Expected global suitability of coffee, cashew and avocado due to climate change.

Coffee, cashew and avocado are of high socio-economic importance in many tropical smallholder farming systems around the globe. As plantation crops with a long lifespan, their cultivation requires long-term planning. The evaluation of climate change impacts on their biophysical suitability is therefore essential for developing adaptation measures and selecting appropriate varieties or crops. In this study, we modelled the current and future suitability of coffee arabica, cashew and avocado on a global scale based on climatic and soil requirements of the three crops. We used climate outputs of 14 global circulation models based on three emission scenarios to model the future (2050) climate change impacts on the crops both globally and in the main producing countries. For all three crops, climatic factors, mainly long dry seasons, mean temperatures (high and low), low minimum temperatures and annual precipitation (high and low), were more restrictive for the global extent of suitable growing regions than land and soil parameters, which were primarily low soil pH, unfavourable soil texture and steep slopes. We found shifts in suitable growing regions due to climate change with both regions of future expansion and contraction for all crops investigated. Coffee proved to be most vulnerable, with negative climate impacts dominating in all main producing regions. For both cashew and avocado, areas suitable for cultivation are expected to expand globally while in most main producing countries, the areas of highest suitability may decrease. The study reveals that climate change adaptation will be necessary in most major producing regions of all three crops. At high latitudes and high altitudes, however, they may all profit from increasing minimum temperatures. The study presents the first global assessment of climate change impacts on cashew and avocado suitability.

Introduction

Plantation crops such as coffee, cashew and avocado are among the most important cash crops and contribute substantially to the livelihoods of smallholder farmers around the world. These crops have a lifespan of several decades and therefore long-term agricultural planning considering the expected impacts of climate change is especially important. Global warming of 1.2 up to 3.0°C by 2050 is estimated by the Intergovernmental Panel on Climate Change depending on different greenhouse gas emission pathways [1]. Such changes in temperature will directly affect the climate suitability of growing regions for crops and can therefore cause shifts in production regions or call for adaptation measures in agricultural management, such as more heat or drought tolerant varieties. However, detailed spatial analyses are required to assess the expected global and regional impacts considering both changes in temperature and precipitation patterns. For coffee arabica, a crop highly sensitive to climate change, current and future climate suitability has been studied extensively on global [2, 3] and regional scales [4-11] that have recently been reviewed by Pham et al. [12]. According to these studies, strong reductions in climate suitability are expected for coffee in most current growing regions. In only a few regions, mainly at higher elevations or latitudes, might coffee cultivation profit from climate change. None of these studies, however, have taken into account land and soil characteristics, such as slope, soil pH or texture, in their suitability evaluations. Other tropical perennial cash crops of high socio-economic importance in their main producing countries have received much less attention. For both cashew and avocado, no global assessment of current and future suitability is available. Few land suitability evaluations were undertaken for cashew [13-17], while only one study modelled climate change impacts on cashew suitability in Côte d’Ivoire and Ghana where positive impacts were found [14]. In the case of avocado, a comprehensive assessment of current and future distributions across the Americas was made by Ramírez-Gil et al. [18], including some of the major producing countries. They identified both regions of future expansion and contraction. To our knowledge, only two studies investigated avocado suitability outside these continents [19, 20]. Global biophysical modelling of current and future suitability of coffee, cashew and avocado is therefore essential to take informed decisions in long-term agricultural planning with the aim of maintaining farmers’ livelihoods and of fostering the sustainable use of natural resources.

The objective of this paper is to estimate current and future biophysical suitability for coffee arabica, cashew and avocado production on a global scale and to identify and discuss global and regional trends. We modelled crop suitability based on their biophysical requirements. For the first time, both land and soil (artificial surfaces, protected areas, soil texture, coarse fragments, pH, organic carbon content, salinity) and climate (temperature, precipitation, humidity) parameters were taken into account on a global scale. Climate change projections were based on 14 global circulation models (GCMs) and three representative concentration pathways (RCP 2.6, 4.5 and 8.5) for the year 2050. We investigated climate change impacts (relative decreases and increases in crop suitability) both globally and in the main producing countries of the three plantation crops.

Materials and methods

The GIS-based decision support system CONSUS [21] was used to model the biophysical suitability of coffee (Coffea arabica L.), cashew (Anacardium occidentale L.) and avocado (Persea americana Mill.). The model is based on multi criteria evaluation of biophysical variables and was applied on a global scale for current and future climatic conditions. The land suitability evaluation in CONSUS consists of four steps [21]: niche description, site description, matching and aggregation. In the niche description, the biophysical crop requirements (climate, land and soil parameters such as temperature, precipitation or soil pH) are first identified based on literature search. Then, each parameter is classified into four suitability classes (see Table 1) following the FAO land evaluation approach [22]: S1 (highly suitable), S2 (moderately suitable), S3 (marginally suitable), N (unsuitable). For coffee for example, a soil pH between 4.5 and 5.0 is classified as marginally suitable (Table 1). For the site description, spatial data corresponding to the crop-specific requirements is identified. In this study, publicly available global raster datasets with a resolution of 30 arc seconds were used (see Table 2). The matching of the crop requirements with the spatial data is done for every parameter individually (e.g. soil pH). During this process, each raster cell of the corresponding dataset is reclassified into one of the four suitability values. In the case of coffee, for example, a soil pH of 4.7 (Table 1) of a certain land unit is classified as marginally suitable (S3). This matching of crop requirements with spatial data results in separate suitability maps for each parameter investigated, for example the soil pH suitability map for coffee. The resulting maps are aggregated by the maximum limiting factor [23], rating the suitability of each raster cell by the factor with the lowest value. If the suitability of all crop requirements other than soil pH in the above-mentioned example were rated as S3 or higher, the overall suitability would therefore be S3. The resulting global maps show the potential land suitability of the studied crop under rainfed conditions, without taking into account agricultural management options such as liming or irrigation. The results are based on growth factors that are sufficiently described for the respective crop and where corresponding global spatial datasets are available. Suitability evaluations were undertaken for current and predicted future climatic conditions.

Table 1

Biophysical requirements of coffee (Coffea arabica L.), cashew (Anacadrium occidentale L.) and avocado (Persea americana Mill.) used in the model, classified into four suitability classes (S1: Highly suitable, S2: Moderately suitable, S3: Marginally suitable, N: Unsuitable).

The classification was done based on several sources for coffee [24–28], cashew [13, 16, 24, 25]and avocado [24, 25, 29, 30].

	Coffee				Cashew				Avocado
Criteria	S1	S2	S3	N	S1	S2	S3	N	S1	S2	S3	N
Climate
Mean annual temperature (°C)	17–22	22–2515–17	25–2812–15	>28<12	24–28	28–3120–24	31–3415–20	>34<15	18–26	26–3015–18	30–4510–15	>45<10
Mean minimum temperature of coldest month (°C)	10–19	19–217–10	21–234–7	>23<4	>10	8–10	4–8	<4	>16	13–16	8–13	<8
Mean annual precipitation (mm)	1400–1800	1800–23001000–1400	2300–4200750–1000	>4200<750	1000–2250	2250–3200800–1000	3200–4500500–800	>4500<500	1200–1800	1800–20001200–1000	2000–2500750–1000	>2500<750
Length of dry season (months)	1–4	4–50–1	5–6-	>6	0–4	4–5	5–6	>6	1–4	4–6<1	>6	-
Mean relative humidity of driest month (%)	40–70	70–8030–40	80–9020–30	>90<20	>30	25–30	20–25	<20	-	-	-	-
Land and Soil
Artificial surfaces (type)a	0,2,3,4,5,6,7,8,9,10,11	-	-	1	0,2,3,4,5,6,7,8,9,10,11	-	-	1	0,2,3,4,5,6,7,8,9,10,11	-	-	1
Protected areas (category)b	-	-	-	1,2,3,4,5,6,7,8,9,10	-	-	-	1,2,3,4,5,6,7,8,9,10	-	-	-	1,2,3,4,5,6,7,8,9,10
Slope (%)	0–8	8–16	16–30	>30	0–8	8–16	16–30	>30	0–8	8–16	16–30	>30
Soil texture (USDA class)c	1,3,4,5,7,8,10	6	9	2,11,12	1,2,3,4,5,6,7,8,9,10,11	-	12	-	4,5,6,7,8,9,10,11,12	2,3	1	-
Coarse fragments (vol%)	0–15	15–35	35–55	>55	0–15	15–35	35–55	>55	0–15	15–35	35–55	>55
Soil organic carbon (%)	>1.2	0.8–1.2	<0.8	-	>0.8	0.5–0.8	0.1–0.5	<0.1	>1.2	0.8–1.2	<0.8	-
Soil pH	5.5–6.5	6.5–7.05.0–5.5	7.0–7.54.5–5.0	>7.5<4.5	5.2–7.0	7.0–7.54.8–5.2	7.5–8.04.5–4.8	>8.0<4.5	5–6.5	6.5–7.54.5–5	7.5–8.34.3–4.5	>8.3<4.3
Soil salinity (ECe)	0–0.5	0.5–1.5	1.5–2.5	>2.5	0–2	2–3	3–4	>4	0–3	3–4	4–5	>5

aArtificial surfaces were classified according to FAO global land cover SHARE ([31]; 1 = artificial surfaces, 2 = cropland, 3 = grassland, 4 = tree covered areas, 5 = shrubs covered areas, 6 = herbaceous vegetation, aquatic or regularly flooded, 7 = mangroves, 8 = sparse vegetation, 9 = bare soil, 10 = snow and glaciers, 11 = water bodies).

bProtected areas were classified according to the IUCN management categories ([32]; 1 = Ia, 2 = Ib, 3 = II, 4 = III, 5 = IV, 6 = V, 7 = VI, 8 = not reported, 9 = not applicable, 10 = not assigned).

cSoil texture was classified according to USDA soil taxonomy (1 = clay, 2 = silty clay, 3 = sandy clay, 4 = clay loam, 5 = silty clay loam, 6 = sandy clay loam, 7 = loam, 8 = silty loam, 9 = sandy loam, 10 = silt, 11 = loamy sand, 12 = sand).

Table 2

Data sources used for the modelling of the crops’ climate, land and soil suitability [31–38].

All datasets are in a raster format and have a resolution of 30 arc seconds.

