3 194 613	3.87	1 662 826	3.13	460 969	0.73
Spatialised	61 665 699	94.48	67 043 508	95.41	65 644 594	95.60	78 611 685	95.26	50 968 344	95.83	61 580 821	97.76
— including ZIP codes redistributed	242 845	0.37	239 946	0.34	243 678	0.35	154 765	0.19	108 059	0.20	161 731	0.26
Total QSA	65 270 820	100.0	70 270 993	100.00	68 665 151	100.00	82 526 464	100.00	53 186 226	100.00	62 990 592	100.00

47Given all the data related restrictions, map 1 is a representation of the spatialised QSA per ha at the French municipalities level (all product families combined, all receiving areas) in 2020.

48The classes presented are the percentiles of QSA per ha for the municipalities, with highlights of the municipalities with the greatest quantities spatialised divided by their UAA (utilised agricultural area). We can observe that, on this map, the zones that stand out are associated with permanent crops (vineyards and orchards). Two supplementary information are added in the legend: the threshold QSA/ha of the classes as well as the percentage of the UAA that is in the class. For instance, the penultimate class (between the 95th and the 99th percentile), we can read that the QSA per hectare is somewhere in between 10 and 21 kg/ha, and that 3.9% of the UAA is inside this class.

5. Data produced and reproducibility

49The outputs of the spatialisation are the following:

• Tables “com_insee_qsa” and “com_insee_amm” contain the data per substance (respectively per product), per municipality, per year. The fields are the name of the substance, its CAS number, its classification, with the addition of a mention on the classification if any, the Sandre code, the function, and the QSA in kg (respectively the AMM number, the packaging, the possibility to use or not use the product by amateur gardeners (EAJ) and the quantity of the product).

• Table “com_insee_famille_simple_qsa”: this table aggregates per function the substances into 4 simplified classes (fungicide, herbicide, insecticide, and other) at the municipality level. It allows simpler manipulation of the data with roughly 30 times lower number of lines comparing to the table “com_insee_qsa” (the latter has more than 4 million lines in 2020, while its aggregation per simplified function has 135 000) at the expense of the per substance detail.

• Table “com_insee_occsol”: contains the area per crop group, per municipality, per year. This table is the result of the combination of the land use layers described in Section 2.2. The land use groups are organised in 34 classes, the different grasslands set asides and fallow land, as well as railroads are considered untreated and their area can be ignored. Pesticides are distributed on the agricultural areas and the ZNAs. This table can be used to identify zones of production depending on the crops of interest (select viticultural regions or those in arable crop, for instance).

• Table “bnvd_statut”: unlike previous tables, this one is at the ZIP code level. It details whether the PPP purchase (combination of ZIP code and AMM) has been spatialised or not, and if not the reason for the failure. This table provides a full inventory of the purchases of the year. It allows to check if a particular product presents limitations as to its spatialisation. For instance, in 2020, the AMM No 2000125 (a Bordeaux mixture) has been spatialised over 1300 ZIP codes but not in the case of 47 other ones because of the lack of a receiving land use or unknown ZIP code.

50The figure 4 presents the tables made available and their fields as a scheme.

51More details about the tables are provided in the Supplemental Materials, with namely the description of their different fields and the associated data sources. The tables are in two file formats “.csv” and “.parquet”. While the “.csv” is well known and readable with any text editing application, the format “.parquet” allows: i) faster loading into R, Python, DuckDB, etc.; ii) the fields type is conserved (important when it comes to municipalities codes); and iii) remote data querying via the persistent URLs of files allowing to better target the download of information. This functionality in terms of API (Application programming interface) can facilitate the development of applications using the data.

52All of the data treatment has been done in R (data loading, orchestrations and some processing such the calculation of the application rates) or in Python (data loading and orchestration) with the creation of a PostgreSQL database (version 12.9 and PostGIS extension, version 3.1) allowing us geometrical processing such as the intersection of the land use layer and other zonings. In terms of packages used, for R these are “tidyvers”, “RPostgreSQL”, “here”, “readxl” as well as certain functionalities of R Studio. For Python the packages are “configparser”, “datetime”, “psycopg2”. Given the size of the tables, the choice to use PostgreSQL was self-evident. For more ergonomics and processing automatisation, the compilation and execution of the SQL queries is done via R or Python depending on the time necessary with slower queries (the construction of the land use layer) executed directly on the remote server of the database. These scripts are commented and available on the Git repository; sensitive information in terms of server connections or paths is erased. The link to the project repository is: https://forge.inrae.fr/​alungarska/​obnvds-publique. Some tables necessary for the processing are also available, they allow, among other things: i) the adjustments in the E-Phy database (through the script src/e-phy_v6.R); ii) the attribution of the ZIP codes (see Section 3.1.1).

53The data processing is carried out on a Debian server with 16 processors (Intel® Xeon® Gold 6230R CPU @ 2.10GHz; PostgreSQL is using in parallel up to 5 cores depending on the parameters) and 16 GB of RAM. Table 20 presents the approximate time and hard drive size necessary for the processing of one year’s data. In order to accelerate junctions between tables, indexes, including spatial ones, where put in place.

6. Results analysis

54An indicator of interest when studying pesticides’ use is the QSA per hectare of agricultural land. This indicator takes into account the variation of the BNV-D data among ZIP codes. In figure 5, we show the distribution of the coefficients of variation of the QSA/ha. These coefficients are calculated at the municipality level and summarised per region and active substance. The majority of the coefficients are inferior to 5, this illustrates a low variability of the QSA/ha across regions. A greater variability is observed for wheat in the Paris region (Île-de-France). This is probably the result of the disparity between the ZIP code of farms’ headquarters which can be located in a territory different than the parcels. This way, municipalities with few agricultural plots can find themselves attributed important per hectare quantities.

55At the regions level, figure 6 and figure 7 present the QSA/ha for 8 crop groups of the spatialisation (among the 24 used) for the regions Nouvelle Aquitaine (major city Bordeaux) and Hauts-de-France (major city Lille) respectively. We find similar orders of magnitude for the different crops in terms of values and ranking except for orchards. These latter cover productions differing between the regions. By way of comparison, the ranking is quite similar to the one of the treatment frequency index (IFT) derived from the surveys on “agricultural practices” despite the differences in measurement units between QSA and what is referred to as “treatment” (Agreste, 2023).

