*Article* **Effects of Land Use and Cropping on Soil Erosion in Agricultural Frontier Areas in the Cerrado-Amazon Ecotone, Brazil, Using a Rainfall Simulator Experiment**

**Marco Aurélio Barbosa Alves <sup>1</sup> , Adilson Pacheco de Souza <sup>2</sup> , Frederico Terra de Almeida <sup>2</sup> , Aaron Kinyu Hoshide 3,4, Handrey Borges Araújo <sup>2</sup> , Apoliano Francisco da Silva <sup>2</sup> and Daniel Fonseca de Carvalho 5,\***


**Abstract:** Agricultural soils provide ecosystem services, but the removal of natural vegetation reduces water infiltration capacity, increasing surface runoff. Thus, monitoring erosion is critical for sustainable agricultural management. Sediment losses and surface runoff were evaluated using a simulated rainfall of 75 mm/h in areas with crops and pastures in both the Caiabi River and Renato River sub-basins of the Teles Pires River watershed in Mato Grosso State, Brazil. In both the Caiabi and Renato sub-basins, data were collected from 156 observations in the upper, middle, and lower regions where (1) soybeans, (2) maize, and (3) pasture were grown alone, with another crop, or with soil that was scarified. Erosion occurred independent of soil texture and was closely related to the management and use of systems involving fewer crops and more soil scarification, regardless of sub-basin location. In uncovered, scarified soil, the soil losses from erosion were greater compared to covered soil, regardless of sub-basin and sub-basin region. In the Renato River sub-basin, soil losses in cultivated areas not planted with crops but with scarification were 66.01, 90.79, and 60.02 g/square meter in the upper, middle, and lower regions, respectively. Agricultural producers need to increase the planting of crops throughout the year and minimize soil disturbance, which will reduce soil erosion and improve sustainability.

**Keywords:** Cerrado-Amazon; crops; geoprocessing; GIS; land use; mapping; rainfall simulator; satellite images; soil erosion

### **1. Introduction**

Soil is essential for both macro- and microscopic life and provides ecosystem services, ensuring a stock for carbon, nutrient cycling, water retention and infiltration, and food production [1]. However, most soils in Brazil and around the world are compromised due to contamination, pollution, and erosive processes that have contributed to soil degradation [2,3]. In addition to adversely affecting farming production and ecosystem services, degraded soils contribute to global warming and hydrological extremes since they have lower capacities to store carbon and facilitate the infiltration and storage of water [4–6].

Erosion is one of the main causes of soil loss in Brazil and around the world, specifically water erosion [2,7], which is responsible for detaching soil particles, transporting them, and depositing them in areas at lower altitudes. Therefore, soil carbon stocks are exposed and lost through decomposition, mineralization, and transport, while mineral particles

**Citation:** Alves, M.A.B.; de Souza, A.P.; de Almeida, F.T.; Hoshide, A.K.; Araújo, H.B.; da Silva, A.F.; de Carvalho, D.F. Effects of Land Use and Cropping on Soil Erosion in Agricultural Frontier Areas in the Cerrado-Amazon Ecotone, Brazil, Using a Rainfall Simulator Experiment. *Sustainability* **2023**, *15*, 4954. https://doi.org/10.3390/ su15064954

Academic Editors: Maurizio Lazzari and Teodor Rusu

Received: 6 January 2023 Revised: 1 March 2023 Accepted: 7 March 2023 Published: 10 March 2023

**Copyright:** © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

silt up rivers, streams, and lakes [8]. While soil compaction, surface crusting, and soil erosion can be reduced by soil organic matter, which can also provide nutrients to plants, soil erosion can also be reduced by vegetative cover [9]. Uncovered and/or badly managed soils accelerate this process since the absence of vegetative cover (e.g., crops) promotes direct exposure to raindrops [10]. Factors such as the duration of rain and the slope of the area are also responsible for increases in soil losses by erosion, as they influence surface runoff [11,12], while soil turning (scarification) promotes the breakdown of aggregates, making the particles more susceptible to the erosive process [13].

