3.2 Model performance and significant environmental variables
We calculated the model accuracy of ANN, CTA, FDA, GBM, GLM, MARS, MaxEnt, RF, and EM using ROC, TSS, and KAPPA values (Fig. S4). The mean ROC of the eight models were 0.885, 0.943, 0.948, 0.960, 0.947, 0.952, 0.951, and 0.983, respectively (Table S2). The mean TSS of eight single models were 0.666, 0.818, 0.773, 0.809, 0.775, 0.778, 0.785, and 0.880, respectively. The mean KAPPA of eight single models were 0.685, 0.810, 0.768, 0.808, 0.777, 0.782, 0.778, and 0.880, respectively. The mean ROC, TSS, and KAPPA of the EM were 0.984, 0.851, and 0.856, respectively, which were higher than those of single models, indicating that the EM predicted PGSH of S. halepense was reliable.
We analyzed the significant environmental variables to predict the PGSH of S. halepense using EM, and the mean contributions are listed in Table 1. Precipitation of coldest quarter (bio19, 0.317) was the most significant environmental variables, followed by precipitation of driest quarter (bio17, 0.108), min temperature of coldest month (bio6, 0.091), annual precipitation (bio12, 0.085), mean diurnal range (mean of monthly (max temp-min temp) (bio2, 0.078), max temperature of warmest month (bio5, 0.067), precipitation seasonality (coefficient of variation) (bio15, 0.048), and LUC (0.019). The response curves of significant variables are shown in Fig. S7. When the suitability probability achieved the maximum values, the survival probability of S. halepense was more reliable. Bio19 and bio17 were the most significant environmental variables for the predicted PGSH of S. halepense in the three land-use types, achieving approximately the maximum values at 200 mm and from 140 mm, respectively.
3.3 The PGSH of S. halepense under the current and future climate scenarios
The PGSH of S. halepense extracted from the three land-use types was mainly distributed in eastern, southeastern, central, and western Asia; western Europe; southern North America; southeastern South America; east-central, southwestern, and northern Africa; and southeastern and southwestern Oceania projected in the near current climate and future scenarios (Figs. 2 and 3). Compared with the near current climate, the predicted global total cropland and urban areas would increase under future scenarios, yet the predicted total grassland area would decrease to a certain extent (Table 2). As a result, the PGSH of S. halepense notably increases in the three land-use types, achieving a maximum under SSP5-8.5 in the 2030s and the 2050s.
In cropland, the PGSH of S. halepense was located in eastern Asia (eastern Japan, South Korea), southeastern Asia (southern and eastern China), central Asia (southern and northern India, northeastern Pakistan, Afghanistan, southeastern Uzbekistan, western Tajikistan, and southern Kazakhstan), and western Asia (Iran, northern Iraq, northern Syria, and Turkey,), near total western countries in Europe, southern North America (central and western United States, Mexico, Cuba, and the Dominican Republic), southeastern South America (southern and eastern Brazil, center Bolivia, Paraguay, eastern Argentina, and Chile), northern Africa (northern Morocco, northern Algeria, northern Tunisia), east-central Africa (Ethiopia, Kenya), southwestern Africa (Uganda, Malawi, Mozambique, Zimbabwe, and South Africa), southeastern and southwestern Oceania (southeastern and southwestern Australia).
In grassland, the PGSH of S. halepense was mainly located in central Asia (Pakistan, Afghanistan, Tajikistan, Uzbekistan, and Kazakhstan), western Asia (Turkey), western Europe (northern Mediterranean Sea and western coastal), southern North America (the United States and Mexico), southeastern South America (central Bolivia, eastern and southwestern Brazil, Paraguay, Uruguay, and northeastern Argentina), southern Africa (Ethiopia, eastern Mozambique, Lesotho, southern Madagascar, and South Africa), southeastern and southwestern Oceania (Australia and New Zealand). However, in urban areas, the PGSH of S. halepense was widely distributed worldwide.