56Map 2 and map 3 present the QSA/ha for the fungicides and the glyphosate spatialised at the municipality level. The numerator is the QSA grouped per function (or filtered at the substance). The denominator is the sum of the receiving areas (crops and ZNA). Patterns clearly emerge from the two maps. In most cases the contrasting zones are those where permanent crops are present (vineyards and orchards) such as the Mediterranean rim, the Gironde (Bordeaux), the Champagne (Reims) as well as the Rhône corridor. The classes on the maps are obtained thanks to the “Jenks” method set for 5 classes. The classes with deeper colours show higher levels of QSA/ha. Grey colour stands for municipalities with no pesticide sales associated. For instance, on map 2, the municipalities in the Southeast part of the Occitania region (Montpellier) form a homogenous bloc coloured in red-orange. This colour corresponds to the second highest class, with a spatialised QSA/ha in between 5 and 25 kg/ha.

57Map 4 presents the QSA/ha of fungicides in the Gers department (Auch) in 2020 as modelled in the BNVDs (map in the upper part of the figure at the municipality level) and as reported by the BNV-D (sales that are not spatialised are excluded from the ZIP codes map in the lower part of the figure). By comparing the two maps, we can note that the data at the municipality level makes apparent some local variations nuancing the gradients. Indeed, the heterogenous land use between the municipalities from the same ZIP code drives the spatialisation of more or less of the products. This makes municipalities change their class of values comparing to the one of their ZIP code. This is also shown in map 5 which presents the spatialisation of glyphosate in the Rhône department (Lyon) in 2020.

58Finally, map 6 focuses on the spatialisation at the municipality level for a drinking water catchment area (AAC). The boundaries of the ZIP codes, the municipalities, and the AAC are drawn on this map (respectively, in red, grey, and blue). The value of the QSA/ha spatialise at the municipality level is shown in the centre of the polygons in order to show the heterogeneity in the distribution within the ZIP code. We keep only the municipalities (or ZIP codes) intersecting the AAC of interest. In the case of the AACs, it is better to build on the results of the spatialisation at the parcel level for a more precise estimation of the QSA. This is even the primary interest of the proposed method so as to overcome the geographical limits of administrative units.

7. Limits and perspectives

59The spatialisation method allows us to associated the PPP purchases data to a land use and to geographical units of finer scale than the ZIP code currently available. As all method for downscaling, the results are subject to biases such as those discussed in the introduction (Openshaw, Taylor, 1979; Eagleson et al., 2002). Furthermore, our method involves assumptions on the PPP’s usage and this brings about some precautions when interpreting the results. Thus, numerous limits of the method are yet to be solved and should be taken into account.

7.1. Limits related to the source data

• [BNV-D] Since these are sales data, it is possible that certain quantities bought are not used the same year. We do not know the exact date of purchase and these stocking biases are possible. Moreover, the BNV-D covers the sales that occurred current the civil year and not the agricultural campaign.

• [BNV-D] Some sales are associated with the ZIP code ‘00000’ which hinders spatialisation. For other ZIP codes, less than five farms are present and thus the sales data is not rendered public. There is also the case of ZIP codes absent from the land use layer, for some the sales are redistributed as presented in Section 3.1.1.

• [BNV-D and land use layer] The ZIP code of a given purchase is the one of the farm’s headquarters. It is possible that farm’s parcels are located within another ZIP code. As a consequence, the spatialisation is done on the basis of the ZIP code’s land use where it would be preferable to do it on the basis of the plots of the farms associated with the ZIP code (figure 8). The data allowing to match farms and their respective parcels do exist but are considered too sensitive to be freely diffused. Nevertheless, it is estimated that about 20% of the LPIS parcels are associated to a ZIP code different from their farm’s headquarters’ one. Accessing these data could improve the quality of the spatialisation as demonstrated in Lungarska et al. (2023).

• [E-Phy] Some products do not dispose of an application rate in the E-Phy catalogue, we attribute them a homogenous one equal to 1. The list of these products is provided in the appendices (table 18). Moreover, the doses from E-Phy are the legal ones. Farmers still can apply different rates in practice. These rates do not inform us on the maximal number of treatments, or at least, and exhaustively.

• [Land use] The LPIS not being exhaustive (see Section 2.2.2), its complementary layer, the RPG complété, is the result of statistical modelling and/or data not frequently updated. This results in some uncertainties as to the land use classes.

7.2. Limits related to the spatialisation method

• The calculation of the coefficients is done on the basis of median application rates. In the absence of exhaustive information on targeted bioagressors and on the actual PPP’ usage by farmers, it is impossible to have the precise application rate. Furthermore, the method does not take into account the number of treatments done and it can vary significantly among crops. This results in an important bias in the zones with arable crops and vineyards (particularly true in the case of fungicides such as sulphur).

• The spatialisation method does not account whether the parcels are exploited with organic farming practices or not. Organic farming parcels receive the same products as those in conventional agriculture even if they should only be treated with a subset of approved products.

• The spatialisation does not account for products’ run-off and drift. Although, these phenomena are identified in the literature (Bailey et al., 1974; Damalas 2015), they are out of the scope of this modelling exercise. In order to do this, we would need a precise treatment calendar for each parcel, combined with further topographic, edaphic and climatic data at a fine scale.

8. Conclusion

60The data presented in this paper are the result of a spatialisation method being applied to the phytopharmaceutical products purchases data provided by the French National Bank of PPP sales by the authorised distributors (BNV-D). This method allows us to distribute the quantities of products or active substances among numerous agricultural and non-agricultural land use classes. Thanks to the definition of an application rate specific to each product and each land use class, the quantities of products can be attributed to the parcels. The land use layer used is the most comprehensive possible by combining the data from the LPIS and the RPG complete as well as information on non-agricultural zones potentially receiving treatments. The spatialisation at the plot level facilitates the change of scale towards geographical units other than the ZIP code, the initial level of the BNV-D “Register” database. To the best of our knowledge, because of its comprehensiveness, our spatialisation of the BNV-D seems to be the most accomplished one up until now.

61However, there are numerous limits to our method that should be accounted for when using the data for the purpose of research or public policies’ assessment. Some of these limits are data related while others are due to our methodological choices. Among other, the most important ones are: i) the possible discrepancy between the ZIP code of the headquarters and the physical code of farm’s parcels; ii) the application rates do not distinguish sufficiently the practices between crops; iii) the organic farming is not accounted for. We have identified potential solutions to these 3 issues and we are pursuing developments in future version of the data.

62In order to nurture future empirical research in agronomics, economics or epidemiology, we propose here an aggregation at the municipality level of the results of the spatialisation. This way, we increase the number of annual observations from slightly lower than 6 000 ZIP codes to more than 34 000 municipalities and facilitate the interoperability with other databases.

Supplemental Materials

63The datasets are integrated into an R Shiny application allowing the creations of maps and figures ‘on-the-fly’ as well as the access to all or subsets of the data following different criteria (active substance, function, toxicity). This application, in French, is available here: https://shiny.sk8.inrae.fr/​app_direct/​odr-shiny-obnvds.

