Land sharng versus land sparng to protect water from pestcde polluton? Sophe Legras Elsa Martn Vrgne Pguet CESAER UMR INRA AgroSup Djon Very prelmnary verson Summary: In ths paper we nvestgate the concept of land sharng and land sparng strateges n the context of water polluton by pestcdes. A land planner wllng to reduce the rsk of transfer of pestcdes to neghbourng water bodes may ether nduce the adopton of ntegrated pest management practces by farmers through the use of ncentves the land sharng strategy or take agrcultural land out of producton by purchasng t the land sparng strategy. We propose an ex ante assessment framework to help choosng between the sharng and sparng strateges at a water catchment level to acheve water qualty goals wth respect to pestcde polluton. After proposng a theoretcal model of parcel selecton we develop an emprcal procedure on the Sene Ource rver catchment n Burgundy France. We confrm that the targetng crteron based on the maxmzaton of envronmental gans under an economc cost constrant provdes hgher envronmental gans than the other targetng crtera where the land planner maxmzes the surface selected under a budget constrant or the total envronmental gans under a surface constrant both for the land sharng strategy and for the land sparng strategy. We show that a combnaton of both strateges maxmzes envronmental gans a concluson that allows movng forward n the land sharng/sparng debate. Keywords: water polluton pestcdes ntegrated pest management ste selecton JEL classfcaton codes: Q18 Q24 Q25 Q51 Introducton The world agrcultural producton has ncreased consderably over the last decades thanks to technologcal nnovatons and to the ncreasng use of nputs n partcular of pestcdes. However the massve use of pestcdes has many negatve effects on the envronment (see Pmentel et al. 1992 for more detals) a major one beng the polluton of water bodes (see Ongley 1996 for more detals). Ths assessment s partcularly true n Europe where publc polces such as the European Water Framework Drectve or the Drectve 2009/128/EC on sustanable use of pestcdes am at reducng water polluton by pestcdes. To acheve the water qualty goals defned n the European 1
legslaton European local decson makers can mplement two man strateges: a land sharng one or a land sparng one. These strateges were orgnally defned n the case of bodversty conservaton (Green et al 2005) opposng wldlfe frendly farmng that would mprove wld populaton on farmland at the expense of decreasng agrcultural yelds and land sparng that would reduce demand for farmland through an ncrease n yelds. However these concepts also prove relevant to analyse water polluton snce for nstance Hascc and Wu (2006) show that land uses sgnfcantly affect the level of water polluton n a water catchment. Appled to the case of water polluton by pestcdes the land sharng strategy would consst n mplementng economc nstruments such as taxes or subsdes to gude farmers toward sustanable pest management strateges (see for nstance Sexton et al. 2007 for more detals) whle the land sparng strategy would consst n purchasng and excludng from agrcultural producton the lands wth the hghest rsk of pestcde contamnaton. One queston of nterest s then to assess the best strategy to mplement. The answer s not straghtforward and depends on the case consdered. In the water polluton case the most relevant scale of analyss s the water catchment level (see Coner et al. 2001). Langpap et al. (2008) provde an ex post analyss of the best land use polces wth respect to the protecton of water catchment ecosystems. We propose n ths work to develop an ex ante method that helps n choosng between a land sharng and a land sparng strategy at a water catchment level to acheve water qualty goals wth respect to pestcde polluton. The lterature on reserve ste selecton manly based on conservaton bology s very helpful n defnng an ex ante land sparng strategy. For nstance Margules et al. (1988) produce a method to select land to preserve that conssts n maxmzng bologcal bodversty. Extensons of ths work propose to swtch from a maxmzaton of the number of speces preserved only to a strategy that also consders the cost of land purchase. For nstance Ando et al. (1998) compare both methods and show that the herarchy of lands to preserve can be very dfferent between methods. The frst contrbuton of our work conssts n adaptng ths lterature to the problem of water polluton by pestcdes. To do so we need to measure the economc costs and the envronmental gans lnked to the purchase of land for conservaton purpose. Concernng the economc cost of land purchase we follow Newburn et al. (2006) by mplementng the frst stage of the hedonc method (see Palmqust 2005 for more detals on the method and Bastan et al. 2002 for an applcaton next to ours because of beng based on geographc nformaton system). Indeed ths measurement s a good approxmaton of the opportunty cost of puttng land nto reserve. The envronmental gans from land sparng are measured by the rsk of pestcde contamnaton before purchase snce the afterpurchase land use the reserve s assumed to bear a zero rsk of pestcdes contamnaton. Babcock et al. (1996) measure the envronmental gans lnked to surface water qualty n a land sharng case wth the dstance of land from water bodes. Babcock et al. (1997) measure envronmental gans lnked to groundwater vulnerablty to pestcde leachng wth an ndex provded by Kellogg et al. (1992) that s a functon of sol leachng potental pestcde leachng potental precptaton and chemcal use. We mplement a cumulated I Pest ndcator that measures the rsk assocated wth pestcde applcaton (see van der Werf and Zmmer 1998 for more detals on the method) at the plot level. 2
