Environmental factors

Prediction of global marginal land resources for Pistacia chinensis Bunge by a machine learning method

Accuracy of global suitability prediction for P. chinensis

Layering the global occurrence points of our plant on its environmental suitability map (Fig. 1), we find the consistency that these points of existing plants mostly appear in regions of high suitability, which could indicate the good performance of BRT model.

Figure 1

Distribution of global P. chinensis occurrence records layering over the global environmental suitability for P. chinensis to visually verify the accuracy of model prediction [Figure was created with ArcGIS Desktop (ESRI, Inc, Version 10.2, https://desktop.arcgis.com)].

To further evaluate the accuracy of our simulation, we calculated the statistical index under tenfold cross-validation procedure, obtaining the training data AUC = 0.995 ± 0.001 and validation data AUC = 0.995 ± 0.003, which could give credibility to the outputs of the BRT model .

In addition, we used the standard deviation values ​​to quantify the uncertainty of the spatial prediction. The visualized quantification in Fig. 2 presented relatively low uncertainty, which could further validate the results of model simulation.

Figure 2
figure 2

Visualized uncertainty in spatial prediction on the basis of standard deviation values ​​computed for each pixel across the model ensemble. [Figure was created with ArcGIS Desktop (ESRI, Inc, Version 10.2, https://desktop.arcgis.com)].

Relative contribution of the environmental covariates

Given that the accuracy of our model prediction has been validated through multiple means, the reliability of variables’ relative contribution in the simulating process were therefore guaranteed. Table 1 shows the relative contribution of the environmental covariates related to predicting the global distribution of P. chinensisconcurring with our expectation that climate variables substantially influence identifying global land resources for growing P. chinensis. The top four variables determining the suitability for P. chinensis all fall into the category of representing climate conditions, which are mean annual water vapor (58.76% [95%CI 57.54–59.98]), average annual temperature (24.33% [95%CI 23.08–25.58]), average solar radiation (5.49% [95%CI 4.67–6.31]), annual cumulative precipitation (5.26% [95%CI 4.17–6.35]), accounting for totally 93.84% of the contribution. By comparison, soil and topographical factors exhibit far less importance: soil water content (3.90% [95%CI 3.26–4.53]), soil class (0.41% [95%CI 0.28–0.53]), soil depth (0.24% [95%CI 0.28–0.53]), elevation (1.34% [95%CI 1.14–1.54]), slopes (0.27% [95%CI 0.22–0.33]).

Table 1 The relative contribution of each spatial predictor variable.

Potential marginal land resources suitable for P. chinensis

Global distribution of potential marginal land resources for P. chinensis

A necessary step prior to estimating the worldwide potential marginal land available for proper cultivation of P. chinensis is evaluating the global suitability for growing this plant.

In the global map (Fig. 3), we distinguished the environmentally suitable regions for growing P. chinensis, covering a broad span of latitudes from 45 degrees north and 45 degrees south. Generally, the map exhibits a descending trend in suitability from coasts toward inland, excluding a few suitable regions in Africa and areas along borders between countries.

Figure 3
figure 3

Predicted global environmental suitability for P. chinensis ranging from high (dark greenish blue) to low (light gray) [Figure was created with ArcGIS Desktop (ESRI, Inc, Version 10.2, https://desktop.arcgis.com)].

To identify potential land resources for planting, we determined whether a 5 × 5 km2 grid cell is suitable for P. chinensis by setting 0.5 as the value of the threshold. After that, when distinguishing potential marginal land out of all land resources for growing P. chinensis, it is our understanding that in case of compromising environment preservation and productivity, we should include only a few land-use types: savannas, shrublands, and grasslands. The resulting map is presented in Fig. 4.

Figure 4
figure 4

Map of the potential marginal land resources suitable for P. chinensis planting classified by land use type [Figure was created with ArcGIS Desktop (ESRI, Inc, Version 10.2, https://desktop.arcgis.com)].

The quantified results show that globally, there are 1311.85 million hectares of marginal land in total for proper P. chinensis cultivation, mostly distributed in Southern Africa, the southern part of North America, the western part of South America, Southeast Asia, Southern Europe, and eastern and southwest coasts of Oceania.