Dataset Description

64Common elements:

• separator: “;”

• encoding: UTF-8

• code for missing values: “NA”

• decimal separator: “.”

65Files naming convention: the files are organised into folders depending on their year.

66File tree:

67The same plan applies to the files with ‘.parquet’ extension.

.
├── 2015
│ ├── bnvd_statut_2015.csv
│ ├── com_insee_famille_simple_qsa_2015.csv
│ ├── com_insee_occsol_2015.csv
│ ├── com_insee_amm_2015.csv
│ └── com_insee_qsa_2015.csv
├── 2016
│ ├── bnvd_statut_2016.csv
│ ├── com_insee_famille_simple_qsa_2016.csv
│ ├── com_insee_occsol_2016.csv
│ ├── com_insee_amm_2016.csv
│ └── com_insee_qsa_2016.csv
├── 2017
│ ├── bnvd_statut_2017.csv
│ ├── com_insee_famille_simple_qsa_2017.csv
│ ├── com_insee_occsol_2017.csv
│ ├── com_insee_amm_2017.csv
│ └── com_insee_qsa_2017.csv
├── 2018
│ ├── bnvd_statut_2018.csv
│ ├── com_insee_famille_simple_qsa_2018.csv
│ ├── com_insee_occsol_2018.csv
│ ├── com_insee_amm_2018.csv
│ └── com_insee_qsa_2018.csv
├── 2019
│ ├── bnvd_statut_2019.csv
│ ├── com_insee_famille_simple_qsa_2019.csv
│ ├── com_insee_occsol_2019.csv
│ ├── com_insee_amm_2019.csv
│ └── com_insee_qsa_2019.csv
├── 2020
│ ├── bnvd_statut_2020.csv
│ ├── com_insee_famille_simple_qsa_2020.csv
│ ├── com_insee_occsol_2020.csv
│ ├── com_insee_amm_2020.csv
│ └── com_insee_qsa_2020.csv
└── LISEZMOI.md

Information specific to the data in the files com_insee_occsol.csv/com_insee_occsol.parquet

68File containing the aggregated areas per municipality and per land use group.

69List of variables/heading of the columns:

• com_insee: code of the municipality (INSEE 2018 references), these values are assigned to each parcel depending on the municipality where lays the majority of its area

• occsol: land use groups (agricultural and non-agricultural zones, see the methodology note Ramalanjaona et al., 2020)

• surf_ha: aggregated area (computed) of the parcels of a given land use and municipality, expressed in ha with precision of 4 digits after the decimal point.

Information specific to the files com_insee_amm.csv/com_insee_amm.parquet

70File containing quantities of pesticides products spatialised per municipality.

71List of variables/heading of the columns:

• com_insee: code of the municipality (INSEE 2018 references), these values are assigned to each parcel depending on the municipality where lays the majority of its area.

• amm: marketing authorisation number (Autorisation de mise sur le marché) as provided in the BNV-D.

• conditionnement: “kg/ha” or “L/ha” as provided in the BNV-D (the “/ha” part is added by commodity).

• eaj: Use authorised in gardens (Emploi Autorisé dans les Jardins), “Oui” or “Non”, products allowed to the general public as detailed in the BNV-D.

• quantite_produit: pesticides’ quantity (the unit is provided in the “conditionnement” column) from the BNVDs per municipality.

Information specific to the files com_insee_qsa.csv/com_insee_qsa.parquet

72File containing the active substances’ quantities as spatialised per municipality.

73List of variables/heading of the columns:

• com_insee: code of the municipality (INSEE 2018 references), these values are assigned to each parcel depending on the municipality where lays the majority of its area.

• substance: name of the active substance as presented in the BNV-D.

• cas: Chemical Abstracts Service number of the substance as presented in the BNV-D.

• classification: following the BNV-D with regards to the annual decrees defining the list of substances subject to the diffuse pollution tax; for the 2015-2018 period, the modalities are “N Organique”, “N minéral”, “T, T+, CMR”, “Autre”; for the 2019-2020 period, the modalities are “Santé A”, “Env B”, “Env A”, “CMR”, “Autre”.

• classification_mention: provided by the BNV-D following the decrees fixing the list of active substances subject to the diffuse pollution tax; “Substitution” or “Exclusion”.

• code_sandre_substance: number in the National administrative service on data and reference about water (derived from the BNV-D).

• fonction_substance: function defined in the BNV-D with the following modalities: “Fongicide”, “Herbicide”, “Herbicide – Antimousse”, “Insecticide”, “Insecticide – Acaricide”, “Insecticide – Médiateur chimique”, “Molluscicide”, “Nématicide”, “Rodenticide – Taupicide”, “Régulateur de croissance”, “Bactéricide – Virucide”, “Autre usage – Répulsif”, “Autre usage – Adjuvant”, “Autre usage – Coformulant/Phytoprotecteur/Synergiste”, “SA non phyto – Divers”, “SA non phyto – Désinfectant – Biocide”.

• qsa: quantity of active substance (in kg) resulting from the BNVDs for the municipality.

Information specific to the files com_insee_famille_simple_qsa.csv/com_insee_famille_simple_qsa.parquet

74File containing the active substances’ quantities as spatialised per municipality aggregated following a simplified function.

75List of variables/heading of the columns:

• com_insee: code of the municipality (INSEE 2018 references), these values are assigned to each parcel depending on the municipality where lays the majority of its area.

• famille_substance: simplified function with the following modalities “Fongicide”, “Herbicide”, “Insecticide” et “Autres”.

• qsa: quantity of active substance (in kg) resulting from the BNVDs for the municipality.

• nb_substances: number of active substances aggregated in the same simplified function class in the municipality.

Information specific to the files bnvd_statut.csv/bnvd_statut.parquet

76List of variables/heading of the columns:

• code_postal_acheteur: buyer’s ZIP code as declared in the BNV-D.

• amm: marketing authorisation number (Autorisation de mise sur le marché) as provided in the BNV-D.

• quantite_substance: quantity of active substance contained in the product.