In the land sharng strategy both envronmental gans and economc costs are more dffcult to compute at a water catchment level. There s a wde lterature about the land sharng strategy appled to the bodversty conservaton problem partcularly wthn the framework of the Amercan Conservaton Reserve Program (see for nstance Wu and Boggess 1999). Indeed ths framework provdes bds made by farmers as a measurement of economc costs. Such datasets do not exst n the European case. Followng Coner et al. (2001) we focus on a specfc French water catchment. The costs of the land sharng strategy are valued wth feld surveys related to farms accountng data. The surveys also allow us to nvestgate an orgnal land sharng strategy: Integrated Pest Management (IPM). Ths s our second man contrbuton. Stern et al. (1959) frst developed the IPM concept. It s a multdscplnary concept that promotes physcal and bologcal regulaton strateges to control pests whle reducng the relance on pestcdes. Lechenet et al. (2014) show that t can theoretcally be an effcent tool for reducng pestcde use at the least cost n French arable farmng. Boussemart et al. (2011) show that less pestcde use per ha may be preferable for producers to reduce ther costs. Wlson and Tsdell (2001) examne why farmers contnue to use pestcdes despte the ncreasng costs; n partcular they emphasze a lock n phenomenon n pestcdes use. To solve ths dffculty Jacquet et al. (2011) shows that fscal schemes can be helpful n the French case. It s why we decded to focus on farm accountng data to value the economc cost of the land sharng strategy. Once we have computed the envronmental gans and economc costs of land sharng and land sparng strateges we follow Babcock et al. (1996) and mplement three alternatve targetng crtera for both: the maxmzaton of the sze under an economc cost constrant the maxmzaton of envronmental gans under a sze constrant and the maxmzaton of envronmental gans under an economc cost constrant. We confrm that for a gven budget the targetng crteron based on the maxmzaton of envronmental gans under an economc cost constrant provdes hgher envronmental gans than the other targetng crtera both for the land sparng strategy and for the land sparng strategy. Fnally followng the lterature on land sharng and land sparng relatve to bodversty conservaton (see Fsher et al. 2014 for more detals) we propose to compare both strateges. We show that a combnaton of both strateges maxmzes envronmental gans a concluson that allows movng forward n the land sparng/sharng debate. Ths s our thrd contrbuton. Secton 1 presents our theoretcal framework. Our emprcal strategy s detaled n secton 2. Fnally the results are presented and dscussed n secton 3. Secton 4 provdes some concludng remarks. 1. Theoretcal framework We analyse the stuaton where a planner responsble for the water polcy ntervenes on the agrcultural land market whch conssts of I felds. Let 1 I be an ndcator for each ndvdual feld of sze s. In the land sharng case we assume that each unt of land can generate ether net returns c h or envronmental gans b h. In the land sparng case the net returns are denoted c p and the envronmental gans b p. c p can be nterpreted as the opportunty cost of agrcultural producton whch we take to be the mnmum amount that the planner has to spend n order to take ths land 3
out of producton. c h s the mnmum amount the planner has to bd n order to make farmers move toward IPM. The envronmental gan s n terms of pestcde polluton avoded n water bodes. We are gong to assume three types of land plannng behavours usable for the land sharng strategy for the land sparng one and for the strategy consstng n mxng both: The frst assumpton A1 s one of a land planner who only has an economc objectve n mnd: the planner only consders the cost of the strategy and ams at maxmzng the sze of land on whch to ntervene under an economc cost constrant. In ths case the amount of money spent cannot be hgher than the budget B. The second assumpton A2 s one of a land planner who only has an envronmental objectve n mnd. The planner only consders the envronmental gans of the strategy that he ams at maxmzng. Wthout a constrant such a strategy would result n ntervenng on the whole land under the jursdcton of the planner. Ths would nduce a problem of socal acceptablty that can be consdered by addng a sze constrant to the optmzaton of envronmental gans. In ths case the sze of land on whch the land planner ntervenes cannot be hgher than the area A that s assumed lower than the total area under the jursdcton of the land planner. The thrd assumpton A3 s one of a land planner who has both envronmental and economc objectves n mnd: the planner maxmzes the envronmental gans under an economc cost constrant. 1.1 Land sharng and land sparng strateges separately Let x h denote the amount of unt subsdzed by the land planner for the agrcultural converson to IPM and x p the amount of unt purchased by the land planner for ts water qualty amentes. Obvously x h and x p l. We propose to detal the three types of behavours (A1 to A3) n the land sharng case (A1h to A3h) but one can smply change notatons to obtan the land sparng case h becomes p. In A1h case the optmzaton problem s: max x s.t. x c andx s xh I I h h h h 1 1 Let 1 be the shadow prce of the budget constrant and 1 be the shadow prce of the th unt s producton capacty. The Lagrangan functon for ths problem s: I I I L x B x c s x h 1 h h 1 h 1 1 1 Let the values of the optmal soluton be gven by are: x * h the Kuhn Tucker condtons of the problem 4
I I * * Bxh ch 0 1 0 1 Bxh ch 0 1 1 * * s xh 0 1 0 1 s xh 0 1... I * * 1 1 c h 1 0 x h 0 xh. 1 1 c h 1 0 1... I A1h Ths system of equatons mples that: * sf1 ch 1 * * * xh f1 ch 1 where 1 denotes the optmal shadow value of the ncrease n the area 1 * xh f1 ch 1 converted to IPM assocated wth an ncrease n the budget constrant. Unts of land wth the nverse * of converson cost greater than 1 wll be converted whle unts wth strctly lower ratos wll stay n producton. In A2h case the optmzaton problem s: max x. b s.t. x andx s xh I I h h h h 1 1 Let the values of the optmal soluton be gven by x and ** h ** 2 for the optmal shadow prce of the sze constrant. We deduce from the new Lagrangan functon and from the Kuhn Tucker condtons appled to ths functon that: ** s fbh 2 ** ** ** xh 0f bh 2 where 2 denotes the optmal shadow value of the ncrease n 2 ** xh fbh 2 envronmental amenty assocated wth an ncrease n the sze constrant. Unts of land wth ** envronmental gans greater than 2 wll be converted whle unts wth strctly lower envronmental gans wll stay n producton. In A3h case the optmzaton problem s: max x. b s.t. x c andx s xh I I h h h h h 1 1 Let the values of the optmal soluton be gven by x and *** h *** 3 for the optmal shadow prce of the budget constrant. We deduce from the new Lagrangan functon and from the Kuhn Tucker condtons appled to ths functon that: 5