In order for Brazil to reduce its soil and water losses, producers need to adopt management techniques such as the no-tillage system, contour lines, and pasture management with rotational grazing. In contrast, areas covered with native forest, crop straw residues, and pasture tend to have less soil loss than areas with exposed and tilled soils [13–15], minimizing the environmental damage from soil erosion. Undisturbed natural vegetation can also minimize soil erosion. For example, in an area with native forest located in the Amazon biome, the measured surface runoff and soil losses were close to zero due to the greater rainfall infiltration capacity of the soil [16]. The removal of natural vegetation from forest areas and its transformation into crops and pastures cause soil and water losses and the subsequent destruction of biodiversity, due to the reduction of carbon stocks [11]. When this removal is accompanied by farming practices involving continuous soil disturbance, such as scarification, these losses are even more marked since soil disruption favors erosive processes [2,17]. Long-term research studies over decades demonstrate that shifting grasslands to progressively more disturbed or scarified (e.g., tilled) cropping systems can decrease soil microbiome diversity, make soil microbiome processes more variable, and increase the prevalence of pathogenic soil organisms [18].

Located in the Cerrado-Amazon ecotone and holding great socioeconomic and environmental importance, the upper and middle Teles Pires regions in Mato Grosso State, Brazil, are agricultural frontiers that have undergone constant native vegetation removal and are susceptible to severe soil, water, and nutrient losses [17]. These areas represent the dynamics of land occupation and land use in the region; soil erosion monitoring is an important tool for decision-making, especially regarding sustainable agricultural production. Due to difficulties in collecting and quantifying the runoff in the experimental plots and the temporal variability of the intensity of natural rainfall, rainfall simulators have been used in different classes, uses, and occupations [12] and in studies about water infiltration into the soil [19], revealing distinct classifications and operational characteristics [20–22]. Additionally, the simulation equipment allows control of the duration and rainfall intensity, and the size and speed of the droplets impacting the soil [20]. Known as sprinkler infiltrometers, these simulators take less time and have a lower cost when conducting field research, compared to experimental natural rainfall plots, and can accurately control water input, thereby reducing the errors associated with natural rainfall variability [23].

The use of rainfall simulators to measure soil erosion in agricultural production systems in the Cerrado-Amazon transition region is unprecedented. Our research can serve as a guide for agricultural producers, ranchers, the public agencies of agrarian policies, and non-governmental agencies seeking to improve the management of soil and water resources in the tropics. The goal of our study is to quantify soil losses when under different agricultural land uses in the Caiabi and Renato River sub-basins of the Teles Pires River watershed in Mato Grosso State, Brazil.

#### **2. Materials and Methods**

#### *2.1. Study Area*

The Teles Pires River watershed is located in the states of Mato Grosso and Pará, Brazil (Figure 1). Despite being located, in hydrological terms, in the Amazon region, the Teles Pires watershed has variable vegetative cover, with its upper and lower regions in the Cerrado and Amazon biomes, respectively. The upper and middle Teles Pires regions correspond to 26.2 and 57.71% of the basin area; they have a population density of 45.9 and 27.5% of the total population of the basin and are responsible for 66.3 and 18.7% of the gross domestic product (GDP) obtained in the two Teles Pires River areas, respectively. The two regions together represent more than 17% of the GDP of Mato Grosso. Analyses of the soil and water losses were conducted in two drainage sub-basins, the Caiabi River (upper) and Renato River (middle), with drainage areas of approximately 500 and 1450 km<sup>2</sup> , respectively (Figure 1). the Cerrado and Amazon biomes, respectively. The upper and middle Teles Pires regions correspond to 26.2 and 57.71% of the basin area; they have a population density of 45.9 and 27.5% of the total population of the basin and are responsible for 66.3 and 18.7% of the gross domestic product (GDP) obtained in the two Teles Pires River areas, respectively. The two regions together represent more than 17% of the GDP of Mato Grosso. Analyses of the soil and water losses were conducted in two drainage sub-basins, the Caiabi River (upper) and Renato River (middle), with drainage areas of approximately 500 and 1450 km2, respectively (Figure 1).