Table 2
Potential global suitable habitats (PGSH) of Sorghum halepense in three land-use types between the near current climate and different future scenarios
Land-use (×104 km) | cropland | grassland | urban |
suitable area | total area | | suitable area | total area | | suitable area | total area | |
Near current | 921.99 | 2177.42 | 42.34% | 280.41 | 1675.80 | 16.73% | 50.42 | 61.89 | 81.47% |
2030s,SSP1-2.6 | 1026.27 | 2150.03 | 47.73% | 297.00 | 1555.53 | 19.09% | 75.24 | 89.92 | 83.67% |
2030s,SSP2-4.5 | 1073.00 | 2279.08 | 47.08% | 317.87 | 1638.99 | 19.39% | 72.92 | 87.09 | 83.73% |
2030s,SSP5-8.5 | 1079.24 | 2340.77 | 46.11% | 315.38 | 1613.94 | 19.54% | 78.81 | 93.78 | 84.04% |
2050s,SSP1-2.6 | 1060.59 | 2178.33 | 48.69% | 281.28 | 1448.65 | 19.42% | 85.00 | 101.57 | 83.69% |
2050s,SSP2-4.5 | 1091.27 | 2345.28 | 46.53% | 315.51 | 1577.10 | 20.01% | 82.24 | 99.20 | 82.90% |
2050s,SSP5-8.5 | 1113.91 | 2377.42 | 46.85% | 334.58 | 1609.21 | 20.79% | 95.63 | 112.74 | 84.82% |
The suitable areas of S. halepense in cropland were mainly located in Europe under the near current climate, approximately 311.95 × 104 km, accounting for 33.83% of the PGSH of S. halepense, followed by Asia (228.21 × 104, 24.75% ), South America (127.91 × 104, 13.87%), North America (117.19 × 104, 12.71%), Africa (79.29 × 104, 8.60%) and Oceania (54.21 × 104, 5.88%). In the 2030s, future suitable cropland areas achieved the maximum under SSP2-4.5 in Europe, approximately 379.84 × 104 km, accounting for 33.83% of the PGSH of S. halepense, followed by Asia (249.93 × 104, 24.35% ), South America (151.74 × 104, 14.38%), North America (147.60 × 104, 14.38%), Africa (85.07 × 104, 8.29%) and Oceania (55.69 × 104, 5.43%). In the 2050s, future cropland areas of S. halepense were mainly distributed in Europe, approximately 407.34 × 104 km, accounting for 36.57% of the PGSH of S. halepense, followed by Asia (254.32 × 104, 22.83%), North America (160.65 × 104, 14.42%), South America (148.04 × 104, 13.29%), Africa (87.94 × 104, 7.89%), and Oceania (52.28 × 104, 4.69%).
Suitable areas of S. halepense in grassland were mainly located in North America under the near current climate, approximately 95.63 × 104 km, accounting for 34.10% of the PGSH of S. halepense followed by Europe (55.47 × 104, 19.78% ), South America (46.70 × 104, 16.65%), Oceania (43.16 × 104, 15.39%), Africa (19.51 × 104, 6.96%) and Asia (18.2 × 104, 6.49%). In the 2030s, future suitable grassland areas achieved the maximum under SSP5-8.5 in North America, approximately 116.18 × 104 km, accounting for 36.84% of the PGSH of S. halepense followed by Europe (66.83 × 104, 21.19% ), South America (43.55 × 104, 13.81%), Oceania (39.18 × 104, 12.42%), Asia (25.53 × 104, 8.09%) and Africa (22.30 × 104, 7.07%). In the 2050s, future grassland areas of S. halepense were mainly distributed in North America, approximately 131.39 × 104 km, accounting for 39.27% of the PGSH of S. halepense, followed by Europe (81.90 × 104, 24.48%), South America (39.69 × 104, 11.86%), Oceania (33.83 × 104, 10.11%), Asia (26.96 × 104, 8.06%), and Africa (18.82 × 104, 5.62%).
Suitable areas of S. halepense in urban were mainly located in Europe under the near current climate, approximately 14.98 × 104 km, accounting for 29.71% of the PGSH of S. halepense, followed by North America (14.18 × 104, 28.12% ), Asia (13.00 × 104, 25.78%), South America (3.02 × 104, 5.99%), Africa (2.50 × 104, 4.96%) and Oceania (1.24 × 104, 2.46%). In the 2030s, future suitable urban areas achieved the maximum under SSP5-8.5 in North America, approximately 25.01 × 104 km, accounting for 31.73% of the PGSH of S. halepense followed by Europe (20.47 × 104, 25.97% ), Asia (20.05 × 104, 25.44%), Africa (4.81 × 104, 6.10%), South America (4.06 × 104, 5.15%) and Oceania (2.46 × 104, 3.12%). In the 2050s, future urban areas of S. halepense were mainly distributed in North America, approximately 34.10 × 104 km, accounting for 35.62% of the PGSH of S. halepense, followed by Europe (24.43 × 104, 25.52%), Asia (21.42 × 104, 22.37%), Africa (5.65 × 104, 5.90%), South America (4.17 × 104, 4.36%), and Oceania (3.67 × 104, 3.83%).