*** sfbh ch 3 *** *** *** xh fbh ch 3 where 3 denotes the optmal shadow value of the ncrease n the 3 *** xh fbh ch 3 envronmental amenty assocated wth an ncrease n the budget constrant. Unts of land wth ratos *** of envronmental gans to costs of converson greater than 3 wll be converted whle unts wth strctly lower ratos wll stay n producton. 1.2 Combnaton of land sharng and land sparng strateges We now turn to the case n whch the land planner can mplement ether a land sharng strategy or a land sparng one on each unt of land. We make the same assumpton as before on the planner behavour except that now he combnes strateges (A1hp to A3hp). In A1hp case the optmzaton problem wth a combnaton of strateges becomes: max x x s.t. x c x c andx x s xh xp I I I I h p h h p p h p 1 1 1 1 Let 1 be the shadow prce of the budget constrant and 1 be the shadow prce of the th unt s producton capacty. The Lagrangan functon for ths problem s: I I I I I L x x B x c x c s x x h p 1 h h p p 1 h p 1 1 1 1 1 Let the values of the optmal soluton be gven by problem are: x and ' h x ' p the Kuhn Tucker condtons of the I I I I ' ' ' ' Bxh ch xp cp 0 1 0 1 Bxh ch xp cp 0 1 1 1 1 ' ' ' ' s xh xp 0 1 0 1 s xh xp 0 1... I ' ' 11 ch 1 0 xh 0 xh. 11 ch 1 0 1... I ' ' 1 c 0 x 0 x. 1 c 0 1... I 1 p 1 p p 1 p 1 A1hp Ths system of equatons helps n choosng between a land sharng strategy and a land sparng one. When the budget constrant s bndng unts of land wth costs of converson hgher than purchase costs are purchased and unts of lands wth costs of converson hgher than purchase costs are converted. In ths case system of equatons (A1hp) mples that: 6
' ' xh sandxp 0f ch cp 0 ' ' xh 0and xp sf ch cp 0 ' 1 ' 1 ' 1 xh xh xp xp andxh xh sfch cp 0 When the budget constrant s not bndng the logc s dfferent especally f both costs are equal snce (A1hp) then mples that 1 =0 and: x x x x andx x sf 1 ' 1 ' 1 ' 1 ' h h p p h h 1 Hence the value of the shadow prce of the th unt s producton capacty plays a crucal role n the choce between a land sharng or a land sparng strategy. In A2hp case the optmzaton problem wth a combnaton of strateges becomes: max x. b x. b s.t. x x andx x s xh xp I I I I h h p p h p h p 1 1 1 1 Let the values of the optmal soluton be gven by x and functon and from the Kuhn Tucker condtons appled to ths functon that: '' h x '' p. We deduce from the new Lagrangan '' '' xh s andxp 0 fbh bp 0 '' '' xh 0 andxp sf bh bp 0 '' 2 '' 2 '' '' xh xh xp xp andxh xp sfbh bp 0 Here the envronmental gans are decsve n the choce between a land sharng or a land sparng strategy: the strategy nducng the hghest envronmental gans wll be chosen. In A3hp case the optmzaton problem wth a combnaton of strateges becomes: max x. b x. b s.t. x c x c andx x s xh xp I I I I h h p p h h p p h p 1 1 1 1 Let the values of the optmal soluton be gven by x ''' h ''' ''' x p and 3 the optmal shadow prce of the budget constrant. We deduce from the new Lagrangan functon and from the Kuhn Tucker condtons appled to ths functon that when the budget constrant s bndng: 7
x sand x 0 f b b c c x s x b c b c x x s b b c c x x s b c b c x x x x x x x s b c b c ''' ''' ''' h p h p h p 3 ''' ''' h and p 0 f h h p p 0 ''' ''' ''' h and p f h p h p 3 ''' ''' h and p f h h p p 0 ''' 3 ''' 3 ''' ''' ''' h xh xp xp andxh xp sf bh bp ch cp 3 ''' 3 ''' 3 ''' ''' h h p p and h p f h h p p 0 The choce between a land sharng and a land sparng strategy s even more complex than for A1hp case snce t depends () on the dfference of the beneft to cost rato between each strategy () on ''' the rato of beneft dfference to cost dfference and () on 3 the optmal shadow value of the ncrease n the envronmental amenty assocated wth an ncrease n the budget constrant. Consequently dependng on the budget and the dstrbuton of costs and benefts over land unts and strateges a land planner may select only unts for purchase or only unts for converson or a mx of both whatever hs strategy s (A1 or A2 or A3). 2. Emprcal procedure Data from the Sene Ource catchment are used to analyse the mplcatons and dscuss the effcency of strateges A1 A3 descrbed above. Located n the heart of Burgundy France t s a mostly agrcultural area spannng over 71 muncpaltes and 80 000 hectares. Area under agrculture represents 59 300 hectares of whch 21% are permanent pastures 63 % are dedcated to cereal crops and 16 % to fodder producton. Levels of pestcdes above the regulatory thresholds are regularly measured n the area s waterbodes (ARS Bourgogne 2009). The emprcal procedure conssts n three steps: () the computaton of envronmental gans n the catchment under study and for both strateges () the computaton of the economc costs and () the smulaton of land planner decson for the three strateges under study. 2.1. Computaton of the envronmental gans To estmate the envronmental gans attached to each strategy we rely on I Phy (van der Werf and Zmmer 1998; Bockstaller et al. 2008) a predctve ndcator that assesses the envronmental mpacts of pestcde use as the rsk of contamnaton of the ar and surface and ground waters. It s calculated at the parcel scale based on dfferent nputs : envronmental varables (leachng potental runoff potental dstance to rver) farmng practces varables (crop type actve ngredent sowng date sprayng date banded sprayng equpment characterstcs presence of crop resdues) pestcdes characterstcs (actve ngredent moblty half lfe toxcty exposure transfer rate). The applcaton rate ntervenes at the last step of the calculaton once the rsks of contamnatons of ar surface and groundwater have been assessed 1. 1 The «ar» component s not accounted for n the analyss due to our focus on water polluton. 8
The French database systems from IGN (Insttut Géographque Natonal the French Natonal Geographc Insttute) and from GIS Sol (Groupement d Intérêt Scentfque Sol Sol Scentfc Interest Group) are used to gather topographcal and pedologcal data at the parcel scale: sol texture organc matter content slope etc. Ths allows us to compute the runoff and leachng potentals and dstance to the rver for each parcel. To characterze current farmng practces at the catchment scale we adopt a two step procedure. From feld farm surveys we dentfed sx typcal current feld crop producton systems dependng on the type of crop rotaton (rapeseed wheat barley or rapeseed wheat slage maze wheat) and tllage practces (no tllage systematc tllage tllage on rapeseed and maze and no tllage on wheat and barley). Each typcal producton system s precsely characterzed wth respect to tllage sewng fertlzaton harvestng and crop protecton practces (lke n Lechenet et al. 2014). Then to dentfy the parcels concerned by these typcal producton systems n the Sene Ource