The Teles Pires River watershed is located in the states of Mato Grosso and Pará, Brazil (Figure 1). Despite being located, in hydrological terms, in the Amazon region, the Teles Pires watershed has variable vegetative cover, with its upper and lower regions in

*Sustainability* **2023**, *15*, x FOR PEER REVIEW 3 of 23

*2.1. Study Area* 

**Figure 1.** The Teles Pires River watershed and the location of the Caiabi and Renato River basins (data source: Ref. [24]).

Regarding the soils in the Caiabi sub-basin, the most recurrent classes are the inceptisols, oxisols, and entisols [24,25], formed from metasedimentary rocks belonging to the Cuiabá Group and the Raizama and Araras formations (Upper Paraguai Group). Conversely, the Renato sub-basin presents pedological characterization, with ultisols, oxisols, entisols, and Plinthic oxisols [24,25], formed from the granitic and rhyolitic rocks of the Juruena magmatic arc, with several gold occurrences: sandstones from the Dardanelos Formation and the Beneficiente Group, with sandstones, siltstones, and claystones from the Upper Tapajós basin (Capoeiras Formation).

In the Caiabi River sub-basin, located in the Cerrado-Amazon ecotone, monoculture areas of soybean (*Glycine max* L.), immediately followed by maize (*Zea mays* L.) in the same production year succession, are predominant, while in the Renato River sub-basin, there is also a predominance of native Amazon rainforest under forest management. According to the Köppen climate classification model, the climate of the region is the Aw type,

considered a tropical wet and dry climate, with a dry period between June and August [26]. The mean annual temperature is 25 ◦C, with a minimum temperature below 16 ◦C and a maximum temperature above 34 ◦C. The mean annual precipitation varies by approximately 1900 mm [27]. In 2019, the percentages (%) of land use and occupation in the Caiabi subbasin were most for crops (51.96%) followed by water (22%), native forest (31.78%), pastures (3.6%), and burned areas (0.08%), respectively. Likewise, in the Renato River hydrographic sub-basin, these same classes of land use and occupation were highest for native forest (61.32%) followed by water (27%), pastures (9.7%), crops (9.57%), and burned areas (0.18%), respectively (Figure 2). *Sustainability* **2023**, *15*, x FOR PEER REVIEW 5 of 23

**Figure 2.** Location of the sampling points of simulated rainfall and land use, along with occupation in the Caiabi and Renato sub-basins, which are tributaries of the Teles Pires river. **Figure 2.** Location of the sampling points of simulated rainfall and land use, along with occupation in the Caiabi and Renato sub-basins, which are tributaries of the Teles Pires river.

Rainfall simulator experimental tests were carried out on 12 farms, two in each region of each sub-basin. In general, in the region each property operates in, only one type of farming activity occurs (either cropping or pasture). In the Caiabi river sub-basin, 84 tests were carried out on six farms, distributed as follows: (1) on the farms occupied by soybeans, only four soil covers were evaluated at each sampling point (soybean + straw, straw only, uncovered soil, and scarified soil), totaling 48 tests; (2) on the farms occupied by pastures, three soil covers were evaluated at each sampling point (pasture, uncovered soil, and scarified soil), totaling 36 tests. In the Renato sub-basin, 72 tests were carried out, since in both of the areas occupied by corn and by pastures, three soil covers were evaluated (with vegetation, uncovered soil, and scarified soil), resulting in 36 tests in areas occupied by corn and 36 tests in areas occupied by pastures. In Figure 2, the four repetitions evaluated in the same farm are not so evident, due to the spatial scale. However, the tests with simulated rainfall were carried out considering a minimum distance of 500 m between repetitions (examples can be seen in Figure 3). The areas selected for carrying out these tests had spent at least 5 years under the same land use. *Sustainability* **2023**, *15*, x FOR PEER REVIEW 6 of 23