Criteria	Data source	URL	Reference
Climate
Mean annual temperature (°C), current (1970–2000) and future (2041–2060)	WorldClim: Global climate and weather data, version 2.0 (BIO 1)	www.worldclim.org	[33]
Mean minimum temperature of coldest month (°C), current (1970–2000) and future (2041–2060)	WorldClim: Global climate and weather data, version 2.0 (BIO 6)	www.worldclim.org	[33]
Mean annual precipiataion (mm), current (1970–2000) and future (2041–2060)	WorldClim: Global climate and weather data, version 2.0 (BIO 12)	www.worldclim.org	[33]
Length of dry season (months)	WorldClim: Global climate and weather data, version 2.0 (monthly precipitation)Global aridity and PET database (potential evapotranspiration)	www.worldclim.orgcgiarcsi.community/data/global-aridity-and-pet-database	[33,37,38]
Mean relative humidity of dryest month (%)	CliMond: Global climatologies for bioclimatic modelling (relative humidity at 9 am)	www.climond.org/ClimateData.aspx	[36]
Land and Soil
Artificial surfaces (type)	FAO Global Land Cover (GLC-SHARE) Beta-Release 1.0 Database	www.fao.org/geonetwork/	[31]
Protected areas (category)	The World Databae on Protected Areas (WDPA)	www.protectedplanet.net	[32]
Slope (%)	Derived from ESRI Terrain Service
Soil texture (USDA class)	SoilGrids—global griddes soil information (soil texture at 15 cm depth)	www.isric.org/explore/soilgrids	[35]
Coarse fragments (vol%)	SoilGrids—global griddes soil information (coarse fragments at 15 cm depth)	www.isric.org/explore/soilgrids	[35]
Soil organic carbon (%)	SoilGrids—global griddes soil information (soil organic carbon at 15 cm depth)	www.isric.org/explore/soilgrids	[35]
Soil pH	SoilGrids—global griddes soil information (soil pH at 15 cm depth)	www.isric.org/explore/soilgrids	[35]
Soil salinity (ECe)	Harmonized Soil Database, version 1.2	www.fao.org/soils-portal/data-hub/soil-maps-and-databases/harmonized-world-soil-database-v12/	[34]

Current suitability modelling

For the modelling of the current global land suitability for the three plantation crops, their biophysical requirements were identified via a literature search. Table 1 lists their climatic, land and soil requirements that were used in the model, classified into the four suitability classes. This classification was done based on Sys et al. [24] taking into account additional scientific literature (see Table 1). Whenever several authors defined different thresholds between suitability classes of a criterion, the threshold was set conservatively resulting in a higher crop suitability or it was set in agreement with the majority of indications. Global climate and land/soil raster data with a spatial resolution of 30 arc seconds were used from established global climate models and satellite-data based global land use datasets and are listed in Table 2. Artificial surfaces according to the FAO Global Land Cover SHARE dataset [31] and protected areas based on the World Database on Protected Areas [32] were rated as ‘unsuitable’ (see Table 1). The length of the dry season was based on the monthly precipitation (P) and the potential evapotranspiration (PET), and was calculated as the number of months where P < ½ PET [24].

Coffee, cashew and avocado have similar ecological niches (Table 1). However, there are still important differences in their biophysical requirements. While avocado has the highest suitable temperature range, coffee is most susceptible to high temperatures. At the same time, avocado is more susceptible to low temperatures than the other crops during the coldest month. With regard to precipitation, cashew has the highest suitable range, tolerating both higher and lower values than coffee and avocado. Avocado is most susceptible to high precipitation, but more tolerant to longer dry seasons. Both avocado and coffee have a slightly reduced suitability in regions without a dry season. When looking at edaphic factors, coffee has a narrower ecological niche than the other crops. Compared to coffee, both cashew and avocado are more tolerant to high and low soil pH and less restricted regarding soil texture. Combining all these factors, cashew has the broadest ecological niche, followed by avocado and coffee.

For each criterion listed in Table 1, the crop suitability was calculated globally. The results for all five climate requirements and for the eight land and soil requirements were then aggregated by the lowest suitability value to obtain the overall climate suitability and land and soil suitability, respectively [22]. Finally, the climate and the land and soil suitability were combined once again by the lower rate, resulting in the overall current suitability. For the identification and discussion of most limiting factors in the different regions, the individual suitability maps for the different requirements were compared. All analyses were done using the ESRI ArcGIS Pro software version 2.5.0.

Future suitability modelling

To estimate the potential effects of climate change on future suitability, the impacts were modelled based on three representative concentration pathways (RCPs; [39]) for the future climatic conditions of 2050 (average of projected climate from 2041–2060). Downscaled CMIP5 data [40] of 14 global climate models (GCMs: BBC-CSM1-1, CCSM4, CNRM-CM5, GFDL-CM3, GISS-E2-R, HadGEM2-AO, HadGEM2-ES, IPSL-CM5A-LR, MIROC-ESM-CHEM, MIROC-ESM, MIROC5, MPI-ESM-LR, MRI-CGCM3 and NorESM1-M) were used for RCP 2.6 (low emissions), 4.5 (intermediate emissions) and 8.5 (high emissions), available in 30 arc second resolution in WorldClim version 2.0 (www.worldclim.org). Three relevant bioclimatic variables (BIO 1, 6 and 12) were retrieved from the data of all 14 GCMs by calculating means (see Table 2). These three variables were used for the modelling of future crop suitability in this study. All other parameters listed in Table 1 were kept constant. The overall future climate suitability was then calculated using the future projections of the three bioclimatic variables described above. For the overall future suitability, the climate and the land and soil suitability were aggregated again by the maximum limiting factor. All the calculations were done for RCP 2.6, 4.5 and 8.5 separately, using the ESRI ArcGIS Pro software version 2.5.0.

Calculation of expected changes

To visualize the projected overall suitability change between 2000 and 2050, the difference between the overall suitability of 2000 and 2050 was calculated for each raster cell. The resulting values between -3 and +3 indicate by how many suitability classes the different regions are expected to decrease or increase (e.g. a value of -2 indicating a negative change from S1 highly suitable to S3 moderately suitable).

Additionally, the changes in suitability by 2050 for the three different RCPs were calculated individually for each crop for the four main producing countries. The main producing countries were identified using FAOSTAT (www.fao.org/faostat) data concerning the quantity produced in 2018. For coffee, these are Brazil, Vietnam, Indonesia and Colombia with a share of 64% of the global production. For cashew, Vietnam, India, Côte d’Ivoire and Benin (Benin was chosen instead of the Philippines because of very similar quantities produced but a much bigger area harvested in 2018) were identified, representing 73% of the global production. Finally, the main avocado producing countries are Mexico, the Dominican Republic, Peru and Indonesia, accounting for 58% of the global production. For these countries, the changes in suitable areas were calculated based on the area per suitability class that was calculated after equivalent projection to the Equal Earth coordinate system in ESRI ArcGIS Pro version 2.5.0.

Results

An overview of the results relevant for coffee, cashew and avocado is first presented. Subsequently, the current and future suitability and the expected changes are described individually for each of the three crops.

Factors determining suitability

The modelled suitable areas of all three plantation crops show a similar global extent due to their similar biophysical requirements (Figs 1–3). Globally, cashew has the largest suitable growing areas globally, followed by avocado and coffee (see Tables 3–5). However, as far as the most limiting factors and the extent of highly (S1) and moderately (S2) suitable areas are concerned, there are important differences between the crops. For all three, the modelled climate suitability is more restrictive than the land and soil suitability for the global growing regions. All climatic factors used in the models (see Table 1) are important suitability limiting factors in several regions, except for the relative humidity of the driest month. For coffee, humidity is a co-limiting factor in very few areas in Central Africa and Southeast Asia, while it is no requirement in the avocado model and does not limit cashew suitability at all.

Fig 1

Overall current suitability for coffee (aggregated climate, land and soil suitability).

A Central and South America, B West and Central Africa, C South and Southeast Asia.

Fig 3

Overall current suitability for avocado (aggregated climate, land and soil suitability).

A Central and South America, B West and Central Africa, C South and Southeast Asia.

Table 3

Suitable coffee growing areas globally and in main producing countries (S1: highly suitable, S2: Moderately suitable, S3: Marginally suitable, N: Unsuitable) for current (2000) and future (2050) conditions under three RCPs: 2.6 (low emissions), 4.5 (intermediate emissions), 8.5 (high emissions).

Expected changes in suitable areas are given as a percentage.

		RCP 2.6		RCP 4.5		RCP 8.5
Suit Class	2000 (km2)	2050 (km2)	Δ (%)	2050 (km2)	Δ (%)	2050 (km2)	Δ (%)
	World total
S1	36,240	16,540	‒54.4	16,777	‒53.7	14,678	‒59.5
S2	5,709,608	3,951,207	‒30.8	3,679,863	‒35.5	3,369,550	‒41.0
S3	14,709,645	15,118,407	2.8	13,995,976	‒4.9	12,787,405	‒13.1
N	104,044,240	105,413,581	1.3	106,807,118	2.7	108,328,100	4.1
	Brazil
S1	5,934	1,421	‒76.1	1,268	‒78.6	161	‒97.3
S2	1,822,032	1,311,548	‒28.0	1,161,921	‒36.2	1,040,958	‒42.9
S3	2,430,089	2,536,454	4.4	2,427,693	‒0.1	1,939,711	‒20.2
N	4,099,828	4,508,459	10.0	4,767,001	16.3	5,377,052	31.2
	Vietnam
S1	683	358	‒47.6	196	‒71.3	99	‒85.5
S2	141,637	106,814	‒24.6	89859	‒36.6	75,422	‒46.7
S3	108,773	143,838	32.2	146498	34.7	149,801	37.7
N	68,291	68,373	0.1	82829	21.3	94,060	37.7
	Indonesia
S1	0	0	0.0	0	0.0	0	0.0
S2	42,862	35,247	‒17.8	26,828	‒37.4	20,914	‒51.2
S3	1,391,935	1,191,058	‒14.4	922,242	‒33.7	698,400	‒49.8
N	387,893	596,385	53.7	873,620	125.2	1,103,376	184.5
	Colombia
S1	332	123	‒63.0	108	‒67.5	83	‒75.0
S2	75,494	55,729	‒26.2	50,886	‒32.6	47,650	‒36.9
S3	574,239	375,858	‒34.5	273,066	‒52.4	224,152	‒61.0
N	473,101	691,455	46.2	799,106	68.9	851,282	79.9

Table 5

Suitable avocado growing areas globally and in main producing countries (S1: Highly suitable, S2: Moderately suitable, S3: Marginally suitable, N: Unsuitable) for current (2000) and future (2050) conditions under three RCPs: 2.6 (low emissions), 4.5 (intermediate emissions), 8.5 (high emissions).

Expected changes in suitable areas are given as a percentage.