catchment we use the RPG (Regstre Parcellare Graphque Land parcel dentfcaton system) a cartographc feld pattern regstry contanng at parcel level all productve land uses recevng payments from the EU CAP and wth a typology of agrcultural producton n 26 classes that s suffcently fne for selectng plots under typcal producton systems. We select the plots that have been regstered as growng our typcal crops of nterest over the 2006 2009 perod to specfy the sample of plots to whch the land sharng or sparng strateges can be appled. The pestcdes nputs component of the I Phy calculaton s nformed by the depcton of the typcal producton systems. Consequently we obtan an I Phy score for each parcel for years 2006 to 2009 based on actual land uses and typcal producton systems; we average these scores to produce an I Phy score for each parcel under study before land plannng. Then the envronmental gan scores used n the optmzaton are the dfferences between I Phy scores after and before land plannng mplementaton an assessment of the reducton n the rsk of contamnaton of waterbodes by pestcdes. In the land sparng strategy the I Phy score after land plannng mplementaton s set equal to ts maxmum (10) snce we assume no agrcultural producton once the plots have been acqured by the land planner; n the land sharng strategy t s computed wth the assumpton that IPM producton systems are mplemented. The IPM producton systems are characterzed based on the man objectve of reducng the dependency of farmers on pestcdes and more precsely herbcdes hence the envronmental mpact of neghborng waterbodes. Two typcal IPM producton systems are conceved (wth and wthout fodder crops). They span over 6 years nclude a dversty of crops such as pea barley mxes and are more relant on sol labour than conventonal producton systems. However pestcdes are not completely banned but greatly reduced (see Appendx 1). Accordng to the crop or mxed crop/lvestock orentaton of the farms they are allocated an IPM producton system; the I Phy score after land plannng decson s averaged over the sx years of the IPM rotaton. 2.2. Computaton of the economc costs The computaton of the economc costs dffers dependng on whch strategy s analysed. The man orgnalty of our work s that they are estmated before the mplementaton of a strategy ex ante. 9
The land sharng strategy conssts n provdng ncentves to farmers to adopt IPM producton systems to reduce ther dependency to pestcdes n order to mprove water qualty. Consstent wth the current European polcy framework we concentrate on the case of a subsdy to convert from conventonal to IPM crop producton. Ideally an optmal subsdy scheme would be ndvdualzed to account for dfferences n margnal costs and benefts assocated wth the converson for each plot under scrutny. However actual programs rely on fxed per hectare payments. The sem net margns of each producton system analysed above (conventonal and IPM) are used to compute the per hectare subsdy level n the Sene Ource catchment. Wthn each producton system for each crop the producton costs are computed for a typcal set of farmng practces nformed by the feld surveys and expert knowledge. To compute the mechanzaton and fuel costs we use data from the Bureau de Coordnaton du Machnsme Agrcole (French agrcultural mechanzaton coordnaton offce) about purchase and mantenance costs workng output and fuel consumpton of the agrcultural machnes used n the producton systems (whch are more numerous n the IPM cases). Seed prces are taken from a natonal synthess from the Groupement Natonal Interprofesson des Semences (French natonal unon of seed producers) for the 2010/2011 campagn pestcdes prces from the Crter database from INRA (French natonal nsttute for agronomc research see Fortno and Reau 2010) and fertlzers prces from the Agreste statstcs (French natonal agrcultural statstcs). Refer to Appendx 1 for a descrpton of the producton costs n the dfferent producton systems. Fnally the sem net margn s calculated under the assumpton that yeld objectves hence gross products are the same for conventonal and IPM producton systems. A global sem net margn before land plannng mplementaton s calculated averaged over the four years of land occupaton for whch we have extracted data from the RPG. The global sem net margn after land plannng decson s based on the assumpton of a generalzed adopton of the IPM producton systems assumng that permanent pastures are not affected and that parcels that were under a cereal or cereal/fodder crop orentaton reman as such under IPM. The opportunty cost of the land sharng strategy s then establshed at 50 euros per hectare per year for a cereal only farm orentaton and 15 euros per hectare per year for a cereal and fodder crop farm the amounts necessary to ensure no loss n sem net margn at the catchment scale for each orentaton. These levels of subsdy are of the same order of magntude as agrcultural payments currently n place n France such as the Agro Envronmental Measures for the dversfcaton of cultures set at 32 euros per hectare per year. In the land sparng strategy we assume that the cost of the land plannng decson s equal to the cost of the purchase of land. To estmate ths cost the frst step s to use the hedonc prce method that conssts n breakng up the prce of a land nto ts characterstcs. Ths estmaton s based on sales that really occurred. The water catchment of Sene and Ource s too small to conduct such an estmaton. In France analysts of agrcultural land markets (SAFER Socété d Aménagement Foncer et d Etablssement Rural French companes for agrcultural real estate management and rural development) work at the Département level. Thus we work at the level of Côte d Or that s the French department of 8 763 km² located n Burgundy n whch the Sene and Ource water 10