Africa, though no P. chinensis has been found there yet, has the greatest amount of land resources for the planting, which is 528.19 million hectares in total, for growing this plant. According to the map, the majority of Angola, Zambia, Mozambique, Tanzania, Zimbabwe, and Malawi, and a considerable proportion of Zaire, South Africa, Ethiopia, Madagascar, and Kenya has potential for future cultivation. South America has the next largest amount of marginal land to Africa. The total 304.44 million hectares are mainly distributed in West Brazil, central Argentina, Uruguay, Paraguay. We also find a relatively small portion of land resources in Peru, Bolivia, Chile, Ecuador, Venezuela, and Colombia. North America possesses 222.99 million hectares of plantable marginal land for our plant, mainly found in the United States and Mexico. Asia, where P. chinensis is native to, only owns third to last marginal land for growing it. The majority of the total 106.18 million hectares of land resources are from countries in Southeast Asia, including China, Myanmar, and Vietnam. Also, part of Turkey, India, Iran, Iraq, and the bordering region between Pakistan and Afghanistan are also found suitable for potential cultivation. In Oceania, marginal land resources are mostly distributed in coastal regions. The total amount of land resources is 101.04 million hectares, of which 97.99% are found in Australia. In Europe, we find the least land resources among all continents, which are mainly distributed in Spain, Portugal, Ireland, the United Kingdom, Greece, Italy, and France, amounting to 49.01 million hectares.

Global land-use composition of potential land resources for P. chinensis

Classification of the global marginal land resources by land-use type could reveal the land-use structure of available land for P. chinensis all over the globe, indicating that the land resources mostly consist of woody savannas, savannas, grasslands, and open shrublands, leaving only a tiny fraction of marginal land as closed shrublands.

Our calculation shows that woody savannas take up the largest proportion of land resources, accounting for 37.72% (494.80 million hectares) of the total. In addition, savannas (358.71 million hectares), grasslands (257.10 million hectares), and open shrublands (198.65 million hectares) share fair proportions of all the plantable land resources, which are 27.34%, 19.60%, and 15.14%, respectively. In comparison, close shrublands (2.59 million hectares) make an almost negligible contribution to the marginal land for planting, sharing 0.20% of the total.

Moreover, we managed to calculate the percentage of all types of land use for planting P. chinensis in each continent, respectively, presented in the pie charts made below (Fig. 5). It is not surprising that we find woody savannas in the leading position in marginal land resources of several continents, including Africa, Asia, Europe, and Oceania. The following type is savannas, comprising the largest proportion of marginal land in South America.

Figure 5
figure 5

Pie charts depicting the size (thickness of the circles) and land-use composition (illustrated by different greens on the circles, representing different types of land use) of available marginal land in each continent for planting P. chinensis [Figure was created with Microsoft Excel (Microsoft, Version 16.0, https://www.microsoft.com/en-us/microsoft-365/excel) and Adobe Illustrator (Adobe Inc, Version 25.0, URL: https://www.adobe.com/products/illustrator.html)].

Furthermore, quantification results for countries with more than 10 million hectares of land resources for P. chinensis Cultivation in all continents are itemized in descending order by size in Table 2. The table shows that 9 of the total 21 countries in the list are from Africa, amounting to 443.95 million hectares, accounting for 33.84% of the total land resources. However, the country with the largest amount of available marginal land is Brazil (190.58 million hectares) from South America, which is mainly comprised of savannas. In terms of the land-use type in the leading position of these countries, 7 of the 21 countries are primarily distributed with woody savannas (236.83 million hectares in total), while 6 of them are mainly distributed with grasslands (156.90 million hectares), followed by 5 countries with savannas (235.73 million hectares), 3 countries with open shrublands (87.58 million hectares). The results on the national scale are generally consistent with that on the continental scale, indicating the substantial share of countries in Africa as well as woody savannas in total available marginal land.

Table 2 Potential marginal land resources suitable for P. chinensis in major global regions and the countries with the area of ​​suitable land > 10 million hectares.

Related Articles

Leave a Reply

Your email address will not be published.

Check Also
Back to top button