		RCP 2.6		RCP 4.5		RCP 8.5
Suit Class	2000 (km2)	2050 (km2)	Δ (%)	2050 (km2)	Δ (%)	2050 (km2)	Δ (%)
	World total
S1	790,882	682,815	‒13.7	626,928	‒20.7	464,091	‒41.3
S2	8,150,219	9,121,096	11.9	9,447,321	15.9	9,763,580	19.8
S3	12,503,581	13,248,676	6.0	13,234,520	5.8	13,497,956	8.0
N	103,062,026	101,454,122	‒1.6	101,197,940	‒1.8	100,781,081	‒2.2
	Mexico
S1	829	1,548	86.7	1,420	71.3	1,373	65.6
S2	93,712	107,578	14.8	111,375	18.8	112,949	20.5
S3	416,740	411,218	‒1.3	405,629	‒2.7	406,587	‒2.4
N	1,417,134	1,408,071	‒0.6	1,409,991	‒0.5	1,407,506	‒0.7
	Dominican Republic
S1	3,964	1,632	‒58.8	1,194	‒69.9	596	-85.0
S2	15,333	19,644	28.1	20,221	31.9	20,275	32.2
S3	15,754	14,111	‒10.4	13,314	‒15.5	13,320	-15.5
N	11,807	11,470	-2.9	12,128	2.7	12,667	7.3
	Peru
S1	11,814	5,399	-54.3	4,043	‒65.8	2,817	‒76.2
S2	92,971	91,394	‒1.7	93,042	0.1	95,393	2.6
S3	162,120	164,745	1.6	155,237	‒4.2	156,678	‒3.4
N	1,005,178	1,010,545	0.5	1,019,761	1.5	1,017,195	1.2
	Indonesia
S1	8,286	5,002	‒39.6	3,607	‒56.5	2,874	‒65.3
S2	153,705	156,022	1.5	144,982	‒5.7	143,972	‒6.3
S3	422,596	440,551	4.2	372,775	‒11.8	370,961	‒12.2
N	1,242,637	1,225,648	‒1.4	1,305,859	5.1	1,309,415	5.4

Overall, four of the biophysical land and soil requirements (artificial surfaces, protected areas, slope and coarse fragments) are not crop-specific, and therefore the suitability maps for these parameters are identical for all three crops (see Table 1). However, whether these parameters are limiting factors in the overall suitability still differs between the three crops depending on their other biophysical requirements. The artificial areas that are classified as not suitable (N) are concentrated around the global metropolises and represent only a minor share of the global suitable areas. In contrast, the protected areas, also rated as N, represent a substantial share in all growing regions with a global terrestrial coverage of about 15% [41]. For all three crops, the slope criterion restricts agricultural suitability where it is above 16% (S3) and above 30% (N), both of which obviously represent mountainous and hilly regions such as the Himalayas or the Andes, but also smaller-scale mountain chains across all growing regions. The coarse fragments of the soil hardly have any effect on crop suitability in any growing region. Additionally, both soil salinity and organic carbon content are not relevant criteria for the suitability of coffee, cashew and avocado and will therefore not be described further below.

Current coffee suitability

Globally, the highest overall current coffee suitability (S1 & S2) is found in Central and South America (esp. Brazil), in Central and West Africa, and in parts of South and Southeast Asia (Fig 1). The northern and southern extents of the global growing regions are restricted by climate factors, mainly by three parameters: long dry seasons (northern and southern boundaries of the growing regions in Africa, India, Australia, Eastern Brazil), high mean annual temperatures (West Africa, certain Southeast Asia regions, Central America) and low mean minimum temperatures of the coldest month (northern and southern boundaries in America, China, certain Southeast Asia regions, some mountainous areas). In some of the climatically suitable regions, the land and soil criteria greatly restrict the suitability of coffee cultivation. Low soil pH limits coffee suitability in South America (Amazon basin), Central Africa (Congo basin) and Southeast Asia (Sumatra, Malaysia, Borneo, New Guinea). In few regions, unsuitable soil texture (e.g. Florida) or steep slopes (e.g. North India) are the limiting factors.

The main coffee producing countries investigated (Brazil, Vietnam, Indonesia, Colombia) have very diverse agroclimatic conditions. Therefore, different climatic requirements (annual temperature, annual precipitation, length of dry season and minimum temperature of coldest month) are important limiting factors determining current coffee suitability (Table 3) depending on the region. For example, in Central and Southern Vietnam (see Fig 4), high annual temperatures limit current suitability, while in the South (too high) and in the northern mountains (too low) the limiting factor is the minimum temperatures of the coldest month, and in Central and Northeast Vietnam the high annual precipitation.

Fig 4

Current suitability and expected changes by 2050 for coffee in Vietnam.

A current landscape and soil suitability, B current climate suitability, C current overall suitability, D suitability change under RCP 2.6 (low emissions), E suitability change under RCP 4.5 (intermediate emissions), F suitability change under RCP 8.5 (high emissions).

Future coffee suitability

Taking into account climate change scenarios (Table 3 and Fig 5), the suitability of coffee will drastically decrease by 2050. The highest suitability (S1) will decrease by more than 50% in all three RCPs and the moderately suitable (S2) regions decrease by 31% (RCP 2.6) to 41% (RCP 8.5). In the RCP 4.5 and 8.5, even marginally suitable (S3) locations will decrease by 5% and 13%, respectively, while areas not suitable for cultivation (N) will increase in all scenarios. Negative changes in suitability will mainly be caused by increasing mean annual temperatures. Most current growing regions (Fig 5) are expected to decrease by at least one suitability class (Central and South America, Central and West Africa, India, Southeast Asia). Only a few regions, especially at the northern and southern borders of the growing areas, are expected to profit from climate change (e.g. Southern Brazil, Uruguay, Argentina, Chile, USA, East Africa, South Africa, China, India, New Zealand) due to increasing minimum temperatures of the coldest month.

Fig 5

Suitability change for coffee by 2050 according to RCP 4.5 (intermediate emissions).

A Central and South America, B West and Central Africa, C South and Southeast Asia.

The main coffee producing countries investigated (Brazil, Vietnam, Indonesia, Colombia) are all seriously affected by climate change with a strong decline in suitable areas (S1:48–97% reduction; S2:18–51% reduction) and an increase in unsuitable areas by 2050 (Table 3). For example, increasing mean annual temperatures are mainly responsible for negative changes in suitability in Vietnam in all three emission scenarios (Fig 4).

Current cashew suitability

Globally (Fig 2), the regions manifesting highest levels of current cashew suitability (S1 & S2) are Central (e.g. Mexico) and South America (e.g. Brazil, Venezuela), the Caribbean islands (e.g. Cuba, the Dominican Republic), Central (e.g. the Democratic Republic of the Congo) and West Africa (e.g. Nigeria), Madagascar and South (e.g. Sri Lanka) and Southeast Asia (e.g. Vietnam, the Philippines). The climate suitability is mainly limited by long dry seasons (e.g. India, northern boundaries in Africa, Brazil, Australia) and low mean annual temperatures (e.g. certain mountainous areas in South America and Southeast Asia). Additionally, the low mean minimum temperatures of the coldest month limit the northern and southern boundaries of some growing regions (e.g. China, USA, Brazil), and high annual precipitation limits suitability in some wet regions in Central and South America, West Africa, India and Southeast Asia. For the land and soil requirements it is mainly low soil pH limiting cashew suitability in South America (Amazon basin), Central Africa (Congo basin) and Southeast Asia (e.g. Malaysia, Sumatra, Borneo). Soil texture is not relevant for the suitability of cashew.

Fig 2

Overall current suitability for cashew (aggregated climate, land and soil suitability).

A Central and South America, B West and Central Africa, C South and Southeast Asia.

The suitability of cashew in the four major cashew producing countries investigated (Vietnam, India, Côte d’Ivoire, Benin) are different due to diverse agroclimatic conditions prevailing in these countries (Table 4). For India as an example, current suitability of different climate parameters is shown in Fig 6. It is mainly limited by long dry seasons, low minimum temperatures of the coldest month and high annual precipitation.

Table 4

Suitable cashew growing areas globally and in main producing countries (S1: Highly suitable, S2: Moderately suitable, S3: Marginally suitable, N: Unsuitable) for current (2000) and future (2050) conditions under three RCPs: 2.6 (low emissions), 4.5 (intermediate emissions), 8.5 (high emissions).

Expected changes in suitable areas are given as a percentage.

		RCP 2.6		RCP 4.5		RCP 8.5
Suit Class	2000 (km2)	2050 (km2)	Δ (%)	2050 (km2)	Δ (%)	2050 (km2)	Δ (%)
	World total
S1	1,909,945	2,236,263	17.1	2,271,876	18.9	2,199,364	15.2
S2	9,512,602	9,705,504	2.0	9,817,983	3.2	10,120,530	6.4
S3	8,628,977	9,472,204	9.8	9,570,975	10.9	9,725,958	12.7
N	104,448,211	103,085,763	‒1.3	102,838,901	‒1.5	102,453,881	‒1.9
	Vietnam
S1	51,682	70,861	37.1	68,820	33.2	68,827	33.2
S2	169,992	154,264	‒9.3	157,493	‒7.4	158,329	‒6.9
S3	44,266	41,088	‒7.2	40,028	‒9.6	39,299	‒11.2
N	53,442	53,170	-0.5	53,042	-0.7	52,928	‒1.0
	India
S1	9,278	7,770	‒16.3	6,546	‒29.4	5,568	‒40.0
S2	157,456	163,086	3.6	164,258	4.3	171,194	8.7
S3	315,272	328,034	4.0	332,251	5.4	331,303	5.1
N	2,566,359	2,549,476	‒0.7	2,545,311	‒0.8	2,540,302	‒1.0
	Côte d’Ivoire
S1	18,165	18,084	‒0.4	15,222	‒16.2	12,307	‒32.2
S2	190,087	190,168	0.0	193,030	1.5	195,945	3.1
S3	38,135	38,135	0.0	38,135	0.0	38,135	0.0
N	70,773	70,773	0.0	70,773	0.0	70,773	0.0
	Benin
S1	2,848	1,280	‒55.1	623	‒78.1	2	‒99.9
S2	25,283	26,850	6.2	27,507	8.8	28,128	11.3
S3	14,917	14,917	0.0	14,917	0.0	14,917	0.0
N	72,079	72,079	0.0	72,079	0.0	72,079	0.0

Fig 6

Current climate suitability for cashew in India.

A current suitability of mean annual temperature, B current suitability of mean minimum temperature of the coldest month, C current suitability of mean annual precipitation, D suitability of the length of the dry season, E overall current climate suitability.