catchment s located. We mplement the followng sem logarthmc model of hedonc prce on 6 477 sales recorded between 1992 and 2008 by the SAFER: l t p v lnc where c ndcates the prce of the parcel n constant Euros l the vector of the characterstcs of the locaton of the parcel t that of ts topographc characterstcs p of ts pedologc characterstcs and v a vector of control varables; α represents the constant and β γ and the margnal effects of each of the explanatory varables; ε s a random error term. Most of these varables are not ncluded n the SAFER database. Consequently we use exogenous geographcal nformaton. More precsely the characterstcs of locaton of the parcel are the followng contnuous varables that we computed from ODOMATRIX (see Hlal 2010 for detals): Eucldan dstance to the town hall of the muncpalty travel dstances to shoppng and servces (closest centre of retal and servces place) to employment (closest urban centre) or to Djon the regonal captal whch makes t possble to tap the expectatons of urbanzaton n connecton wth access to some hgher functons. The topographc characterstcs also gather contnuous varables such as the surface area of the parcel ts alttude and ts slope that we computed from the data provded by IGN (Insttut Géographque Natonal the French natonal geographc nsttute). These varables gve us nformaton on the capacty n terms of agrcultural producton of the parcel studed. Most of the varables descrbed above do not make t possble to take nto account the sol characterstcs of the parcels whch consttute however one of ther essental productve attrbutes. In order to measure ths attrbute we have chosen to concentrate on the texture of the sols whch s defned by the sze of ts component partcles and provded by Donesol database (GIS Sol Sol Scentfc Interest Group). Apart from these varables we have used other more general varables that we have called control varables and that are avalable n the SAFER database: varables lnked to the year of the transacton to the fact that there s a farmer who s the purchaser of the parcel of nterest 2 or to the fact that the parcel s cropland as opposed to beng a meadowland. The converson of the prces nto constant Euros enables us to take nto account the effects of nflaton but not the evoluton between years of the real estate market for the perod under consderaton (1992 2008). We ntroduce annual dchotomous varables to control these effects. The second step of the estmaton of the purchase cost of agrcultural lands n the water catchment under consderaton conssts n usng the model estmated to predct the prce of agrcultural parcels that were or were not the object of sales. For ths purpose we needed to compute the values of our explanatory varables for all the agrcultural lands of the water catchment. Furthermore we found that the predctor of Meulenberg (1965) suts the best for our dataset. Indeed our assumpton of a sem logarthmc model calls for the addton of a correctve term to a nave predcton that would consst n smply mplementng the followng computaton c exp ˆ ˆ l ˆ t p v * * * * 2 In France the farmer of a parcel has a pre emptve rght on the sale of that parcel (see Bonon 2003 for more detals on French real estate polcy). 11
where the hats represent the value of the estmated parameters and the stars the values of the explanatory varables for parcels for whch the prce s to be predcted. Meulenberg (1965) proposes ntroducng the correctve term 1 ' x X ' X x such that: [exp( )] exp 0.5* 1 and usng c * E[exp( ˆ )] as the predcted prce of the agrcultural parcel. E ˆ ˆ 2 Fnally we proceed under the followng assumptons: f the land sharng strategy s adopted farmers are subsdzed at 50 euros per hectare per year for a cereal only farm orentaton and 15 euros per hectare per year for a cereal and fodder crop farm for a contract length of 15 years accountng for an actualsaton rate of 4%; when the land sparng strategy s chosen land s purchased for once over an nfnte tme horzon. In the Sene Ource water catchment we dentfy 2 273 lands that are relevant to mplement the optmzaton strateges descrbed n secton 1. Table 1 presents some descrptve statstcs of our man varables of nterest. Table 1: Descrptve statstcs of land surface costs and gans Mean Standard Devaton Mnmum Maxmum Surface (ha) 3.76 5.33 0.005 73.65 Land sparng cost ( ) 11410 14570 115 112359 Land sharng cost ( ) 1809 2750 0.95 38896 Land sparng gan 8.98 0.79 4.98 9.99 Land sharng gan 2.89 0.62 0.065 3.83 Land sparng gan/cost 0.005 0.008 0.00007 0.083 Land sharng gan/cost 0.029 0.103 0.00003 1.773 2.3. Smulaton of land planner decsons We smulate the decsons of three types of land planners accordng to assumptons A1 to A3 n our catchment. To do so we follow a rankng based on both Babcock et al. (1996) and our theoretcal framework. Ths rankng detaled below s a one shot rankng wthout tme delay. If the rankng crteron s equal for two or more parcels when the budget constrant bnds the soluton s not unque. Then to avod addng another ad hoc crtera those parcels are ranked randomly to determne the full set of parcel selected. When land sharng and land sparng strateges are consdered separately rankngs are qute smple and are the same as n Babcock et al. (1996). In A1h and A1p cases the land planner selects land accordng to a rankng based on costs: land wth the lowest costs are selected frst. In A2h and A2p 12
cases the planner selects land accordng to a rankng based on envronmental gans: land wth the hghest envronmental gans are selected frst. In A3h and A3p cases the planner selects land accordng to a rankng based on the beneft to cost rato: land wth the hghest rato are selected frst. When land sharng and land sparng strateges are consdered jontly rankngs are more complex. They drectly result from our theoretcal framework. In A1hp case lands are ranked (n decreasng order) accordng to the mnmum cost between a land sharng or a land sparng strategy: land wth the lowest mnmum cost wll be selected frst and assgned to the best strategy wth the lowest cost. In A2hp case the planner selects land accordng to a rankng based on the maxmum envronmental gan between a land sharng and a land sparng strategy: land wth the hghest maxmum envronmental gan wll be selected frst and assgned to the strategy wth the hghest beneft. To sum up n A1hp case a change n the strategy gven by the rankng accordng to the mnmum cost would not brng addtonal unts of surface (the crteron whch s maxmzed) whereas such a change would spend unts of budget. In A2hp case a change n the strategy gven by the rankng accordng to the maxmum envronmental gan would not brng addtonal envronmental gan (the crteron whch s maxmzed) whereas such a change would consume unts of surface. Ths s qute dfferent n A3hp case where a change n the strategy gven by the rankng accordng to the maxmum beneft to cost rato would brng an addtonal beneft (the crteron whch s maxmzed) at the expense of unts of budget. In other words such a change would not brng addtonal beneft to cost whereas t could generate addtonal beneft wth respect to addtonal costs. Therefore n A3hp case