• statut: result of the spatialisation with the following modalities:

  • « Spatialisée »: the sale has been successfully spatialised;

  • « Pas d’occsol récepteur »: the sale has not been spatialised; the ZIP code is known to the land use layer but there are no parcels potentially receiving treatments of the product (with a land use class authorised for the product);

  • « CP dispatché »: the sale has been spatialised; the ZIP code of the sale is not among those in the land use layer but the sale has been attributed to one or few other ZIP codes disposing with parcels allowed to receive treatments;

  • « Pas de possibilité de dispatcher le CP »: the sale is not spatialised; impossible to attribute the sale to another ZIP code disposing with parcels allowed to receive treatments;

  • « CP inconnu pour l’occsol »: the sale has not been spatialised; its ZIP code is unknown to the other data sources (SIRENE, ZIP codes database);

  • « CP à 00000 »: the sale has not been spatialised; the ZIP code of the buyer was not attributed by the distributors;

  • « AMM de stockage »: the sale has not been spatialised; the product is only authorised for the stocking (silos, etc.) and thus is not spatialised on the agricultural and non-agricultural parcels;

  • « AMM pas dans ephy »: the sale has not been spatialised; the information on crops allowed to receive treatments is not provided in the E-phy database;

  • « Unités ephy et bnvd incompatibles pour le CP »: the sale has not been spatialised; there is a discrepancy between the packaging in the BNV-D and in the E-phy databases;

  • « QSA non renseignée »: the sale has not been spatialised; the product does not dispose of a measure of its active substance (if any).

Localisation of the data and files’ description

77Data in “csv” format are available here: https://doi.org/​10.57745/​EJQWUZ.

78Data in “parquet” format are available here: https://doi.org/​10.57745/​FCQKYQ.

Acknowledgments

79The authors would like to express their gratitude to the three anonymous reviewers of the paper for their valuable advices. This paper is the result of a long-term work that mobilised numerous scientists and experts. We would like to thank the piloting committee: Remy Ballot, Corentin Barbu, Nowlenn Bougon, Antoine Camus, Marco Carozzi, Florence Fernandez, Laurence Guichard et Philippe Martin. We would also like to mention the present and former members of the Observatoire du développement rural (ODR) team: Eric Cahuzac, Pierre Cantelaube, Marie Carles, Pascal Filippi, Eva Groshens, Benjamin Lardot, Lovasoa Ramalanjaona, Claire Séard et Camille Truche.

Hypertext links

80BNV-D Traçabilité: https://ventes-produits-phytopharmaceutiques.eaufrance.fr/​

81E-Phy on Data.gouv: https://www.data.gouv.fr/​fr/​datasets/​donnees-ouvertes-du-catalogue-e-phy-des-produits-phytopharmaceutiques-matieres-fertilisantes-et-supports-de-culture-adjuvants-produits-mixtes-et-melanges/​#/​resources

82RPG through IGN: https://geoservices.ign.fr/​rpg

83RPG complété: https://entrepot.recherche.data.gouv.fr/​dataverse/​odr_sysagri?q=&fq2=keywordValue_ss%3A%22RPG%2C+RPG+compl%C3%A9t%C3%A9%22&fq0=subtreePaths%3A%22%2F123960%2F9%2F123452%2F123672%22&fq1=dvObjectType%3A%28dataverses+OR+datasets%29&types=dataverses%3Adatasets&sort=dateSort&order=

84Annual agricultural statistics (Statistique agricole annuelle): https://agreste.agriculture.gouv.fr/​agreste-web/​disaron/​SAA-SeriesLongues/​detail/​

Data samples

BNV-D “Registre” (table “achats” of substances per ZIP code)

RPG

E-Phy

Table 18: Products and active substances’ quantities (in kg) lacking application rates allowing spatialisation. We assign them a unique dose of 1 for the crops subject to treatments.