Future cashew suitability

The future suitability models reveal that, depending on the growing region, both positive and negative changes in cashew suitability can be observed (Table 4 and Fig 7). In total, areas of high suitability (S1) increase by about 17% globally, and areas of moderate and marginal suitablility (S2 and S3) by 2–13% (see Table 3). Unsuitable areas are expected to slightly decrease globally. The main regions of positive change (Fig 7) are the USA, South America (Brazil, Paraguay, Uruguay, Argentina), East Africa around Lake Victoria, South Africa, Angola, North India, Vietnam, China and Australia. Primarily, rising minimum temperatures of the coldest month and in some regions (e.g. East Africa, Angola, Australia) rising annual temperatures are responsible for these positive changes in suitability. The main negative changes (Fig 7) are modelled to occur in Central and South America (Panama, Colombia, Venezuela), West Africa (e.g. Nigeria) and South and Southeast Asia (Sri Lanka, the Philippines, Indonesia, Cambodia, Myanmar), primarily due to increasing annual temperatures and in few regions (e.g. parts of Panama or Sri Lanka) due to higher precipitation.

Fig 7

Suitability change for cashew by 2050 according to RCP 4.5 (intermediate emissions).

A Central and South America, B West and Central Africa, C South and Southeast Asia.

The expected changes in cashew suitability due to climate change in the four major cashew producing countries is shown in Table 4. Both in Côte d’Ivoire and Benin, a large proportion of the highly suitable (S1) areas are expected to become less suitable (S2) by 2050 due to increasing annual temperatures. In Côte d’Ivoire, the expected reduction in S1 is 16% (RCP 4.5) to 32% (RCP 8.5), and in Benin 55% (RCP 2.6) to nearly 100% (RCP 8.5). In the northern parts of both countries, long dry seasons limit cashew suitability. In India and Vietnam, both positive and negative changes in suitability might occur. In North India, positive changes are expected due to higher minimum temperatures of the coldest month, while in South India and Northeast India the negative changes will dominate, mostly due to increasing precipitation. In total (Table 4), the highly suitable (S1) areas will decrease by 16% (RCP 2.6) to 40% (RCP 8.5), whereas both S2 and S3 will increase by 4% (RCP 2.6) to 9% (RCP 8.5). In Northern and Central Vietnam, positive suitability changes are expected, while in the South suitability decreases. For both positive and negative changes, it is mainly increasing annual temperatures which are responsible in the respective regions. In Vietnam (Table 4), S1 will increase by about 35% in all scenarios, while S2 will decrease by 9% (RCP 2.6) to 7% (RCP 8.5) and S3 by 7% (RCP 2.6) to 11% (RCP 8.5). Unsuitable areas will show a slight decrease in general.

Current avocado suitability

The avocado suitability under current climatic conditions (Fig 3) revealed large regions with high suitability (S1 & S2) in Central and South America (e.g. Honduras, Venezuela, Bolivia, Brazil), in West and Central Africa (e.g Côte d’Ivoire, Cameroon, Central African Republic, the Democratic Republic of the Congo, the Republic of the Congo, Uganda), and in South and Southeast Asia (e.g. India, Sri Lanka, Vietnam, Cambodia, Thailand, Myanmar, Indonesia). The minimum temperature of the coldest month and the annual precipitation are the two most limiting factors in the suitability model. The boundaries of the avocado growing regions in North and South America, southern Africa and northern Asia, as well as in certain mountain regions, are limited by low minimum temperatures in the coldest month. The annual precipitation limits avocado suitability both in wet (Central America, West Africa, Southeast Asia) and dry regions (Eastern Brazil, East Africa, Australia, northern boundaries in Africa) because avocado has a narrow precipitation optimum (see Table 1). Only in a few regions do long dry seasons limit avocado suitability. Compared to climatic criteria, land and soil criteria only play a minor role in the global avocado model with low soil pH (Amazon basin, Congo basin, Borneo) and unfavourable soil texture (Central America, Central Africa), limiting suitability in few regions.

In the four main producing countries investigated (Mexico, the Dominican Republic, Peru, Indonesia), avocado suitability is also mainly limited by low and high precipitation and low minimum temperatures of the coldest month (Table 5). Regarding the land and soil criteria, only limited regions are affected by soil texture and slope in all four countries and by low soil pH in Peru and Indonesia.

Future avocado suitability

The climate change scenario models revealed both positive and negative impacts on avocado suitability in different regions by 2050 (Fig 8). Positive changes due to increasing minimum temperatures in the coldest month are mainly identified at the northern and southern boundaries of the growing regions in America (the USA, Brazil, Uruguay, Paraguay, Argentina), Africa (Angola, Zambia), Asia (North India, China) and Australia. In sub-Saharan and East Africa (e.g. Burkina Faso, Nigeria, Chad, Ethiopia, Uganda, Kenya) and parts of India, positive changes in suitability are caused by increasing precipitation. Regions of mainly negative impacts are due to drier (e.g. Venezuela, Eastern Brazil) or wetter conditions (e.g. Central Africa, Indonesia, the Philippines). Both positive and negative impacts are expected in Central America, West Africa and Southeast Asia (e.g. Vietnam, Myanmar) based on changes in temperature or precipitation. On the one hand (Table 5), the highly suitable (S1) areas will decrease globally by 14% (RCP 2.6) to 41% (RCP 8.5). On the other hand, S2 areas will increase by 12% (RCP 2.6) to 20% (RCP 8.5) and S3 areas by 6% (RCP 2.6) to 8% (RCP 8.5), while the global unsuitable areas will slightly decrease.

Fig 8

Suitability change for avocado by 2050 according to RCP 4.5 (intermediate emissions).

A Central and South America, B West and Central Africa, C South and Southeast Asia.

According to the model, in all four main producing countries investigated (Mexico, the Dominican Republic, Peru, Indonesia), both positive and negative changes in avocado suitability due to climate change will occur in different regions by 2050 (Table 5). They can be explained mainly by the expected changes in precipitation patterns (both wetter and drier conditions causing positive or negative changes) and to a smaller extent by increasing minimum temperature of the coldest month (positive changes). In Mexico (Fig 9 and Table 5), the positive changes tend to dominate, with increases in S1 (by 87% [RCP 2.6] to 66% [RCP 8.5]) and S2 (by 15% [RCP 2.6] to 21% [RCP 8.5]), and S3 and N areas slightly decreasing. Based on the RCP 4.5 intermediate emissions scenario (Fig 9), future avocado suitability in Mexico is mainly restricted by climate suitability, which is mainly due to the expected mean annual precipitation and the minimum temperature of the coldest month. In the Dominican Republic, the highly suitable (S1) areas will decrease by 59% (RCP 2.6) to 85% (RCP 8.5) and the marginally suitable (S3) areas by 10% (RCP 2.6) to 16% (RCP 8.5), while the moderately suitable (S2) areas will increase by about 30% in all scenarios. In Indonesia, S1 areas will decrease by 40% (RCP 2.6) to 65% (RCP 8.5). Except for RCP 2.6, where S2 and S3 areas will increase by 2% and 4%, respectively, S2 (-6%) and S3 (-12%) suitability is also expected to decrease. In Peru, the negative changes are expected to dominate in all scenarios with S1 decreasing by 54% (RCP 2.6) to 76% (RCP 8.5).

Fig 9

Future suitability for avocado in Mexico and expected change by 2050 based on RCP 4.5 (intermediate emissions).

A Overall future suitability, B suitability change by 2050, C current landscape and soil suitability, D future climate suitability, E future suitability of mean annual temperature, F future suitability of mean minimum temperature of the coldest month, G future suitability of mean annual precipitation, H suitability of the length of the dry season.

Discussion

Current crop suitability compared to main producing regions

As could be expected by the biophysical requirements of coffee, cashew and avocado (Table 1), both temperature and precipitation criteria represented the most limiting factors for the suitability of the three crops in different regions. However, for all three crops, low soil pH was shown to be an additional important limiting factor in South America (Amazon basin), Central Africa (Congo basin), Southeast Asia (e.g. Sumatra, Malaysia, Borneo, New Guinea). In certain regions, also steep slopes and unfavourable soil texture were identified as important limitations. We therefore conclude that the integration of topographic and soil factors is crucial for improving current and future crop suitability modelling, especially on a regional or local scale.

The modelled global suitable regions of coffee, cashew and avocado comprise most major producing countries based on FAOSTAT (www.fao.org/faostat). For coffee arabica, suitable growing regions were identified in all major producing countries. In our analysis, however, certain major cashew producing countries (India, Senegal, Guinea-Bissau, Mali, Burkina Faso) resulted in low cashew suitability. Some of the growing regions in these countries were rated as unsuitable because they have a dry season lasting longer than six months. We therefore conclude that this criterion is either too restrictive with cashew being a drought tolerant crop [24], or that irrigation during the dry season is required in these regions. When changing the length of the dry season for unsuitable areas (N) from more than six to more than seven months, the major growing regions in India and West Africa lie within the suitable area. Similarly, suitable avocado growing areas were found in most but not all major producing countries and regions according to our model. Some of the growing regions of Peru (coastal area), the USA (California), Chile, South Africa, Spain, Morocco, Israel and Australia were rated as not suitable for avocado cultivation. On the one hand, this limitation is due to insufficient annual precipitation in some of these regions, where avocados are in fact irrigated [30]. On the other hand, low minimum temperatures in the coldest month co-limit suitability in some of these regions, indicating that this criterion could be too restrictive. This is supported by Wolstenholme [30] who reported growing areas with minimum temperatures during the coldest month of below 8°C (rated as ‘unsuitable’ in our model) in parts of California, Israel, New Zealand, South Africa and southern Australia. However, the same author also mentioned the importance of good site selection and management practices from the point of view of frost in order to reduce frost damage in these marginal areas.

Shifts in crop suitability due to climate change

When taking into account climate change scenarios, shifts in suitable growing regions are expected for all three crops, with both some of today’s suitable growing regions disappearing and new ones emerging. Compared to cashew and avocado, climate change has the highest negative impact on currently suitable coffee growing regions because of its greater susceptibility to high temperatures. While cashew suitability is also negatively affected by rising temperatures in certain regions, avocado is more affected by changes in precipitation (both negatively and positively). All three crops, however, profit from increasing minimum temperatures at high latitudes and high altitudes. The impacts of the three RCPs (2.6, 4.5 and 8.5) applied in our model generally show similar patterns. For coffee, the negative impacts on current growing regions clearly intensify from RCP 2.6 to 8.5. The same is true for cashew and avocado, where both positive and negative global as well as regional trends intensify from RCP 2.6 to 8.5.