lands are frst ranked (n ncreasng order) accordng to the maxmum beneft to cost rato between a land sharng or a land sparng strategy. To be selected each land must generate a maxmum beneft to cost hgher than the maxmum dfference of beneft to cost nduced by a change of the selecton strategy chosen for the prevous land selected j : max( bh ch bp cp ) max(( bs2j bs1j) ( cs2j cs1j)) j. If t s not the case the prevous land j selecton strategy ncrmnated wll swtch from s1 to s2. Ths procedure s mplemented untl the budget constrant bnds. 3. Results and dscusson We can frst deduce from Table 1 that on average a land sharng strategy s preferred n A1 case (the mean cost s lower) and n A3 case (the mean gan/cost s hgher). A land sparng strategy s preferred n A2 case (the mean gan s hgher). When land sharng and land sparng strateges are consdered separately ths result s partly confrmed outsde the average. Indeed Table 2 shows that for a gven budget the number of parcels the surface and the envronmental gan (percentage of the total possble gans f all lands were selected wth the strategy) of lands selected are hgher wth the sharng strategy than wth the sparng one. For a gven surface the envronmental gan and the number of parcels converted are hgher wth the sharng strategy than wth the sparng one; the cost s then lower wth the sharng strategy than wth the sparng one. Thus outsde average t s better to choose a land sharng strategy than a land sparng one. 13
More partcularly we know from Table 2 that for a gven budget equal to 200 000 the envronmental gan and the number of parcels s hgher n A3h and A3p cases than n A1h and A1p cases. The global surface selected s hgher n A1h and A1p cases than n A3h and A3p cases. Ths means that for a gven budget the envronmental economc land planner (A3 maxmsaton) ncreases envronmental gans wth respect to a pure economc land planner (A1 maxmsaton). Furthermore an envronmental economc land planner favours small parcels (mean sze of 0.45 ha wth a land sharng strategy and of 0.12 ha wth a land sparng strategy) especally wth respect to a pure economc land planner (mean sze of 3.23 ha wth a land sharng strategy and of 10.47 ha wth a land sparng strategy). For a gven surface (equal ether to A1 surface or to A3 surface) the envronmental gan the cost and the number of parcels are hgher n A2h and A2p cases than n A1h A1p A3h and A3p. Ths means that a pure envronmental land planner (A2 maxmsaton) ncreases the envronmental gan but also the cost of the nterventon. If we look nto more detals at the results we can conclude that for a gven surface a pure envronmental land planner favours small parcels (mean sze of 0.84 ha for A1 surface constrant and of 0.45 ha for A3 surface constrant n the land sharng strategy and mean sze of 0.22 ha for A1 surface constrant and of 0.11 ha for A3 surface constrant n the land sparng strategy) wth respect to both the other types of planners. Fnally envronmental objectves favour small parcels whle economc objectves favour large one. Ths s consstent wth economc ntuton snce large parcels are the cheapest. Concernng the envronmental gan the ndcator of pest contamnaton that we use s ndependent of the parcel sze. Ths explans that small parcels are selected frst to maxmze envronmental gan. Table 2: Total cost envronmental gan surface and number of parcels for separated strateges Cost ( ) Env. Gan (and %) Surface (ha) Nb. of parcels Sparng Sharng Sparng Sharng Sparng Sharng Sparng Sharng A1 case 200000 200000 138 (0.7) 631 (9.6) 157 1140 15 353 A2 case (A1 surface) 651346 569917 6481 (31.8) 4020 (61.1) 157 1140 714 1364 A2 case (A3 surface) 277261 234231 3657 (17.9) 3116 (47.4) 43.5 472 404 1057 A3 case 200000 200000 3198 (15.7) 3012 (45.8) 43.5 472 350 1047 Table 4 of Appendx 2 shows how results of Table 2 evolve wth the budget constrant. Every varables (envronmental gan number of parcels surface and mean sze of parcel selected) ncrease wth the budget except the mean sze of parcel selected n A1 case. Indeed wthn a land sparng strategy a pure economc land planner selects parcels of smaller mean sze when the budget ncreases. Ths s due to the fact that ths land planner purchases frst the larger parcels that become less abundant when the budget ncreases. Wthn a land sharng strategy there s no clear trend concernng the sze of parcels. One can explan ths by the fact that the heterogenety of costs of converson lnked to the dfferentaton between cereal only farm and fodder crop farm s more 14
mportant than the fact that these costs ncrease wth the sze of the parcel (they are proportonal to the surface). At ths step when consdered separately t s better to choose a land sharng strategy than a land sparng one n the Sene Ource catchment. However our theoretcal framework brngs to the fore that t s not that smple. Let s now turn to the case where the land planner has the possblty to combne strateges by acqurng some parcels and subsdzng the adopton of IPM on others. Table 3 shows that wth a total budget of 200000 t s optmal to enforce a combned strategy under assumpton A3: 253 parcels of total surface 2.86 hectares are purchased at a cost of 126444 and for an envronmental gan of 2319 whle 520 parcels of total surface 174 hectares are converted to IPM at a cost of 73556 and for an envronmental gan of 1508. Two thrd of the total budget s allocated for the purchase of 14% of the total surface concerned by land plannng that provde 60% of the resultng envronmental gans. Wth the same budget of 200000 under assumpton A1 only the sharng strategy s enforced: 356 parcels are converted to IPM over a total surface of 1140 hectares for an envronmental gan of 640. The land plannng decson covers a wder area than under assumpton A3 wth a lower envronmental gan. The per hectare cost dfferental between the land sharng and land sparng strateges explans n part the lack of combned strateges mplemented by the land planner. An mmedate extenson of ths paper wll be to alter contract length or the dscount rate (set at 4%). Under assumpton A2 only the sparng strategy s enforced. Wth a surface constrant set at the total area of land under land plannng (sharng or sparng) under assumpton A3 (202.6 hectares) 793 parcels are purchased for a total cost of 808779 euros and for an envronmental gan of 7182. Implement land plannng over the same surface comes at a hgher prce (4 tmes hgher) wth ncreased envronmental gans (below twce the benefts). Wth