AMM	QSA 2015	QSA 2016	QSA 2017	QSA 2018	QSA 2019	QSA 2020
2000097	1 299	988	40	47	68	566
2000124	0	0	0	0	0	0
2000136	273	104	0	0	0	0
2000228	492	32	2	1	7	5
2000364	2	6	14	14	6	0
2000378	1 230	2	0	0	0	0
2000512	0	0	0	0	0	0
2000536	41 256	5 794	3 544	3 452	3 904	3 702
2010121	0	0	0	0	0	0
2010142	1	0	1	1	1	45
2010143	3 115	2 905	1 454	842	111	12
2010261	3 281	5 258	6 635	8 860	5 544	9 437
2010273	71	0	2	9	1	4
2010275	1	9	5	6	0	0
2010378	5	4	7	2	0	0
2010430	0	0	0	0	0	0
2010439	84	68	243	45	0	0
2010495	82 299	84 038	52 928	13 694	3 786	2 993
2010585	1 893	1 960	1 919	969	10	0
2010604	15	0	0	0	0	0
2020002	0	0	0	0	0	0
2020234	13 327	14 462	14 313	6 454	961	873
2020303	8	8	8	1	1	1
2020305	3	0	0	0	0	0
2020365	0	0	0	0	0	0
2030056	6 415	193	4	4	0	11
2030069	5	3	1	0	0	0
2030070	2 339	2 395	1 202	212	87	82
2030072	21 948	20 956	15 568	3 529	1 074	701
2030073	35 814	38 657	31 589	10 410	6 687	2 560
2030110	6 540	10 705	16 147	7 017	2 969	1 182
2030169	80	4	0	0	0	0
2030173	133	87	0	16	3	0
2030226	998	1 252	1 395	1 532	1 045	71
2030406	0	0	0	0	0	0
2030429	2 490	2 641	940	118	26	11
2030430	38 863	44 291	33 907	12 287	4 040	2 247
2040063	209	44	6	12	0	0
2040076	5	0	0	0	0	0
2040205	121	40	0	1	0	0
2040321	1	0	0	0	0	0
2040322	0	0	0	0	0	0
2040323	0	0	0	0	0	0
2040336	40 514	15 446	1 350	115	0	0
2050078	4 308	4 591	4 917	6 177	4 818	5 976
2050136	3 904	2 943	3 508	3 961	3 169	4 584
2050142	0	5	0	5	0	1
2060111	278	68	364	390	398	366
2060114	5 575	643	654	25	21	0
2060175	37 002	31 709	29 107	19 303	14 449	16 759
2080025	10 758	11 568	11 879	14 178	17 393	16 515
2080049	10 814	13 392	20 681	20 475	1	0
2080061	6 921	11 192	10 764	15 980	952	0
2080091	201	62	147	4	1	1
2090030	0	0	0	0	0	0
2090051	20 556	19 184	22 171	24 602	12 860	4 221
2090062	210	48	15	0	12	0
2090109	491	1 171	2 141	835	12	35
2090146	2	3	5	6	10	17
2090160	495 941	1 679 818	3 169 104	3 949 044	2 624 836	1 830 687
2090163	73 495	86 481	110 678	139 671	130 492	104 937
2090169	59	28	20	9	14	17
2090200	30	53	62	39	31	43
2100035	196	122	40	68	0	0
2100036	594	1 087	971	249	102	76
2100124	1 438	25	2	0	0	0
2100154	59 210	107 120	1 678	183	186	222
2100175	429	801	850	791	719	1 133
2100208	2 509	2 326	3 236	2 294	1 626	8
2100241	78	162	116	304	48	56
2100242	0	0	0	0	0	0
2110009	287	635	755	619	748	716
2110017	34	950	26	91	1	41
2110074	2	1	1	0	0	0
2110076	784	835	878	970	919	2 222
2110083	70	14	8	1	1	1
2110084	19	21	32	42	40	1
2110121	67 256	58 506	71 390	68 684	471	112
2110179	2 275	2 452	2 218	2 062	1 846	1 970
2120030	1 840	343	873	480	951	1 114
2120058	22 882	22 300	20 411	26 438	10 470	27 146
2120124	3	0	0	0	0	0
2120130	1 776	2 029	1 422	1 449	2 096	2 949
2130077	11 324	14 434	14 108	21 466	5 537	120
2130108	3 214	491	728	688	620	690
2130109	352	203	64	7	129	303
2130112	210	218	242	69	128	595
2130114	31	27	135	148	0	94
2130115	2 443	3 245	2 139	4 350	3 693	3 530
2130267	0	0	0	0	0	0
2130272	145	186	184	191	95	47
2140041	14	21	28	105	78	11
2140043	942	6 581	17 147	32 919	39 412	5 775
2140044	829	941	2 661	10 926	2 091	611
2140086	2 278	5 268	7 656	11 655	10 657	7 602
2140093	8 262	131 108	300 357	357 993	326 626	593 259
2140143	0	0	0	0	60	13
2140146	0	0	0	0	0	849
2140215	9 157	8 839	15 351	15 835	12 592	15 351
2150086	27	334	338	455	378	432
2150105	0	0	0	0	3 571	5 352
5900316	0	0	0	0	0	0
5900368	4 791	7 447	7 830	4 625	2 096	2 197
6000328	0	0	0	0	0	0
6600397	43	30	14	69	45	29
7000169	1	1	0	0	0	0
7200137	0	0	0	0	0	0
7400610	0	0	24	21	0	0
7400634	30	37	154	16	6	10
7400636	19	0	6	0	0	0
7400648	1	0	0	0	0	0
7400649	10	5	0	0	0	0
7400758	0	0	0	0	0	0
7500589	0	0	0	0	0	0
7700385	149 843	213 690	104 040	66 300	0	0
7700788	0	0	0	0	0	0
7800552	29	0	0	0	0	0
8100268	32	0	0	0	0	0
8100315	0	0	0	0	0	0
8100345	9 246	10 455	11 170	14 527	21 185	19 926
8100462	2 531	2 232	1 625	1 371	983	1 303
8400215	51 802	13 984	6 163	148	0	0
8400276	4	0	0	0	0	0
8400364	0	0	106	62	5	3
8400539	0	0	0	0	0	0
8400549	163	38	4	0	0	0
8500183	63	45	14	1	0	0
8500347	19	18	3	4	0	0
8500348	1	0	0	1	0	0
8500351	36	8	3	1	1	1
8500352	0	0	0	0	0	0
8600027	12 050	12 437	12 354	16 439	6 216	58
8600110	2	0	0	0	0	0
8600349	1 256	858	773	616	607	417
8600395	1 493	931	820	881	827	890
8600659	0	0	0	0	0	1
8700111	1 404	1 255	1 006	1 046	1 031	1 261
8700155	1	1	1	1	1	0
8700304	1 457	737	386	220	132	74
8700305	127	186	217	167	46	1
8700542	51 798	77 045	599	34	0	0
8800001	55	65	22	82	20	12
8800108	0	0	0	0	0	0
8800335	0	0	0	0	0	0
8800423	0	0	0	0	0	0
8800499	0	0	0	0	0	0
8800519	4 696	4 506	4 039	3 840	4 000	3 702
8800550	16	0	0	0	0	0
8800941	60 009	43 056	47 934	30 285	15 247	2 899
8900124	2	142	127	116	0	0
8900132	0	0	0	0	0	0
8900685	953	805	800	828	840	4 367
8900692	216	367	15	0	1	0
9000001	0	0	0	0	0	0
9000319	203	141	281	34	1	0
9000359	93	34	24	12	623	520
9000360	198	143	78	71	62	61
9000361	2 896	2 344	2 323	1 881	1 844	1 347
9000362	2 912	1 594	2 064	861	1 709	70
9000834	17 429	18 821	19 081	17 085	42	4
9100650	12 054	5 387	5 138	1 373	217	160
9200058	0	2	1	1	1	0
9200062	0	1	1	1	0	0
9200145	0	0	0	0	0	0
9200370	169	242	0	0	0	0
9200379	1 125	1 284	1 588	2 203	1 673	3 246
9200380	2 310	2 870	3 375	5 178	2 811	8 745
9200392	9 032	6 936	6 941	186	42	121
9200520	599	0	0	0	0	0
9200532	19 223	20 894	19 792	19 644	16 308	8 538
9300065	11 711	20 599	23 753	5 623	48 557	28 721
9300214	0	0	0	0	0	0
9300379	3	1	0	0	0	0
9400008	6 901	6 189	7 660	5 871	4 402	4 863
9400235	3 418	1 929	1 099	453	281	285
9400246	0	0	0	0	0	0
9400458	1 069	215	114	4	6	5
9400459	324	286	236	273	272	0
9400460	8 827	11 899	11 800	3 709	4 415	14 726
9400462	105	125	105	88	75	61
9400495	19 665	16 245	1 549	787	21	2 312
9500088	38	30	13	3	0	0
9500122	3	0	0	1	0	0
9500135	28 955	32 110	23 763	13 808	2 540	29
9500296	52	33	40	114	27	27
9500467	85	64	100	40	110	80
9500648	220	0	0	0	0	0
9600510	2 469	45	15	11	5	179
9700192	30	66	0	33	0	53
9800125	2	0	0	0	0	0
9800129	2 176	0	0	0	0	0
9800201	19 068	33 718	15 917	6 824	5 043	3 085
9800207	219	156	155	155	139	135
9800255	1	2	0	0	0	0
9800350	35	188	100	22	0	2
9800481	35 544	45 289	46 130	32 530	51 959	51 283
9900018	1	0	0	0	0	0
9900019	0	0	0	0	0	0
9900023	0	0	0	0	0	0
9900123	642	690	419	307	431	352
9900276	247	434	6	5	0	0
9900277	3	6	0	0	0	0
9900333	0	0	0	0	0	0
9900350	26 570	23 938	23 156	23 190	0	0
9900352	0	0	0	0	0	0
9900371	7 321	2 484	1 816	1 132	1 082	624
2000041	0	3	0	0	0	2
2000300	0	0	0	0	0	0
2010124	0	13	0	10	0	0
2040120	0	13	0	0	0	8
2040121	0	6	11	10	12	5
2040140	0	6	2	1	0	0
2090141	0	4 406	19 645	13 448	4 883	14
2100105	0	1	0	1	0	0
2130123	0	25	0	0	0	0
2140091	0	0	0	0	0	0
2150068	0	6	8	0	22	34
2150112	0	22	157	7 285	4 700	4 648
2150179	0	74	360	758	1 009	1 032
2160230	0	47	36	37	108	68
2160273	0	265	1 713	2 057	690	213
2160423	0	4	119	536	413	1 200
2160704	0	1 562	124 388	229 363	145 397	101 008
8500353	0	0	0	0	0	0
8500354	0	0	0	0	0	0
8500616	0	5	81	0	0	0
8700127	0	7	0	0	0	0
9100500	0	1	0	0	0	0
9200056	0	1	0	0	0	0
9800097	0	2	0	0	0	0
2030074	0	0	0	0	0	0
2100130	0	0	8	0	0	0
2150847	0	0	11	29	27	26
2160321	0	0	6	7	10	0
2160383	0	0	49	9	23	17
2160944	0	0	1 334	1 759	830	1 217
2160960	0	0	56 322	110 352	84 822	163 886
2161035	0	0	3	61	275	138
2170042	0	0	288	5 184	8 991	91 456
2170230	0	0	2	435	546	1 054
2170434	0	0	6	135	788	1 102
2170752	0	0	119	296	409	1 675
2170902	0	0	495	76 554	46 169	53 422
2171126	0	0	7	160	396	474
2171253	0	0	0	0	0	0
8400175	0	0	30	0	0	0
9400130	0	0	28	0	0	0
9400332	0	0	4	0	0	0
9500444	0	0	0	0	0	0
2040084	0	0	0	0	0	0
2060046	0	0	0	0	0	0
2150175	0	0	0	46	9	16
2160929	0	0	0	739	586	600
2170392	0	0	0	25	27	1
2170711	0	0	0	206	352	435
2170785	0	0	0	1 624	2 020	1 739
2170907	0	0	0	17 468	23 549	43 099
2170960	0	0	0	69	77	80
2180124	0	0	0	72	1 004	11 051
2180703	0	0	0	113	3 051	6 300
2189997	0	0	0	0	0	0
8700460	0	0	0	1	1	0
9900229	0	0	0	30	0	0
2090145	0	0	0	0	0	0
2120105	0	0	0	0	8	9
2170713	0	0	0	0	594	183
2170871	0	0	0	0	17 741	338
2180226	0	0	0	0	220	144
2180257	0	0	0	0	14	7
2180436	0	0	0	0	52	127
2180549	0	0	0	0	12	160
2180766	0	0	0	0	90	77
2190215	0	0	0	0	1 387	1 203
9300094	0	0	0	0	0	0
9500460	0	0	0	0	1	0
2040003	0	0	0	0	0	9
2060107	0	0	0	0	0	6
2100070	0	0	0	0	0	1
2150077	0	0	0	0	0	3
2190311	0	0	0	0	0	2 055
2200078	0	0	0	0	0	0
9200176	0	0	0	0	0	0