Coffee

The drastic overall decrease in coffee suitability by 2050 that was found in this study is in line with the results of existing climate change impact studies for coffee arabica [2, 3, 10, 11]. Coffee is described as a crop which is highly sensitive to climate change. As in the present study, Bunn et al. [2] and Ovalle-Rivera et al. [3] show overall reductions in global suitable areas for coffee, mainly at low latitudes and low altitudes. According to Ovalle-Rivera et al. [3], the impacts of climate change on coffee suitability are extremely variable at national and global levels, confirming our findings. The decrease in suitable area by about 50% across scenarios by 2050 shown by Bunn et al. [2] is in a similar range compared to the more than 50% reductions in S1 and 30–50% reductions in S2 found in this study. In addition, the main producing regions (Brazil, Vietnam, Indonesia, Colombia) are also expected to experience substantial reductions in suitable areas for coffee cultivation [2, 3, 10]. In accordance with our findings, few regions in East Africa, Asia and South America are described to potentially be able to benefit from climatic change [2, 3, 10, 11]. They are generally at higher elevations or at the latitudinal boundaries of the growing regions. Potential future growing regions around South Brazil, Uruguay and Northern Argentina were explicitly studied by Zullo et al. [11], and have also been identified here. Several other regional studies have investigated coffee arabica suitability changes in Mesoamerica [7, 8], Nepal [9], East Africa [5, 6] and Zimbabwe [4]. However, areas of increasing suitability in South China identified in this study have not been mentioned in previous investigations. In all coffee modelling studies mentioned above, temperature variables were the most important factors explaining decreasing suitability, except for Chemura et al. [4] who found changes in the distribution of precipitation were most important in Zimbabwe. This is in line with our results, where temperature variables are mainly responsible for future changes. Additionally, no previous models took land and soil requirements into account which limited coffee suitability in some areas of all growing regions. This is important in modelling studies for planning new coffee plantations only in areas where coffee is locally adapted and requires a minimum of additional inputs and where there are no major environmental trade-offs.

Cashew

In contrast to coffee, much less research is available about the biophysical suitability of cashew and avocado. For cashew, no global assessment of current and future growing regions is currently available. Cashew suitability maps are available for Ghana and Côte d’Ivoire [14], Malawi [13], India [15] and Lombok Island in Indonesia [16, 17]. Climate change impacts were only modelled for Ghana and Côte d’Ivoire [14]. A few more studies are available on climate change perceptions of cashew growers [42] and the impact of climate factors on cashew productivity [43] in Benin, and about the socio-economic and environmental impacts of cashew expansion in Guinea-Bissau [44-46]. In contrast to CIAT [14] who identified large areas with a positive impact of climate change on cashew suitability in Côte d’Ivoire and Ghana by 2050, no change or slight reductions in suitability were modelled in our study for these regions. This is most probably due to the different methodology (maximum entropy modelling) applied in their study [14], which is based on the current distribution of cashew production, while our model is based on the biophysical requirements of the crop. In fact, the cashew suitability within the current growing regions modelled by CIAT [14] is to a great extent consistent with our findings. While the current cashew suitability in Malawi matches with the suitability map of Benson et al. [13] for rainfed cropping under traditional management, it is not comparable with the findings of Widiatmaka [17] for Lombok Island in Indonesia, due to different criteria applied in the former’s model. However, comparing our results with the major growing regions in India [15] and West Africa [47] reveals that based on long dry seasons, cashew suitability is underestimated in some of these regions, as was mentioned above.

Avocado

Similar to cashew, only a limited number of assessments of avocado growing regions are available. The most comprehensive characterisation of current and future distributions of avocado was undertaken by Ramírez-Gil et al. [18] across the Americas. Avocado suitability was also studied in Mexico [29, 48], Colombia [49], Brazil [50], Turkey [20] and Australia [19]. Substantial differences were identified between the suitable avocado growing regions modelled in our study and by Ramírez-Gil et al. [18]. While in some regions (e.g. Colombia, Honduras or Nicaragua), current avocado suitability is similar, it is higher in their study in parts of Uruguay, Argentina, Chile, Peru, Mexico or the USA, and lower in parts of Brazil, Bolivia, Venezuela, Guyana or Surinam. We explain these differences by the ecological niche modelling approach applied by Ramírez-Gil et al. [18] which is based on current production locations, some of which coincide with arid zones unsuitable for avocado cultivation under rainfed conditions. In accordance with our findings, Ramírez-Gil et al. [18] identified expansion of growing regions due to increasing temperatures mainly in temperate areas (e.g. Argentina). Contractions of suitable ranges were mainly related to temperature and precipitation increases, while in our model they were related to both drier and wetter conditions but not temperature increases. Based on reported heat stress effects during critical periods such as pollination and fruit set [30], we therefore conclude that the requirements at high temperatures are not restrictive enough in our model. The current production locations in Colombia as reported in Ramírez-Gil [49] and in the Paraná River Basin in Brazil described by Caldana et al. [50] lie within suitable areas based on our model. The main producing municipalities in Mexico (Michoacán State) presented in Lira-Noriega et al. [51] are only marginally suitable due to a long dry season and can be confirmed by required supplemental irrigation in this region as flowering and early fruit development occur in a dry period [30]. Based on the analysis of Charre-Medellín et al. [48], avocado suitability will decrease in this region by 2050, an aspect which was not found in our study. Selim et al. [20] identified suitable growing areas for avocado at the Mediterranean coast in Antalya, Turkey, where according to our model, suitability will increase with climate change due to increasing temperatures. In Australia, Putland et al. [19] identified suitable avocado growing regions in southwest Western Australia, along the Murray River and in coastal New South Wales, which were rated as too cold or too dry in our model. As mentioned above, cultivation is only possible in these regions with irrigation or measures against frost.

Modelling approach

The multi-criteria evaluation approach used in this study is based on crop requirements and respective bioclimatic and soil data, unlike other bioclimatic crop modelling approaches that are based on current production locations, such as maximum entropy modelling [52]. It is therefore a transparent approach especially suitable for identifying limiting factors for crop growth but there are also limitations associated with it. As mentioned in the methodology, our model does not take into account management options such as irrigation or liming. Important production locations are therefore rated as not suitable, or only marginally suitable, as was shown above. There are also no interactions modelled between different criteria (e.g. between precipitation and soil texture), that could potentially affect suitability. We also did not discriminate between different varieties of the same crop but took this variation into account when defining the ranges of the different suitability classes. The modelling of climate change impacts on suitability was based on three parameters. However, likely impacts on the length of the dry season, soil factors, pest and disease pressure or risks associated with extreme weather phenomena were not taken into account due to their high uncertainties.

Additionally, using datasets with a coarse resolution of 30 arc seconds (ca. 1km2) may fail to capture the variability of local characteristics [12], particularly topography, soil factors and microclimatic conditions. For global assessments, however, 30 arc seconds can be considered a high resolution that fits the scope of this study.

Conclusions

This study presents the first global evaluation of coffee arabica, cashew and avocado suitability combining both climate and soil factors. It also represents the first global assessment of climate change impact on cashew and avocado suitability. For the potential global growing regions of all three crops, climate requirements were more important limiting factors than land and soil requirements. High annual temperatures, low minimum temperatures, long dry seasons and low or high precipitation were the most relevant climate criteria. However, apart from protected and artificial areas that were rated as unsuitable, low soil pH, steep slopes or unfavourable soil texture were also important limiting factors in some areas. We therefore suggest combining climate and soil parameters in future modelling attempts to increase their significance, especially on a regional or local scale.

Shifts in suitable growing regions due to climate change with both expansions and contractions were found for all three crops. Coffee proved to be most vulnerable to climate change with negative impacts dominating in all growing regions, primarily due to increasing temperatures. Compared to coffee, cashew and avocado were found to be more resilient to climate change. For cashew, which showed the highest suitability range, both positive and negative effects of climate change were found. While globally, the suitable cashew growing areas are expected to increase, in some of the main producing countries (e.g. India, Côte d’Ivoire and Benin), areas of high suitability are expected to decrease. Similarly, for avocado, the suitable areas are expected to expand globally, while the most suitable areas in some of the major producing countries (e.g. the Dominican Republic, Peru, Indonesia) might decrease. All three crops however profit from increasing minimum temperatures at high latitudes and high altitudes.

The study has shown that climate change adaptation will be necessary in most major producing regions of all three crops. Adaptation measures can include site-specific management options, plant breeding efforts for varieties that are better adapted to higher temperatures or drought and in the case of coffee, replacement of arabica with robusta coffee in certain regions [2]. New production locations at higher altitudes and latitudes might create new market opportunities. However, policies and strategies are required to ensure that shifts in production locations will not lead to negative environmental impacts such as deforestation, loss of biodiversity or ecosystem services. Additionally, landowners and farmers in current and future production locations must be willing to change their management or grow a new crop. Therefore, adaptation measures and shifts in production will each have to be addressed in participative approaches that allow the engagement of local stakeholders.

27 May 2021

PONE-D-21-13411

Expected global suitability changes of coffee arabica, cashew and avocado due to climate change

PLOS ONE

Dear Dr. Grüter,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

The reviewers have suggested a more thorough presentation of the modeling framework of the study with explanations on the distinction between cultivated and non-planted species. This is also related to the lack of a proper-definition of 'suitability' in the study that provide the conceptual framework of what the modelling needs to achieve. Sample bias correction and choice of the threshold for determining suitable and unsuitable areas should also be explained in more detail.

Please submit your revised manuscript by Jul 11 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see:  http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at  https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols .

We look forward to receiving your revised manuscript.

Kind regards,

Abel Chemura

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

and

2. We note that you have stated that you will provide repository information for your data at acceptance. Should your manuscript be accepted for publication, we will hold it until you provide the relevant accession numbers or DOIs necessary to access your data. If you wish to make changes to your Data Availability statement, please describe these changes in your cover letter and we will update your Data Availability statement to reflect the information you provide.

3. We note that Figures 1-9 in your submission contain map images which may be copyrighted.

All PLOS content is published under the Creative Commons Attribution License (CC BY 4.0), which means that the manuscript, images, and Supporting Information files will be freely available online, and any third party is permitted to access, download, copy, distribute, and use these materials in any way, even commercially, with proper attribution. For these reasons, we cannot publish previously copyrighted maps or satellite images created using proprietary data, such as Google software (Google Maps, Street View, and Earth). For more information, see our copyright guidelines: http://journals.plos.org/plosone/s/licenses-and-copyright.

We require you to either (a) present written permission from the copyright holder to publish these figures specifically under the CC BY 4.0 license, or (b) remove the figures from your submission:

a. You may seek permission from the original copyright holder of Figures 1-9 to publish the content specifically under the CC BY 4.0 license.

We recommend that you contact the original copyright holder with the Content Permission Form (http://journals.plos.org/plosone/s/file?id=7c09/content-permission-form.pdf) and the following text:

“I request permission for the open-access journal PLOS ONE to publish XXX under the Creative Commons Attribution License (CCAL) CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). Please be aware that this license allows unrestricted use and distribution, even commercially, by third parties. Please reply and provide explicit written permission to publish XXX under a CC BY license and complete the attached form.”