a laxer surface constrant set at the area converted under assumpton A1 1376 parcels are purchased for a total cost of 3644718 euros and envronmental gans of 12398 : ths represents a 5.6 tmes ncrease n benefts for a prce ncrease by 4.5 tmes. The fact that a combned strategy s not enforced s due to a characterstc of our context snce the adopton of IPM producton system wll always produce a rsk of pestcdes contamnaton even f very low hgher than f the same parcel s purchased and put out of producton. Then for the same parcel the envronmental gans of the land sparng opton are always greater than that of the land sharng opton (see Table 1). Consequently a land planner lookng at maxmsng an envronmental objectve wll not have recourse to the sharng strategy even f gven the possblty to do so. Table 3: Total cost envronmental gan surface and number of parcels for combned strategy Cost ( ) Env. Gan Surface (ha) Nb. of parcels Sparng Sharng Sparng Sharng Sparng Sharng Sparng Sharng A1hp case 200000 640 1140 356 A2hp case (A1 3644718 12398 1140 1376 15
surface) A2hp case (A3 surface) 808779 7182 202.6 793 A3hp case 126444 73556 2319 1508 28.7 174 253 520 Table 5 of Appendx 2 shows how the contrbuton of each strategy to total benefts costs and surface under protecton evolves wth the budget constrant under assumpton A3. A tghter budget constrant decreases the share of funds allocated to the land sparng strategy whch concerns a lower porton of total acreage under protecton (2% for a total budget of 50000 for nstance) and provdes a lower share of the total envronmental gans. Note that f the cost and beneft shares of both strateges are of the same order of magntude over the budget range the surfaces nvolved n the land sparng strategy always lay below those concerned by the adopton of IPM producton systems. Fnally a comparson of Table 2 and Table 3 shows that a pure envronmental land planner spends more or the same amount of money when he combnes land sharng and land sparng strateges than when he mplements them separately. The surface and the number of parcels selected when strateges are combned are always hgher or equal than the one obtaned when strateges are separated except for envronmental economc land planner who selects less and smaller parcel when combnng strateges. However the envronmental gan s always hgher when strateges are combned than when they are consdered separately. For nstance for a gven budget of 200000 euros the total envronmental gan s equal to 3827 for an envronmental economc land planner who combnes strateges whle t s equal to 3198 f he only mplements a land sharng strategy. Thus the combnaton of strateges ncreases the envronmental gan of plannng wthout alterng the cost of ths plannng. Concluson Selecton of a land plannng strategy to reduce the rsk of pestcdes transfer from agrculture to surface and underground waterbodes has mportant mplcatons for the total envronmental gan and area of land covered that can be obtaned wth a gven budget. In ths paper two strateges are analyzed: a land sharng strategy subsdzng the adopton of IPM practces or a land sparng strategy purchasng land to put out of producton; these strateges can be used n solaton or combned n a polcy mx. A theoretcal approach brngs to the fore how dfferent types of land plannng objectves maxmzng surface under a budget constrant maxmzng benefts under a surface constrant or maxmzng benefts under a budget constrant mply dfferent selecton crtera. For nstance when the objectve s the maxmzaton of benefts under a budget constrant wth only one strategy avalable t s the rato of beneft to costs of each parcel that determned whether a parcel s selected; when land sparng and sharng can be combned t s also the rato of 16
the dfferences between benefts over the dfferences between costs that gudes the selecton of a parcel. Our emprcal analyss apples the above framework to the Sene Ource rver catchment n Burgundy (France) to analyze both the type of land plannng objectves and the possblty to mx strateges. Frst we show that when only one strategy can be mplemented the best opton for the Sene Ource rver catchment s the land sharng one. Then we proceed to the case where strateges can be combned: for our set of parameters the two strateges are mxed only when the land planner seeks to maxmze benefts under a cost constrant. For the same parcel the envronmental gans of the land sparng opton are always greater than that of the land sharng opton: a land planner lookng only at the benefts sde wll never chose to combne strateges even f gven the possblty to do so. The costs of the land sparng opton are also greater than that of the land sharng opton gven our calculaton of annual subsdes and our assumptons on contract length and dscount rate whch explans the lack of combnaton of strateges when the land planner seeks to maxmze the area of land protected under a gven budget constrant. An mmedate extenson of ths work would assess the senstvty of ths result to some parameters of the model. These prelmnary results put n perspectve the mportance of consderng the possblty to mplement a mx of strateges when comparng targetng optons for the preservaton of water from pestcdes polluton. They also provde a rankng procedure for each land plannng program whch would prove useful to mplement. 17
References Agence Régonal de Santé Bourgogne (ARS Bourgogne) (2009). Qualté des eaux destnées à la consommaton humane en Côte d Or et en régon Bourgogne. Ando A. J. Camm S. Polasky A. Solow (1998) Speces dstrbutons land values and effcent conservaton Scence 279 2126 28 Babcock B.A. P. Lakshmnarayan J. Wu D. Zlberman (1996) The economcs of a publc fund for envronmental amentes: A study of CRP contracts Amercan Journal of Agrcultural Economcs 78 961 971 Babcock B.A. P. Lakshmnarayan J. Wu D. Zlberman (1997) Targetng tools for the purchase of envronmental amentes Land Economcs 73 325 339 Bastan C.T. D.M. McLeod M.J. Germno W.A. Reners B.J. Blasko (2002) Envronmental amentes and agrcultural land values: a hedonc model usng geographc nformaton systems data Ecologcal Economcs 40 337 349 Bockstaller C.Wohlfahrt J. Huber A. Hennebert P. Zahm F. Verner F. Kechnger O. Grardn P. (2008) Les ndcateurs de rsque de transfert de produts phytosantares et leur valdaton: exemple de l ndcateur I PHY. Ingeneres no.86 103 114. Bonon J.P. (2003) Land polcy n France and ts consequences for the farmers Agrcultural Economcs Czech 4 166 