Complementarity between LPIS (RPG) and RPG Complété

Table 19: Number of parcels and areas (in 1 000 ha) per land use group, LPIS (RPG) and RPG complété, 2020, Metropolitan France

Land use group	Land use group name	RPG + RPG complété area (in 1000 ha)	RPG			RPG complété
			Number of parcels	Area (in 1000 ha)	Share of the total	Number of parcels	Area (in 1000 ha)	Share of the total
PP	Prairies permanentes	8 879.3	3 053 607	7 785	88%	2 879 205	1 094	12%
BLE	Blé	4 806.9	963 079	4 767	99%	73 533	40	1%
MAIS	Maïs	3 308.7	932 840	3 281	99%	48 866	28	1%
STH	Surfaces toujours en herbe	2 520.1	326 451	2 227	88%	842 051	293	12%
ORGE	Orge	1 974.3	424 040	1 969	100%	8 783	5	0%
PT	Prairies temporaires	1 800.0	703 561	1 463	81%	864 414	337	19%
COLZA	Colza	1 113.3	182 900	1 111	100%	4 266	3	0%
LFOURR	Légumineuses fourrage	812.3	298 761	812	100%	15	0	0%
TOUR	Tournesol	780.9	164 177	776	99%	7 831	5	1%
VIGNES	Vignes	781.9	481 932	587	75%	625 921	195	25%
GEL	Gel/friches	491.5	523 001	469	95%	67 526	23	5%
BETT	Betterave	429.8	58 647	430	100%	59	0	0%
CEREALES	Céréales	292.5	108 980	287	98%	9 579	5	2%
LEGU	Légumes	251.5	111 665	219	87%	126 685	33	13%
PDT	Pommes de terre	211.8	50 766	212	100%			0%
POIS	Pois	201.1	36 255	201	100%			0%
SOJA	Soja	187.0	44 401	186	100%	1 566	1	0%
FIBRE	Cultures pour fibres	158.8	25 015	159	100%	343	0	0%
VERGERS	Vergers	174.1	88 425	122	70%	127 331	52	30%
FOURR	Fourrages	101.0	43 063	89	88%	9 023	12	12%
FEV	Féveroles	75.8	19 375	76	100%			0%
INDUS	Cultures industrielles	73.1	34 532	71	97%	3 230	2	3%
LGRAIN	Légumineuses grain	59.0	15 855	59	100%			0%
PROTEA	Protéagineux	47.8	12 099	47	99%	727	0	1%
COQUE	Fruits à coque	47.9	34 181	46	97%	4 341	2	3%
OLEA	Oléagineux	38.9	7 373	39	100%	289	0	0%
RIZ	Riz	14.5	1 703	15	100%			0%
OLIVE	Oliveraies	17.2	18 294	13	75%	13 670	4	25%
DIVERS	Divers	2.3			0%	3 846	2	100%
SEMENCES	Cultures pour semences	0.1			0%	64	0	100%
TOTAL		29 653.1	8 764 978	27 516	93%	5 723 164	2 137	7%

Table 20: Computing time for script execution and databases’ construction

Treatments	Most important tables	Script	Space	Computing time
Land use layer construction	parcelles	src/01.occsol.py	> 30 GB	2h
Application rates’ computations	amm_cult_dhmed, amm_occsol_dhref, amm_occsol_dhref_2, amm_surf_cp	src/03.Calc_coef.R	> 4 GB	< 1h
Computation of distributional weights	coef_amm_cp	src/03.Calc_coef.R	> 6 GB	< 1h
Spatialisation of the quantities	bnvd_amm, bnvd_amm_cor, bnvd_sa, bnvd_sa_cor, correctif_cp_coef, com_adm_occsol	src/02.BNVd.R src/04.BNVd_corrige_cp.R src/05.Precalcul_com_adm.py	< 500 MB	< 1h
Aggregation depending on the aimed geographical unit	com_adm_occsol_amm, com_adm_occsol_qsa	src/05.Precalcul_com_adm.py	> 4 GB	< 1h

85Map 8 is at the ZIP code level. These are classified in 4 categories, those containing only one municipality (1820 ZIP codes), those containing from 2 to 4 municipalities (1496 ZIP codes), those from 5 to 9 municipalities (1382 ZIP codes) and those with more than 10 municipalities (1350 ZIP codes). For the latter 3 classes, the spatialisation suggests a gain in the precision with all the precautions concerning the underlying assumptions. For each category, we provide the percentage of the UAA concerned. We show that ZIP codes with only one municipality cover only 3.9% of French metropolitan UAA. The ZIP codes presented here do not include special business ZIP codes (CEDEX).