Please upload the completed Content Permission Form or other proof of granted permissions as an "Other" file with your submission.

In the figure caption of the copyrighted figure, please include the following text: “Reprinted from [ref] under a CC BY license, with permission from [name of publisher], original copyright [original copyright year].”

b. If you are unable to obtain permission from the original copyright holder to publish these figures under the CC BY 4.0 license or if the copyright holder’s requirements are incompatible with the CC BY 4.0 license, please either i) remove the figure or ii) supply a replacement figure that complies with the CC BY 4.0 license. Please check copyright information on all replacement figures and update the figure caption with source information. If applicable, please specify in the figure caption text when a figure is similar but not identical to the original image and is therefore for illustrative purposes only.

The following resources for replacing copyrighted map figures may be helpful:

USGS National Map Viewer (public domain): http://viewer.nationalmap.gov/viewer/

The Gateway to Astronaut Photography of Earth (public domain): http://eol.jsc.nasa.gov/sseop/clickmap/

Maps at the CIA (public domain): https://www.cia.gov/library/publications/the-world-factbook/index.html and https://www.cia.gov/library/publications/cia-maps-publications/index.html

NASA Earth Observatory (public domain): http://earthobservatory.nasa.gov/

Landsat: http://landsat.visibleearth.nasa.gov/

USGS EROS (Earth Resources Observatory and Science (EROS) Center) (public domain): http://eros.usgs.gov/#

Natural Earth (public domain): http://www.naturalearthdata.com/

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: No

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The scientific idea in the manuscript is valid and tangible. However, it requires a lot of editing and the wording requires to be edited. Personally, the authors can split this manuscript into 3 papers to make it easy for the reader. As it stands one gets lost as to which part the author is talking about.

Reviewer #2: This is an interesting attempt to analysis and comprehensive study about of use of GIS-based decision support system methods to examine the niche, biophysical and ecological interactions between ecosystems and agriculture impotent species under climatic change scenarios. I appreciate these types of studies that use DSS methods to examine the distribution interactions between species and ecosystem that may underpin local and regional risks of agricultural species losses patterns. The authors have also made enormous efforts to accumulate relevant data and design a robust DSS methodology. However, I consider that there are some methodological shortcomings remaining, which negate many of the conclusions of their analysis (explained in detail below). Moreover the manuscript lacks clarity and care with language in some places which I have mentioned under minor comments below. I hope these comments will prove useful for the authors to rethink their analysis and incorporated into the manuscript. I have some general and specific comments that should be addressed by the authors. The manuscript is recommended for publication with major revision. In the next paragraph, I explain the comments.

Mayor revision

1. The major drawback of the methodological approach is that the authors presented the final model how a niche description approach. I think niche can have different types (Grinnell or Elton), and is necessary defined if the models are modeling fundamental niche or realized niche or. In addition, the modelling approach based on physiological response of species had been many restriction (please see: DOI:10.1515/eje-2015-0014

2. Oher question in methodological approach used is that the authors have not accounted for spatial biases in recording effort for either species or their ecosystems. It is likely that the process by which planted species are recorded within commercial programs systems or by the research community are very different to the processes by which not commercial zones in which natural conditions when the species can survive under natural suitability (main producing countries). The authors may be simply comparing the niche of “where people like to record planted species” with the niche of “where people like to record not planted species” and comparisons of geographic distributions are therefore very biased and the authors cannot conclude to what % of area of X species distribution will be reduce under climate change scenarios. In my opinion a solution to this is to use a target group to inform the background for the DSS models (one for planted species, one for not planted species), in addition to the accessibility criteria used, prior to comparing niches between panted and not planted species (https://doi.org/10.1890/07-2153.1). The papers by Broenimann et al. 2012 point out the importance of accounting for the differential sampling of different environmental conditions in the recorded distributions of two species when examining niche and distributions (https://doi.org/10.1111/j.1466-8238.2011.00698.x).

3. The authors throw a lot of bioclimatic variables into the model which increases the chances of overfitting and also of detecting niche differences between species, but I can’t see which variables were selected.

4. The processes of DSS is a bit more complex, in which a multi-step must be made in order to improve the capacity of the models and not simply reproduce information of low quality. On the other hand the new approach to correct model under future we should incorporate the sources of variation in future model projections (https://doi.org/10.7717/peerj.6281 and https://doi.org/10.1071/CP19094).

5. Since the DSS approach used and future scenarios present a good approximation when the analysis are local or regional, it is appropriate, practical and methodologically correct to do fine work for a worldwide approach?

6. More details are also needed to describe the analysis of the data so reviewers can ensure their appropriateness for the type of data presented. Additional attention to detail is needed to improve the overall quality of manuscript including the small detail about concept in DSS, modeling, data uses, and computation performance. Please also ensure the relevant parts of the manuscript are in the correct sections (i.e. results confined to the results!).

7. It is not clear when the authors described that GCMs were used, but I only identify the following models: BBC-CSM1-1, CCSM4, CNRM-CM5, GFDL-CM3, GISS-E2-R, HadGEM2-AO, HadGEM2-ES, IPSL-CM5A-LR, MIROC-ESM; CHEM, MIROC-ESM, MIROC5, MPI-ESM-LR, MRI-CGCM3 and NorESM1-M), but is not clear how were represented in the final raster. You can used ensemble model approach? If this concept is the best representation, so the methodology of the REA (Reliability Ensemble Averaging) approach is meaningless.

Minor revision

Title

My suggestion is change the title, because is very general.

Abstract

Try to be specific and write a paragraph more informative, because is very confused the aim, and the relationship between the approach used (DSS approach and climate change). In addition, you can add more information based in data analysis (statistical, among others).

Introduction

In my opinion the introduction are poorly described. My suggestion to the authors is add more information about the use of the DSS algorithms (advantages and disadvantages with ecological niche models approach (ENM)). In addition, try to be more informative about of parameters associated with the algorithms used in DSS, how are used, your mathematical concepts, which are the predictors, how are obtained the predictors, advantages, disadvantages, limitations, among others. This is very important because actually the ENM approach is very popular with. On the other hang the black box model giving rise to models without biological sense.

My suggestion is that aim and hypothesis should be improve, because is not clear, in especial the roll of the approach used.

Material and Methods

It is necessary to contextualize and clearly explain the bioclimatic, topographic variables used, the characteristics of the ecosystem system (e, j., area, currently status), explain more details about countrys sleected (areas, climatic, topographic variability), among others. The methods are superficially described, omitting basic information that is of utmost importance to guarantee reproducibility, a basic criterion in scientific research.

Is no clear the algorithms used. Is necessary add more details about this process as parametrization, performance computational, and evaluated results of model (calibration and validations). In addition, I can’t find the result of evaluation of model under climate change scenarios.

The model processes is a bit more complex, in which a multi-step must be made in order to improve the capacity of the models and not simply reproduce information of low quality. In addition, actually when algorithm is used must be considered an exhaustive evaluation of the parameters associated with the performance computational. On the other hand I suggest that to select the best model approach is necessary incorporate the sources of variation.

Result and discussion

Emphasize on explaining what advantages you have when using these spatial analysis strategies and not others.

How did you relate the genetic and spatial dimensions?

What is the current use of the areas with environmental problem under landslides?

How is the productivity systems in the area tested?

It is important to highlight the results and that these are incorporated into a management program for conservations and what environmental implications and sustainability indicators represent the use of these practices at the government level.

Conclusion

The author should improve the conclusion and focus on the most important data of the study. In addition, the conclusions presented do not represent the importance of the research work.

Supplementary material

For better reproducibility, I suggest that the authors publish the codes and data as supplementary material or in a free repository (Gib-Hub). .

References

Review the correct format used by the Journal.

Figures

Poor resolution and need to be compressible standalone

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Emmanuel Junior Zuza

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

Submitted filename: Review1-Emmanuel Junior Zuza.docx

Click here for additional data file.

11 Oct 2021

Dear Editor, dear Reviewers,

Thank you for your constructive and detailed feedback to our submitted manuscript. We would like to provide a general answer to the main points raised in the review, and then answer to the individual comments of both reviewer 1 and 2 below. We have uploaded our revised manuscript with and without track changes to the editorial management system.

First of all, we feel that there is a misunderstanding regarding the modelling approach applied in our study compared to other modelling approaches applied and described in literature. We therefore thoroughly revised the methodology part of the manuscript to be clear and transparent about our approach, which is also described in Jaisli et al. (2019) in detail, as referenced in the manuscript.

We would like to stress that we did not use a machine learning model (such as maximum entropy modelling) with an underlying algorithm in our study. We applied a very simple multi criteria evaluation approach, based on the aggregation of separate suitability maps of individual climate, soil and land parameters. These individual suitability maps are based on a classification of crop requirements into four suitability classes. The thresholds between these suitability classes were defined for each individual climate, land and soil parameter based on existing literature (see Table 1 of the manuscript) and do therefore not represent values between 0 and 1. For the combination of the different parameters, their suitability maps were aggregated for each land unit based on the lowest suitability class (maximum limiting factor), without using weighting factors. In our opinion, the simplicity and transparency of our approach is also its strength.

This means that our study is not based on present occurrence data of the crops investigated, but on the literature description of their requirements and on respective bioclimatic and biophysical datasets.

We are looking forward to hearing about your decision.

Yours sincerely,

Roman Grüter

Reply to the reviewer comments

Reviewer 1

This study aimed to assess the current and future suitability of coffee, cashew and avocado production on a global and county scale using biophysical parameters. This study is important because it considers land, soil and climatic predictors that influence the production of the three perennial crops. The authors utilized the GIS-based decision support system CONSUS model for their analysis. However, the paper is missing information on the importance and influence of biophysical factors on the crops current and future suitability. This is vital based on the argument that such parameters have not been used in modelling works for these crops, hence filling the scientific gap.

RG: In the first paragraph of the discussion, the importance of topographic and soil factors for crop suitability modelling is now highlighted.

Major comments

• There is a need to revise the submission taking into account grammatical errors in the manuscript.

RG: The manuscript has been thoroughly revised and checked by a native English speaker (EN service of our university).

• Authors need to revise their methodology and should make sure that it is clear on what they did. They need to highlight if they used known areas of occurrence of the crops, i.e. presence data. The authors mention a threshold that was set in the methodology but does not show what it was. Ideally, the authors need to make known what this is, i.e. 0,1 with the ranges (suitable, unsuitable, marginal etc.). If you could explain the maximum limiting factor (MLF) and why you used it.

