Researchers Develop Algorithm to Predict Harvest Potential from Climate Data

Summary

An algo­rithm devel­oped by researchers in Italy and the US can fore­cast an olive grove’s behav­ior and pro­duc­tiv­ity by ana­lyz­ing cli­mate impacts on olive devel­op­ment and har­vests. The algo­rithm, which is pub­licly acces­si­ble, can help stake­hold­ers pre­dict upcom­ing sea­sons and make nec­es­sary adjust­ments in agro­nomic prac­tices based on iden­ti­fied cli­mate stres­sors. The method can be applied beyond Italy to other coun­tries and sec­tors to pro­vide use­ful sea­sonal fore­casts.

Almost one hun­dred pro­duc­tion stake­hold­ers have down­loaded an algo­rithm that may pro­vide the abil­ity to fore­cast an olive grove’s behav­ior and pro­duc­tiv­ity.

The new tech­nol­ogy is based on a thor­ough analy­sis of sea­sonal weather pat­terns dur­ing the olive growth cycle over a long period in Italy.

By com­par­ing the rela­tion­ship between olive devel­op­ment and har­vests to the cli­mate impacts, researchers were able to iden­tify dozens of poten­tial cli­mate stres­sors and how they affect olive tree pro­duc­tiv­ity.

Researchers believe this infor­ma­tion might sup­port national or regional admin­is­tra­tions, olive grow­ers, pro­duc­ers, and other inter­ested par­ties in pre­dict­ing how an upcom­ing sea­son may unfold and mak­ing any agro­nomic or busi­ness adjust­ments.

The new tech­nol­ogy results from a coor­di­nated project involv­ing sci­en­tists from the Italian National Research Council (CNR) and the Agency for New Technologies, Energy and Sustainable Development (ENEA) as well as American researchers from the University of California – Berkeley.

“We are work­ing on under­stand­ing which [cli­matic] dri­vers can trig­ger unfa­vor­able con­di­tions and the asso­ci­ated prob­a­bil­ity of expe­ri­enc­ing detri­men­tal effects on olive pro­duc­tion,” Arianna Di Paola, a researcher at the Italian Institute for BioEconomy at the CNR, told Olive Oil Times.

“Examples of trig­gers are con­di­tions which favor the spread of the olive fruit fly or high win­ter tem­per­a­tures which can alter the olive cycle and impact flow­er­ing and pol­li­na­tion,” she added.

The research ana­lyzed olive har­vests in 66 Italian provinces between 2006 and 2020 to iden­tify the stres­sors using a wide range of data. They were able to uncover how the worst olive har­vests came to be.

“Understanding the ongo­ing sea­son­al­ity allows us to fore­see what we may expect in the near future,” Di Paola said.

“These are not sea­sonal fore­casts, which are required to be reli­able and trans­lated into action­able infor­ma­tion to facil­i­tate the deci­sion-mak­ing process, a whole world of research in itself,” she added. ​“They are short-term sce­nar­ios that might sup­port invest­ments, pre­ven­tive mea­sures, treat­ments or agro­nomic prac­tices.”

The research did not stop at iden­ti­fy­ing the dri­vers of unfa­vor­able con­di­tions.

“While we can­not yet pre­dict the whole phe­no­log­i­cal cycle of the olive, as it is not pos­si­ble to pre­dict the veg­e­ta­tive onset in the sea­son on regional scales, what we can do is, using a cal­en­dar, sim­ply divide the life­cy­cle of the olive into two-month install­ments,” Di Paola said.

By ana­lyz­ing the vari­ables impact­ing olive pro­duc­tion through the years and aggre­gat­ing them every two months, researchers defined a list of the vari­ables and exam­ined how they inter­act over time.

The analy­sis pro­vides a short-term pre­ci­sion fore­cast, which researchers said is three times bet­ter than the analy­sis of a sin­gle vari­able.

“For exam­ple, one thing is to say that we had a warmer win­ter, another is to say that fol­low­ing that warm win­ter, we also had a very wet sum­mer, fac­tors which can add up and fur­ther worsen the sce­nario,” Di Paola said.

Once the analy­sis was ready, the researchers looked at which sea­sonal cli­mate vari­ables were more often asso­ci­ated with extremely bad or high-yield sea­sons, dis­card­ing the mid­dle-range yields.

This selec­tion aimed to iden­tify yields that, on a broad spa­tial scale, were most affected by cli­matic vari­abil­ity given the super­im­po­si­tion of other dri­vers.

“In mid­dle-range sea­sons, the yields might depend on vari­ables such as deploy­ing spe­cific agro­nomic tech­niques by one grower com­pared to another, or to the time spent prun­ing the olives and many more vari­ables,” Di Paola said.

Therefore, researchers were more inter­ested in look­ing at both plen­ti­ful and scarce extreme sea­sons, as the asso­ci­ated con­di­tions had an impact inde­pen­dent of the actions of the sin­gle grower.

“Most of us are used to focus­ing on sin­gle stress fac­tors, such as a freeze or heat­wave, but even if we should man­age to look at those sin­gle stress fac­tors cor­rectly, we would still not be able to asso­ciate them to a spe­cific phe­no­log­i­cal stage with­out proper field obser­va­tions or model sim­u­la­tions,” Di Paola said.

“We tried to smooth out all of these effects to con­sider them together on a large scale and across whole sea­sons,” she added.

Interestingly, the researchers found a link between the cli­mate vari­ables iden­ti­fied by the algo­rithm and the olive fruit fly phe­nom­e­non.

“The algo­rithm will not tell you why a spe­cific sce­nario is going to occur,” Di Paolo said. ​“However, by apply­ing it, we note that the out­puts – worse years in terms of pro­duc­tiv­ity and emerg­ing cli­mate stres­sors – were plau­si­bly asso­ci­ated with olive fruit fly infes­ta­tions.”

“What the algo­rithm tells us is some­thing like: should you have these array of con­di­tions, let’s say five dif­fer­ent vari­ables over a given time, then it is highly prob­a­ble that the olive yield will be excep­tion­ally low,” she added.

Once this warn­ing comes from the algo­rithm, an expert must look at the data to inter­pret it cor­rectly. ​“Is it the olive fruit fly, or are there other fac­tors we should con­sider?” Di Paola noted.

“We stan­dard­ized all the vari­ables to make them com­pa­ra­ble across time and space, and that allowed us to look at things from above,” she added. ​“To make it clear, when the research says that a spe­cific trig­ger is a warmer period than aver­age, that was true across all the provinces in the coun­try.”

By explor­ing a wide range of the ter­ri­tory, the algo­rith­m’s gen­er­al­iza­tion increases, and bet­ter fore­casts for the whole sec­tor in the entire coun­try can be achieved.

“This is a use­ful view of the whole sec­tor for all enti­ties inter­ested in look­ing at the full pic­ture,” Di Paola said.

The algo­rithm, which is pub­licly acces­si­ble and can be down­loaded and inte­grated into their sys­tems, might be help­ful not only for Italy but also for the olive sec­tor.

“The method we applied can be exported to other coun­tries and sec­tors,” Di Paola con­cluded. ​“Once fed with the needed data, the algo­rithm can eas­ily be adapted to make that kind of sea­sonal fore­cast.”

• ENEA