French version of the tables

Table 21: Original version of table 2

Cultures BNVDs	Cultures E-Phy	Cultures SAA
BETT	Betterave industrielle et fourragère, Betterave potagère, Porte graine, Porte graine – Betterave industrielle et fourragère	Betterave industrielle, Betterave potagère
BLE	Blé, Céréales, Céréales à paille	Blé, Triticale
CEREALES	Avoine, Céréales, Céréales à paille, Sarrasin, Seigle	Autres céréales, n.c.a. niv. 1, Avoine, Seigle et méteil
COLZA	Crucifères oléagineuses	Colza grain et navette
COQUE	Châtaignier, Cultures fruitières, Fruits à coque, Noisetier, Noyer	Châtaignes, Fruits à coque, Noisettes, Noix
DIVERS	Arbres et arbustes
FEV	Graines protéagineuses	Féveroles et fèves
FIBRE	Chanvre, Lin, Porte graine, Porte graine – Plantes à fibre	Chanvre papier, Lin textile, Plantes à fibres (y compris semences)
FOURR	Céréales, Choux, Graminées fourragères, Maïs, Porte graine, Porte graine – Graminées fourragères et à gazons, Porte graine – Mais, Porte graine – Maïs	Chou fourrager, Maïs fourrage et ensilage (plante entière), Prairies artificielles et temporaires, Prairies temporaires
GEL	Jachères et cultures intermédiaires, Jachères et cultures intermédiaires	Jachères
INDUS	Canne à sucre, Fines herbes, Fines Herbes, Houblon, Infusions, Infusions (séchées), Porte graine, Porte graine – PPAMC, florales et potagères, Porte graine – PPAMC, Florales et Potagères, PPAMC, PPAM – non alimentaires, Tabac	Canne à sucre, Houblon, Plantes aromatiques, médicinales et à parfum (non compris semences), Tabac
LEGU	Artichaut, Asperge, Bulbes ornementaux, Carotte, Céleri-branche, Céleris, Chicorées – Production de chicons, Chicorées – Production de racines, Choux, Choux à inflorescence, Choux feuillus, Choux pommés, Choux-raves, Concombre, Cresson alénois, Cresson de fontaine, Cucurbitacées à peau comestible, Cucurbitacées à peau non comestible, Cultures florales et plantes vertes, Cultures légumières, Cultures ornementales, Épinard, Fraisier, Framboisier, Haricots, Haricots écossés frais, Haricots et Pois écosses frais, Haricots et Pois non écosses frais, Haricots et pois non écossés frais, Laitue, Légumes racines et tubercules tropicaux, Melon, Navet, Oignon, Plantes d’intérieur et balcons, Poireau, Pois écossés frais, Poivron, Rosier, Salsifis, Tomate, Tomate – Aubergine	Artichauts, Asperges en production, Bulbiculture (bulbe, oignon, tubercule, rhizome, griffe), Cantaloups et autres melons, Carottes, Céleri rave, Céleris branches, Chicorée à café, Chicorées, Choux, Concombre, Cresson, Épinards, Fleurs et feuillages coupés, Fleurs et plantes ornementales, Fraises, Framboises, Haricots à écosser et demi-secs (grain), Haricots frais, Haricots verts (y compris haricots beurre), Laitues, Légumes frais, melons ou fraises, Navet potager, Oignon et échalote, Petits pois, Poireaux, Poivron, piment, gombo, Salsifis et scorsonère, Tomate
LFOURR	Légumineuses fourragères, Porte graine, Porte graine – Légumineuses fourragères, Prairies	Prairie artificielle (luzerne, trèfle violet, etc.)
LGRAIN	Haricots, Légumineuses potagères (sèches), Pois	Haricots secs (y compris semences), Lentilles (y compris semences), Pois secs (pois de casserie) (y compris semences)
MAIS	Céréales, Maïs, Maïs doux, Porte graine, Porte graine – Mais, Porte graine – Maïs, Sorgho	Maïs doux, Maïs grain et maïs semence, Sorgho grain
OLEA	Lin	Lin oléagineux
OLIVE	Cultures fruitières, Fruits à noyau, Olivier	Olives
ORGE	Céréales, Céréales à paille, Orge	Orge et escourgeon
PDT	Pomme de terre	Pommes de terre
POIS	Graines protéagineuses, Pois	Pois protéagineux
PP	Porte graine, Porte graine – Graminées, Porte graine – Graminées, Prairies	Prairies naturelles ou semées depuis plus de 6 ans
PROTEA	Graines protéagineuses	Lupin doux
PT	Gazons de graminées, Porte graine, Porte graine – Graminées, Porte graine – Graminées, Prairies	Prairies temporaires
RIZ	Céréales, Riz	Riz
SOJA	Arachide, Soja	Soja
STH	Porte graine, Porte graine – Graminées, Porte graine – Graminées, Prairies	STH (Superficies toujours en herbe) peu productives (pâturages pauvres)
TOUR	Tournesol	Tournesol
VERGERS	Agrumes, Amandier, Ananas, Avocatier, Bananier, Carambole, Cassissier, Cerisier, Corossol, Cultures fruitières, Cultures tropicales, Figuier, Fruit de la passion, Fruits à noyau, Fruits à pépins, Goyavier, Kaki, Kiwi, Litchi, Manguier, Papayer, Pêcher, Pêcher – Abricotier, Petits fruits, Pommier, Prunier	Abricots, Actinidia (Kiwi), Agrumes, Amandes, Ananas, Avocat, Banane plantain, Cassis et myrtilles, Cerises, Corossol, pomme cannelle, Figues, Fruits à pépins, Fruits tropicaux et subtropicaux, Letchi, longani, ramboutan, Mangue, Maracuja, fruits de la passion, grenadille, Pavies, pêches, nectarines et brugnons, Petits fruits, Pommes à cidre, Pommes de table, Prunes
VIGNES	Vigne	Vignes