RG: The materials and methods section was revised to make it clear what we did. As mentioned above, we did not use present occurrence data of the crops investigated (except for the analysis of the results for the main producing countries) and the analysis is not based on a mathematical model with independent and target variables. We applied a simple multi criteria evaluation approach with a suitability classification of several climate, land and soil parameters that were finally aggregated, based on the factor of lowest suitability for each land unit (the maximum limiting factor is explained in the methodology with soil pH as an example). The thresholds were defined for the four suitability classes of each parameter in its respective unit, i.e. they do not represent values between 0 and 1.

• One of the biggest things when using bioclimatic data from WorldClim is multicollinearity. Would the authors provide a sentence that proves that the bioclimatic variables they used were not correlated, as this has been shown to cause model overfitting?

RG: We did not check for multicollinearity because we did not use a multiple regression model with several independent variables but applied a multi criteria evaluation approach.

• The authors should consider breaking the results into different subheadings, i.e. factors determining suitability, the current suitability (global and country-specific) and future suitability (global and country-specific). As it stands, it becomes confusing what the authors are trying to convey.

RG: We agree and added the following subtitles: Factors determining suitability, Current coffee suitability, Future coffee suitability, Current cashew suitability, Future cashew suitability, Current avocado suitability, Future avocado suitability. We also displaced the information about the main producing countries to these chapters.

• The results highlight that cashew has the largest suitable growing areas globally than avocado and coffee under the current climate conditions. It will be more meaningful if you could provide the percentages (%) of these. We can see these on the maps but perhaps consider having a table/fig in the supplemental materials showing the differences based on the classes.

RG: We give the information in km2 in Table 3 (resp. in the new Tables 3-5). This reference was added to the text.

• In the results, there is a mixture of words that make it difficult to understand what you mean, i.e. the minimum temperature of the coldest month and minimum temperature in the coldest season. The authors need to be consistent with the wording.

RG: We tried to diversify our expressions so that the text would be attractive to the reader. However, we understand the confusion caused by the mixture of words and changed the expression ‘coldest season’ to ‘coldest month’ throughout the manuscript. We also changed ‘agricultural suitability’ to ‘crop suitability’ twice to be more consistent.

• The results in table 3 should be presented per crop and not as a whole thing. It makes the reader move back and forth to see what the authors say in the country-specific suitability.

RG: As suggested, Table 3 was split into three individual tables (per crop).

• The authors need to revise their discussion comparing the current crop suitability with the main producing regions. It would make more sense if they separated the two, i.e. discuss the current crop suitability results from those with the main producing areas as they did in the results.

RG: We added a paragraph in the beginning of the chapter “Current crop suitability compared to main producing regions” to discuss the current crop suitability results and to stress the importance of topographic and soil factors on crop suitability, as was suggested above.

• The authors need to focus on discussing the findings of their study rather than existing literature. For instance, in avocado's discussion from Line 474-479, the authors highlight what others have done. It is recommended to start discussing your results and later compare these to what others have done. This way, the reader understands your findings.

• RG: We discuss the findings of our study in the first paragraph of chapter “shifts in crop suitability due to climate change” and in the chapter “current crop suitability compared to main producing regions” for all three crops. In the individual chapters for coffee, cashew and avocado, we would like to focus on discussing the comparison with existing literature.

Minor comments

Line 1-2: Consider rewording the title "Expected global suitability of coffee, cashew and avocado due to climate change". Do not use the species name in the title.

RG: the title was changed accordingly.

Line 12: Replace the word Coffee arabica with coffee. This applies to the rest of the document.

RG: ‘arabica’ was removed in most of the cases. The differentiation between arabica and robusta coffee is however very important in modelling studies because they have very specific and different growth requirements. We therefore left the expression ‘Coffee arabica’ where we found it important to differentiate it from Coffee robusta (e.g. in the methodology).

Line 15: replace "or" with "and".

RG: this was changed.

Line 18-19: Restructure your sentence to make more sense. "We used climate outputs from 14 global circulation models based on three emission scenarios to model the future climate changes impacts on the crops both globally and in main producing countries".

RG: Thank you for your suggestion. We changed the sentence accordingly.

Line 27-29: Provide how much expansion you are talking about globally, i.e. the percentage (%) increase and the % decrease within the countries (cumulative).

RG: In our opinion this would be too detailed for the abstract since we would have to introduce much more details about the methodology (since there are different levels of suitability), which would go beyond the limits of the abstract.

Line 35-36: Include the scientific names of coffee, cashew and avocado. A reference would be great as well.

RG: Since this is a general statement about these plantation crops, we have not included the scientific names here, but in the materials and methods section below.

Line 36-38: Consider rephrasing this sentence.

RG: Sentence was rephrased.

Line 76-78: Remove the scientific names and include these in the first paragraph of the introduction.

RG: In our opinion, it makes more sense to mention the scientific names of the species investigated in the materials and methods than in the general part of the introduction.

Line 158-159: This is a repetition of what you have already shown. Consider rephrasing this.

RG: The sentence was rephrased and referenced to Table 2.

Line 228-229: Should be N and S boundaries of Africa, remove "in".

RG: In this sentence, the N and S boundaries of the growing regions in Africa are meant, and not the N and S boundaries of Africa. To make this clear, the sentence was rephrased accordingly.

Line 253-256: Consider putting Table 3 in landscape mode as it is not visible in portrait.

RG: Table 3 was split in Tables 3-5.

Line 284: Low soil pH shall mean by how much? Please consider putting the pH number on there.

RG: From Table 1 and the description of the methodology, it should now be clear that a pH<4.5 is rated as unsuitable and a pH between 4.5 and 4.8 is rated as marginally suitable for cashew.

Line 290: Best write 2-13% (see Table 3).

RG: This was changed accordingly.

Line 311: I think the authors mean cashew suitability and not agricultural suitability. If this is the case, please rephrase the sentence.

RG: Thank you. This was changed to cashew suitability.

Line 318: Check the spelling of Vietnam.

RG: This was corrected.

Line 328: Consider revising the sentence.

RG: This was rephrased.

Line 334-337: Rephrase. The sentence is unreadable.

RG: This sentence was rephrased and simplified.

Line 372-373: Consider revising this sentence as it does not make sense. The authors talk about climate suitability restricting avocado suitability. I think you mean weather parameters, not climate suitability per se.

RG: We have rephrased the sentence to make it clearer. It mainly has to do with the methodological approach of the study, in which climate suitability is defined as the aggregation of several climate parameters such as temperature or precipitation (see Table 1).

Line 392-394: Rewrite the sentence.

RG: This sentence was rephrased.

Line 426-427: The author's highlight, "Similar trends were found in general". Which trends are these? It would be important to start highlighting whether climate change will impact coffee production +ve or -ve; thus, the readers know which trends these are. Possibly use the sentence that talks about the reductions first and then highlights what others have done.

RG: We agree with this point and improved the introduction of this paragraph accordingly.

Line 448-450: Could you discuss the implications or the importance of considering land and soil requirements for coffee suitability. This will add more scientific knowledge on why modelling studies need to consider these parameters.

RG: We added a sentence to discuss the implications of considering land and soil requirements.

Line 506-509. The authors need to reconsider these sentences as they have also utilized bioclimatic data provided by WorldClim. Best highlighting that other studies lack the aspect of biophysical parameters?

RG: The use of bioclimatic and soil data was added to the sentence. The important difference is that our modelling approach is not based on present occurrence data, which is necessary for maximum entropy modelling.

Line 527-531: This sentence shows that climatic factors are the most crucial limiting factors for crop production. Basing on this, what could be the best recommendation for the future of crop modelling activities? Could a combination of climatic and soil/land parameters improve the model accuracy?

RG: We added a recommendation for combining climate and soil parameters in future modelling attempts to increase their significance.

Reviewer 2

Major revision

1. The major drawback of the methodological approach is that the authors presented the final model how a niche description approach. I think niche can have different types (Grinnell or Elton), and is necessary defined if the models are modeling fundamental niche or realized niche or. In addition, the modelling approach based on physiological response of species had been many restriction (please see: DOI:10.1515/eje-2015-0014

RG: The model applied in our study is neither a mechanistic, nor a correlative model as described in the literature attached. Our model is based on crop requirements described in literature that are ultimately based on environmental conditions in the growing areas. However, for our study we did not use crop occurrence data as described in the methodology. We just used the major producing countries to qualitatively cross-validate our suitability maps.

2. Oher question in methodological approach used is that the authors have not accounted for spatial biases in recording effort for either species or their ecosystems. It is likely that the process by which planted species are recorded within commercial programs systems or by the research community are very different to the processes by which not commercial zones in which natural conditions when the species can survive under natural suitability (main producing countries). The authors may be simply comparing the niche of “where people like to record planted species” with the niche of “where people like to record not planted species” and comparisons of geographic distributions are therefore very biased and the authors cannot conclude to what % of area of X species distribution will be reduce under climate change scenarios. In my opinion a solution to this is to use a target group to inform the background for the DSS models (one for planted species, one for not planted species), in addition to the accessibility criteria used, prior to comparing niches between panted and not planted species (https://doi.org/10.1890/07-2153.1). The papers by Broenimann et al. 2012 point out the importance of accounting for the differential sampling of different environmental conditions in the recorded distributions of two species when examining niche and distributions (https://doi.org/10.1111/j.1466-8238.2011.00698.x).

RG: as mentioned above, we did not use recorded occurrence data of the crops studied, but the available descriptions of their requirements. We also do not see the point of comparing planted and not planted species, since we only investigated cultivated crops. Furthermore, we are confident about our conclusions on % suitability changes. This is a relative estimation of suitability change due to climate change and not absolute. Of course, additional factors such as breeding efforts and adaptation measures will be critical, as pointed out in the manuscript.

3. The authors throw a lot of bioclimatic variables into the model which increases the chances of overfitting and also of detecting niche differences between species, but I can’t see which variables were selected.

RG: In Table 1 in the manuscript, it is clearly indicated which variables were used for the modelling of each of the three crops. The primary objective of the study was to estimate current and future suitability of the three crops and not to compare niche differences between the crop species.

4. The processes of DSS is a bit more complex, in which a multi-step must be made in order to improve the capacity of the models and not simply reproduce information of low quality. On the other hand the new approach to correct model under future we should incorporate the sources of variation in future model projections (https://doi.org/10.7717/peerj.6281 and https://doi.org/10.1071/CP19094).

RG: In the two publications referenced here, the MaxEnt approach is described and applied, a machine learning algorithm that uses presence-only data and environmental data. As already mentioned, we did not apply a machine learning approach in our study, but a simple multi-criteria evaluation. In our study, we took the uncertainty of future climate simulations into account by using results of a high number (14) of general circulation models (GCMs).