172 Boussemart J.P. Leleu H. Ojo O. (2011) Could socety s wllngness to reduce pestcde use be algned wth farmers economc self nterest? Ecologcal Economcs 70(10) 1979 1804. Coner C. J. Wu S. Polasky (2001) Economc and envronmental mplcatons of alternatve landscape desgns n the Walnut Creek Watershed of Iowa Ecologcal Economcs 38 119 139 Fortno G. Reau R. (2010) Two separate steps for croppng system assessment: charactersaton and fnal evaluaton. In: J. Wery I. Shl Touz A. Pern (Eds.) Agro2010 the XIth ESA Congress Agropols Internatonal Montpeller France August 29th to September 3rd 2010 Fscher J. D.J. Abson V. Butsc M.J. Chappell J. Ekroos J. Hanspach T. Kuemmerle H.G. Smth H. Von Wehrden (2014) Land sparng versus land sharng: movng forward Conservaton Letters 7(3) 149 157 Green P.E. S.J. Cornell J.P.W. Scharlemann A. Balmford (2005) Farmng and the Fate of Wld Nature Scence 307 550 555 Hascc I. J. Wu (2006) Land use and watershed health n the Unted States Land Economcs 82(2) 214 239 Hlal M. (2010) Odomatrx. Calcul de dstances routères ntercommunales. Caher des Technques de l INRA Numéro spécal : Méthodes et outls de tratement des données en scences socales. Retours d expérences 41 63 18
Jacquet F. Butault J.P. Guchard L. (2011) An economc analyss of the possblty of reducng pestcdes n French feld crop Ecologcal Economcs 70(9) 1638 48. Kellogg R. L. M. Margaret D. Goss (1992) Agrcultural Chemcal Use and Groundwater Qualty: Where Are the Potental Problem Areas? Sol Conservaton Servce USDA Langpap C. I. Hascc J. Wu (2008) Protectng watershed ecosystems through targeted local land use polces Amercan Journal of Agrcultural Economcs 90(3) 684 700 Lechenet M. V. Bretagnolle C. Bockstaller F. Bossnot M.S. Pett S. Pett N. Muner Jolan (2014) Reconclng pestcde reducton wth economc and envronmental sustanablty n arable farmng PLOSone 9(6) 1 10 Margules C. A. Ncholls R. Pressey (1988) Selectng networks of reserves to maxmze bologcal dversty Bologcal conservaton 43 63 76 Meulenberg M.T.G. (1965) On the estmaton of an exponental functon Econometrca 33(4) 863 868 Newburn D.A. P. Berck A.M. Merenlender (2006) Habtat and open space at rsk of land use converson: targetng strateges for land conservaton Amercan Journal of Agrcultural Economcs 88(1) 28 42 Ongley E.D. (1996) Control of water polluton from agrculture FAO rrgaton and dranage paper 55 Palmqust R.B. (2005) Property value models chap. 16 of Handbook of Envronmental Economcs Volume 2. Edted by K. G. Mäler and J.R. Vncent Pmentel D. H. Acquay M. Bltonen P. Rce M. Slva J. Nelson V. Lpner S. Gordano A. Horowtz M. D Amore (1992) Envronmental and economc costs of pestcde use BoScence 42 750 760 Sexton S.E. Z. Le D. Zlberman (2007) The economcs of pestcdes and pest control Internatonal Revew of Envronmental and Resource Economcs 1 271 326 Stern V.M. R.F. Smth R. van den Bosch K.S. Hagen (1959) The ntegrated control concept Hlagarda 29 81 101 Van der Werf H. C. Zmmer (1998) An ndcator of pestcde envronmental mpact based on a fuzzy expert system Chemosphere 36 2225 2249 Wlson C. Tsdell C. (2001) Why farmers contnue to use pestcdes despte envronmental health and sustanablty costs Ecologcal Economcs 39(3) 449 462. Wu J. W.G. Boggess (1999) The optmal allocaton of conservaton funds Journal of Envronmental Economcs and Management 38 302 321 19
Appendx 1: mean producton costs for each typcal conventonal and IPM producton systems analysed. The producton systems are defned as follows : PS1 : rapeseed wheat barley PS2 : rapeseed wheat slage maze wheat o wth no tllage (a) systematc tllage (b) or tllage on rapeseed and maze and no tllage on wheat and barley (c) IPM1 : rapeseed wheat ntermedary crop (whte mustard) pea/barley mx wnter barley ntermedary crop (oat) sunflower wheat. IPM2 : rapeseed wheat ntermedary crop (whte mustard) pea/barley mx wnter barley ntermedary crop (oat) slage maze wheat. Fgure 1 presents the mean producton costs detaled by expendture tem for the 6 conventonal and 2 IPM producton systems analysed n the paper. Fgure 1: mean producton costs for conventonal and IPM producton systems 20
Appendx 2: senstvty analyss of results. Table 4: Senstvty analyss of results wthout the possblty of combnng sparng and sharng strateges. Cost Envronmental gans Number of parcels Surface Mean sze of parcels Sparng Sharng Sparng Sharng Sparng Sharng Sparng Sharng Sparng Sharng ha ha A1 maxmsaton 50000 50000 5 176 1 99 42 316 42.08 3.19 100000 100000 30 326 4 185 83 631 20.87 3.41 150000 150000 76 536 9 306 121 947 13.43 3.09 200000 200000 138 631 15 353 157 1140 10.47 3.23 250000 250000 246 717 27 382 191 1235 7.07 3.23 300000 300000 276 792 31 407 223 1329 7.18 3.27 350000 350000 334 920 37 447 254 1424 6.86 3.19 400000 400000 391 967 43 462 284 1519 6.60 3.29 450000 450000 424 1064 47 493 313 1613 6.66 3.27 500000 500000 491 1139 54 518 341 1708 6.32 3.30 A2 maxmsaton (A1 surface) 270628 156729 3599 2732 398 923 42 316 0.11 0.34 421890 312881 4934 3393 543 1151 83 631 0.15 0.55 540115 473336 5809 3811 640 1293 121 947 0.19 0.73 651346 569917 6481 4020 714 1364 157 1140 0.22 0.84 769766 615585 7017 4113 774 1396 191 1235 0.25 0.88 876444 660551 7452 4200 822 1428 223 1329 0.27 0.93 980152 708765 7829 4282 865 1454 254 1424 0.29 0.98 1068448 755035 8143 4359 900 1481 284 1519 0.32 1.03 1152778 802665 8425 4434 931 1506 313 1613 0.34 1.07 1234886 845553 8675 4505 958 1532 341 1708 0.36 1.11 A2 maxmsaton (A3 surface) 88686 57983 1680 1855 186 619 9 116 0.05 0.19 157742 114415 2476 2441 274 823 19 231 0.07 0.28 218138 173100 3105 2827 343 956 31 348 0.09 0.36 277261 234231 3657 3116 404 1057 44 472 0.11 0.45 332700 305612 4181 3371 462 1144 58 617 0.13 0.54 377744 368665 4579 3553 505 1204 71 740 0.14 0.61 426266 430105 4956 3713 546 1259 84 864 0.15 0.69 470016 489863 5308 3850 584 1306 98 981 0.17 0.75 509062 550882 5589 3982 615 1351 111 1102 0.18 0.82 559127 609373 5940 4099 655 1390 127 1219 0.19 0.88 A3 maxmsaton 50000 50000 1236 1782 136 618 9 116 0.06 0.19 100000 100000 1989 2359 218 814 19 231 0.09 0.28 150000 150000 2632 2734 289 950 31 348 0.11 0.37 200000 200000 3198 3012 350 1047 44 472 0.12 0.45 250000 250000 3708 3232 406 1128 58 617 0.14 0.55 300000 300000 4175 3411 457 1188 71 740 0.15 0.62 350000 350000 4607 3567 506 1243 84 864 0.17 0.69 400000 400000 5007 3708 550 1293 98 981 0.18 0.76 450000 450000 5372 3836 590 1337 111 1102 0.19 0.82 500000 500000 5711 3952 627 1377 127 1219 0.20 0.89 21