Table 22: Original version of table 4

Groupe de cultures BNVDs	Libellé	Code culture RPG	Code culture RPG complété
BETT	Betterave non fourragère	BTN	Betterave, Betteraves
BLE	Blé	BTH, BDT, BDP, TTH, EPE, BDH, BTP, TTP	Ble
CEREALES	Autres Céréales	CHH, CPA, CGF, CAG, CHA, CGS, CPZ, CHS, CHT, AVH, CPH, CPT, SGH, CPS, SRS, AVP, SGP, CGP, MCR, CGH	Autres cereales, Autres Cereales, Cereales
COLZA	Colza	CZP, CZH	Colza
COQUE	Fruits à coque	CTG, NOS, NOX, PIS, CAB	Chataigne, Coques, Fruits a coque, Noix
FEV	Féverole	FVL, FVT
FIBRE	Plantes à fibres	LIF, CHV	Chanvre, Plantes a fibre, Plantes a fibres, Plantes fibres
FOURR	Graminées et autres fourrages	CAF, BVF, RDF, GFP, FAG, PAT, FLO, CHF, CPL, PP6, PP5, FSG, DTY, XFE, GAI, PH6, NVF, PH5, FET	Fetuque, Fourrage, Fourrages, Fourrages FFO
GEL	Gel	J6P, JNO, J6S, J5M	Jachere, Jachere (Legumes)
INDUS	Cultures industrielles	BAR, EST, VNB, YLA, MTH, ROM, PMD, MAV, CIB, SRI, MOT, FNO, MLI, VNV, CRF, CMM, ANE, CHR, HBL, VNL, PPP, SGE, ANG, TOT, FNU, CUR, OSE, PAR, BRH, CAV, PPA, PSN, MRJ, TAB, PSL, CUM, CRD, MLP, CML, BAS, PSY, VAL, PPF, THY, LAV, ANI	Cult indus, Culture Indus, Culture industrielle, Culture industrielle (Serre), Culture industrielle (Verveine), Houblon, Lavande, Plantes aromatiques, Plantes aromatiques (Serre), PMD (Ginkgo biloba), Romarin, Sauge
LEGU	Légumes ou fleurs	POR, GER, CES, TOM, PVP, FLA, CAR, PAS, AIL, CEL, CRA, PMV, LBF, CCT, MAC, CSS, DOL, CMB, HSA, HPC, HAR, PPO, MLO, CRN, CCN, ART, FLP, TOP, AUB, PAN, CHU, EPI, MRG, PSE, BUR, CRS, NVT, ROQ, BLT, SFI, FRA, RDI, OIG, POT, LSA, PAQ, RUT, VER	Batiment fleurs, Batiment legumes, Fleurs, Fleurs (batiment), Fleurs (Serre), Fleurs bati, Fraises, Jardin collectif, Legumes, Legumes (Serre), Legumes (Serre), Legumes (batiment), Legumes (Serre), Legumes ART, Legumes bati, Legumes CCN (Serre), Legumes CHU, Legumes CSS, Legumes FLP, Legumes FRA, Legumes MLO, Legumes MOT, Legumes plein champs, Legumes PPO, Legumes PTC, Legumes TOM, Pepiniere (Fleurs), Pepiniere (Fleurs, Serre), Pepiniere (Plants), Pepiniere (Serre), Pepiniere fleurs, Plants Legumes, Plants Legumes (Serre), Plants maraichers, Plants maraichers (Serre), Plants PFR, Plants PFR (Serre), Serres, Serres (divers), Tomates
LFOURR	Légumineuses fourragères	GES, ME5, FFO, LOT, VED, PFP, LH6, LU5, SE5, MH5, ME6, TR5, MC5, LUD, FF6, SA6, LFH, MC6, LUZ, JO5, JOD, LP5, SAD, MEL, LFP, TRD, VE6, JOS, LEF, JO6, LP6, SAI, VES, SED, TRE, ML5, SE6, FF5, MIN, MED, SER, SA5, MLD, TR6, MLG, LU6, PFH, ML6, LH5, MH6, VE5	Legumimeuses, Legumineuses, Luzerne, Sainfoin
LGRAIN	Légumineuses à grains	PCH, LEC	Legumineuses a grains
MAIS	Maïs	MOH, MIE, MLT, MCT, CGO, SOG, MID, MIS	Mais, Miscanthus
OLEA	Autres Oléagineux	NVH, OEH, MOL, OHN, LIP, OEI, NVE, OAG, OHR, OPN, LIH, OPR	Autres oleagineux, Oleagineux
OLIVE	Oliviers	OLI	Olivier, Oliviers, Pepiniere (Oliviers)
ORGE	Orge	ORP, ORH	Orge
PDT	Pommes de Terre	PTC, PTF	Legumes PdT, Pommes de terre
POIS	Pois Protéagineux	PHI, PPR, PPT
PP	Prairies permanentes	PPH, PRL	Prairie, Prairie (chevaux), Prairie (Chevaux), Prairie (Elevage), Prairie (rotation longue), Prairie (Vergers)
PROTEA	Autres Protéagineux	MPT, LDP, FEV, LDT, PAG, MPC, LDH	Proteagineux
PT	Prairies temporaires	RGA, PTR	Prairie Tempo, Prairie temporaire, Prairie Temporaire, Ray Grass
RIZ	Riz	RIZ
SOJA	Soja	SOJ, ARA	Soja
STH	Estives et landes	SPH, SPL, BOP	Bois pature, Landes, Landes (pelouses), SPH, SPL, Surface pastorale (SPH), Surface pastorale (SPL)
TOUR	Tournesol	TRN	Tournesol
VERGERS	Vergers	BCA, ANA, VRG, AVO, VGD, BEI, PFR, AGR, BCI, BCP, CAC, BCF, PWT, BEF, PRU, BER, CBT, PVT, BCR, BEA, BEP, PEP	Agrumes, Amandes, Pepiniere (Vergers), Petit fruit rouge, PFR, PFR (Serre), Verger, Vergers, Vergers (CBT), Vergers PRU
VIGNES	Vignes	VRT, RVI, VRC	Vignes, Vignes (raisins de cuve), Vignes (raisins de table), Vignes (restructuration)
DIVERS	Divers		Arbuste baies, Pepiniere, Pepiniere (arbre), Pepiniere (arbres), Pepiniere (bois), Pepiniere (Callunas), Pepiniere (Callunas, Serre), Pepiniere (ONF), Pepiniere (sapin), Pepiniere (Sapin), Pepiniere bois
SEMENCES	Semences		Semences, Semences (Serre)