5. Since the DSS approach used and future scenarios present a good approximation when the analysis are local or regional, it is appropriate, practical and methodologically correct to do fine work for a worldwide approach?

RG: The analysis is based on global datasets and 14 general circulation models (GCMs), which are valid on a global scale.

6. More details are also needed to describe the analysis of the data so reviewers can ensure their appropriateness for the type of data presented. Additional attention to detail is needed to improve the overall quality of manuscript including the small detail about concept in DSS, modeling, data uses, and computation performance. Please also ensure the relevant parts of the manuscript are in the correct sections (i.e. results confined to the results!).

RG: We thoroughly revised the methodology part of the manuscript to be clear and transparent about our approach, which is also described in Jaisli et al. (2019) in detail, as referenced in the manuscript. The data analysis is clearly described in the methodology, including a detailed description of the datasets used and their sources in Table 2.

7. It is not clear when the authors described that GCMs were used, but I only identify the following models: BBC-CSM1-1, CCSM4, CNRM-CM5, GFDL-CM3, GISS-E2-R, HadGEM2-AO, HadGEM2-ES, IPSL-CM5A-LR, MIROC-ESM; CHEM, MIROCESM, MIROC5, MPI-ESM-LR, MRI-CGCM3 and NorESM1-M), but is not clear how were represented in the final raster. You can used ensemble model approach? If this concept is the best representation, so the methodology of the REA (Reliability Ensemble Averaging) approach is meaningless.

RG: As described in the methodology, three variables, namely mean annual temperature, mean minimum temperature of coldest month and mean annual precipitation were retrieved from the 14 GCMs indicated (see Table 2 for sources), and their mean values were calculated for the future modelling. They were matched with the respective crop requirements (see Table 1), and the rasters reclassified to get a raster of the 4 suitability classes (e.g. future temperature suitability for coffee).

Minor revision

Title

My suggestion is change the title, because is very general.

RG: The title was changed to ‘Expected global suitability of coffee, cashew and avocado due to climate change’ as suggested by Reviewer 1.

Abstract

Try to be specific and write a paragraph more informative, because is very confused the aim, and the relationship between the approach used (DSS approach and climate change). In addition, you can add more information based in data analysis (statistical, among others).

RG: We rephrased the information about the use of GCMs and emission scenarios. The approach used in the study is described in the methods section in detail. We tried to improve the clarity and comprehensibility of this description.

Introduction

In my opinion the introduction are poorly described. My suggestion to the authors is add more information about the use of the DSS algorithms (advantages and disadvantages with ecological niche models approach (ENM)). In addition, try to be more informative about of parameters associated with the algorithms used in DSS, how are used, your mathematical concepts, which are the predictors, how are obtained the predictors, advantages, disadvantages, limitations, among others. This is very important because actually the ENM approach is very popular with. On the other hang the black box model giving rise to models without biological sense.

My suggestion is that aim and hypothesis should be improve, because is not clear, in especial the roll of the approach used.

RG: As mentioned above in the general comments, we used a very simple modelling approach in our study without a complex underlying mathematical concept or algorithm (no black box model!). It is just an aggregation of layers of crop suitability maps of the individual climate, land and soil parameters. The description of the approach in the methodology section was improved and is also available in detail in Jaisli et al. (2019), which is also referenced in the manuscript.

Material and Methods

It is necessary to contextualize and clearly explain the bioclimatic, topographic variables used, the characteristics of the ecosystem system (e, j., area, currently status), explain more details about countrys sleected (areas, climatic, topographic variability), among others. The methods are superficially described, omitting basic information that is of utmost importance to guarantee reproducibility, a basic criterion in scientific research.

Is no clear the algorithms used. Is necessary add more details about this process as parametrization, performance computational, and evaluated results of model (calibration and validations). In addition, I can’t find the result of evaluation of model under climate change scenarios.

The model processes is a bit more complex, in which a multi-step must be made in order to improve the capacity of the models and not simply reproduce information of low quality. In addition, actually when algorithm is used must be considered an exhaustive evaluation of the parameters associated with the performance computational. On the other hand I suggest that to select the best model approach is necessary incorporate the sources of variation.

RG: The materials and methods section was thoroughly revised to make it clear what we did. As mentioned above, the analysis is not based on a mathematical model with independent and target variables. We applied a simple multi criteria evaluation approach with a suitability classification of several climate, land and soil parameters that were finally aggregated, based on the factor of lowest suitability for each land unit. All bioclimatic, land and soil variables and the datasets used in the study as well as their sources are indicated in Tables 1 and 2. In the description, it is clearly explained how they are used in the study. The selection of the main producing countries is based on the crops’ quantity produced in 2018 and justified in the chapter “Calculation of expected changes”. We did a qualitative evaluation of our results based on the delimitations of the main producing countries since we did not use or have global occurrence data available for validation. Apart from this, we did not use any other details about the countries selected in our study and therefore do not want to describe them in more detail in the methodology.

Result and discussion

Emphasize on explaining what advantages you have when using these spatial analysis strategies and not others.

RG: In the first paragraph of the discussion, we added the importance of integrating topographic and soil factors in crop suitability modelling.

How did you relate the genetic and spatial dimensions?

RG: In this study, we focused on the species level of the crops investigated and did not take into account any cultivar effects or any further analysis of the genetic dimension. Apart from this, all our analyses were done on a global level.

What is the current use of the areas with environmental problem under landslides?

RG: We did not address the problem or risk of landslides in this study.

How is the productivity systems in the area tested?

RG: We did not use any present occurrence data of the crops investigated for our analysis and therefore did not assess productivity systems. Generally, we did not address productivity, but suitability.

It is important to highlight the results and that these are incorporated into a management program for conservations and what environmental implications and sustainability indicators represent the use of these practices at the government level.

RG: Potential negative environmental impacts and sustainability issues are addressed in the conclusion section.

Conclusion

The author should improve the conclusion and focus on the most important data of the study. In addition, the conclusions presented do not represent the importance of the research work.

RG: We added a sentence to stress the importance of integrating topographic and soil information in crop suitability modelling. In our opinion, we cover the importance of the research work in the conclusion: i) integration of both climate and soil factors in crop modelling; ii) climate change impact on global coffee, cashew and avocado suitability, iii) implications for climate change adaptation of agroecosystems

Supplementary material

For better reproducibility, I suggest that the authors publish the codes and data as supplementary material or in a free repository (Gib-Hub).

RG: As mentioned in the submission, we will make the data available via the repository figshare.com

References

Review the correct format used by the Journal.

RG: As suggested by PLOS, we use the “Vancouver” style outlined by the International Committee of Medical Journal Editors (ICMJE)

Figures

Poor resolution and need to be compressible standalone

RG: The figures were uploaded in very high quality and can be downloaded from the file inventory of the online editorial manager system. The quality of the figures only got poor in the automatically created PDF manuscript.

Submitted filename: Response to Reviewers.docx

Click here for additional data file.

30 Nov 2021

Submitted filename: Reviewers comments - Copy.docx

Click here for additional data file.

13 Dec 2021

Dear Editor, dear Reviewers,

We are glad that you appreciate our improvements made in the first revision. Thank you for your second feedbacks. We have addressed and answered the second reviewer comments as you can find below. We have uploaded our revised manuscript with and without track changes to the editorial management system.

Based on the journal requirements mentioned in the decision letter, we carefully reviewed the reference list again. Two references were deleted: Duncan (2001) and WorldClim (2016). The relevant information of the former is also included in Widiatmaka et al. (2014). The source of the WorldClim datasets (website) is given in Table 2 and in the text (line 161). Two references were updated with a more recent publication: IPCC (2021), Charre-Medellín (2021). The details in the list of references were updated and harmonized for the following entries: Benson et al. (2016), Wolstenholme (2013), Dudley (2008), UNEP-WCMC (2018), Balogoun et al. (2015). Then, there are a few publications that are not peer reviewed, but still relevant and trustworthy for the definition of the crop requirements for the modelling and for the interpretation of our results: Benson et al. (2016), CIAT (2011), Hombunaka et al. (2016), Kuit et al. (2004) and Ricau (2019).

With this we hope that our manuscript is now ready for publication.

Yours sincerely,

Roman Grüter

Reply to the reviewer comments

Reviewer 1

The authors have managed to address our concerns and now the article is great. It can be Accepted for publication.

Minors.

Line 17: Remove “confirming previous findings”.

RG: Thank you for this comment. We agree and removed the statement.

Line 45: I think what you mean here is that Coffee arabica is highly sensitive to and dependent on weather variables not climate change per se?

RG: Coffee arabica has proven to be highly sensitive to climate change as was for example stated by Bunn et al. (2015). Of course, this is based on the fact that C. arabica is heat sensitive and might thus suffer under warmer temperatures (or weather). We therefore think that the statement is still correct as it stands.

Line 195: Remove “still” from the sentence.

RG: We agree.

Line 232: Climate suitability or climate factors? Please check this to avoid confusing the reader.

RG: Thank you for this hint. “Climate factors” is more precise.

Line 233: Check “in Africa or East Africa” revise this.

RG: Thank you for this comment. We checked our data and removed “East Africa”

Reviewer 2

Under the current scheme, I consider that sufficient contributions were made to be considered for publication. Personally, I do not agree with some answers. My last suggestion would be associated with the current suitability maps changing the blue color for another more informative as red type, indicating danger.

RG: Thank you for your feedback. Regarding the map colours selection, we have intensively discussed and thoroughly tested different colours for the suitability classes and the future changes. We would like to keep the selected colour scheme for the following reasons. First, we do not necessarily relate the suitability class S3 (marginally suitable) with “danger” or “risk”. It might as well indicate “opportunity” or “chance”, as for example in regions of future expansion of suitable growing areas. In these potential future production locations, the suitability might further increase with ongoing climate change in the future. Second, we use different red type colours in the figures where the change in suitability is clearly negative, indicating “danger” (e.g. in Fig. 4, 5, 7, 8). Third, the same colour scheme (green, blue, grey) for the same suitability classes has already been applied in our first publication (Jaisli et al. 2019) based on the same multi-criteria evaluation approach. We would like to be consistent with the visualizations in this publication.

Submitted filename: Response to Reviewers.docx

Click here for additional data file.

15 Dec 2021

Expected global suitability of coffee, cashew and avocado due to climate change

PONE-D-21-13411R2

Dear Dr. Grüter,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Abel Chemura

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

5 Jan 2022

PONE-D-21-13411R2

Expected global suitability of coffee, cashew and avocado due to climate change

Dear Dr. Grüter:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Abel Chemura

Academic Editor

PLOS ONE