3.4 Changes in PGSH of S. halepense
Future increased cropland habitats were virtually distributed in the North Hemisphere, including eastern, central, and western Asia (China, Iran, Kazakhstan, and Turkey); southwestern Europe (Russia and Byelarus); and central North America (Canada and the United States); and partly in South America (Brazil) and Africa (Cote d’ Ivoire) (Fig. 4). The suitable area achieved its maximum under SSP5-8.5 in the 2050s, approximately 172.09 × 104 km2. Future decreased cropland habitat was mainly distributed in central Asia (Pakistan and India), sporadically in southern North America (Mexico), central South America (Argentina, Bolivia, Brazil, and Paraguay), and eastern and central Africa (Ethiopia, Congo, and Zimbabwe). The suitable area achieved the minimum under SSP5-8.5 in the 2050s, approximately 29.21 × 104 km2.
Future increased grassland habitats were virtually distributed in the Northern Hemisphere, including central and western Asia (Afghanistan, Kyrgyzstan, Tajikistan, and Turkey); southern and southwestern Europe (Austria, Norway, Russia, San Marino, Slovenia, Switzerland, and the United Kingdom); central North America (the United States), and partly in southern Africa (Namibia and South Africa) (Fig. S5). The suitable area achieved the maximum under SSP5-8.5 in the 2050s, approximately 72.58 × 104 km2. Future decreased grassland habitat was virtually distributed in eastern, central, and western Oceania (Australia). The suitable area achieved the minimum under SSP5-8.5 in the 2050s, approximately 12.64 × 104 km2.
There was no decrease in suitable urban habitats in the future scenarios, which were distributed in the Northern Hemisphere, including eastern Asia (China); southern Europe (Russia); and central North America (United States) (Fig. S6). The suitable area achieved the maximum under SSP5-8.5 in the 2050s, approximately 3.62 × 104 km2.
3.5 Trend of suitable probability of S. halepense with latitudinal gradients
Compared with the near current climate, the PGSH of S. halepense in croplands tended to have high latitudinal gradients with a higher suitability probability under SSP1-2.6, SSP2-4.5, and SSP5-8.5, in the 2030s and the 2050s (Fig. 5). In the Northern Hemisphere, the suitable probability of S. halepense in croplands was positioned at 28°N–39°N under a near current climate, while it increased at 31°N–42°N, 35°N–50°N, and 58°N–61°N under future scenarios. In the Southern Hemisphere, a suitable probability of S. halepense in croplands was positioned at 22°S–43°S under a near current climate, while remaining firm at 22°S–42°S in future scenarios.
The global suitability probability of S. halepense in grasslands increased slightly in higher latitudinal gradients under SSP1-2.6, SSP2-4.5, and SSP5-8.5 in the 2030s and the 2050s compared with the near current climate (Fig. 6). In the Northern Hemisphere, the suitable probability of S. halepense in grasslands was positioned at 20°N under the near current climate, while it increased at 20°N–26°N under future scenarios. In the Southern Hemisphere, the suitable probability of S. halepense in grasslands was positioned at 30°S–41°S under the near current climate, while it increased at 31°S–46°S under future scenarios.
The global suitability probability of S. halepense in urban areas increased dramatically in higher latitudinal gradients under SSP1-2.6, SSP2-4.5, and SSP5-8.5 in the 2030s and the 2050s compared with the near current climate (Fig. 7). In the Northern Hemisphere, the suitable probability of S. halepense in urban areas was positioned at 20°N–52°N under a near current climate, while it increased at 19°N–61°N under future scenarios. In the Southern Hemisphere, a suitable probability of S. halepense in urban areas was positioned at 22°S–33°S under the near current climate, while it increased at 26°S–41°S under future scenarios.