**Figure 3.** Locations of the simulated rainfall points in (**a**) Continental Farm (headwater sub-basin region), (**b**) Aremisa III Farm (middle sub-basin region), with land use of maize crops and pasture, in the Renato sub-basin, (**c**) São José Farm, and (**d**) Taguá Farm, both in the middle region of the Caiabi sub-basin. The scales in (a, b, c, and d) are 1:54,200, 1:361,000, 1:55,000, 1:62,300, and 1:90,500 **Figure 3.** Locations of the simulated rainfall points in (**a**) Continental Farm (headwater sub-basin region), (**b**) Aremisa III Farm (middle sub-basin region), with land use of maize crops and pasture, in the Renato sub-basin, (**c**) São José Farm, and (**d**) Taguá Farm, both in the middle region of the Caiabi sub-basin. The scales in (**a**–**d**) are 1:54,200, 1:361,000, 1:55,000, 1:62,300, and 1:90,500 cm, respectively.

cm, respectively. The evaluations of soil and water losses occurred for different agricultural crops, depending on the specific cultivation calendar for soybeans and maize in the state of Mato Grosso. Figure 4 summarizes the rainfall in the two hydrographic sub-basins being evaluated, along with the arrangement of crops in the region, regulated by the federal ordinances of the Secretaria de Política Agrícola (SPA) of the Ministério de Agricultura, The evaluations of soil and water losses occurred for different agricultural crops, depending on the specific cultivation calendar for soybeans and maize in the state of Mato Grosso. Figure 4 summarizes the rainfall in the two hydrographic sub-basins being evaluated, along with the arrangement of crops in the region, regulated by the federal ordinances of the Secretaria de Política Agrícola (SPA) of the Ministério de Agricultura, Pecuária e Abastecimento (MAPA) of Brazil. For soybean cultivation, the Portaria SPA/MAPA 249/2022 legislation [28] recommends sowing soybeans between 1 October and 31 De-

Pecuária e Abastecimento (MAPA) of Brazil. For soybean cultivation, the Portaria SPA/MAPA 249/2022 legislation [28] recommends sowing soybeans between 1 October and 31 December in the municipalities located in the Caiabi and Renato hydrographic

a recommended break in the planting of soybeans after 1 January and before 15 September each year. Therefore, in the region studied, maize is planted as the second crop (*safrinha*), and its planting depends on the sowing time and the growth cycle of the soybean cultivar that is planted. This is regulated by the Agricultural Zoning of Climatic Risk (ZARC), established by Ordinance SPA/MAPA 332/2022 [29], which recommends planting maize

between 1 February and 10 March each year.

cember in the municipalities located in the Caiabi and Renato hydrographic basins. This legislation also establishes a "sanitary void" for soybean cultivation, which is a recommended break in the planting of soybeans after 1 January and before 15 September each year. Therefore, in the region studied, maize is planted as the second crop (*safrinha*), and its planting depends on the sowing time and the growth cycle of the soybean cultivar that is planted. This is regulated by the Agricultural Zoning of Climatic Risk (ZARC), established by Ordinance SPA/MAPA 332/2022 [29], which recommends planting maize between 1 February and 10 March each year. *Sustainability* **2023**, *15*, x FOR PEER REVIEW 7 of 23

**Figure 4.** Daily rainfall in (**a**) the Caiabi sub-basin and (**b**) the Renato sub-basin, between 15 September 2020 and 30 June 2021. **Figure 4.** Daily rainfall in (**a**) the Caiabi sub-basin and (**b**) the Renato sub-basin, between 15 September 2020 and 30 June 2021.

#### *2.2. Simulated Rainfall Treatments 2.2. Simulated Rainfall Treatments*

Field trials with the InfiAsper rainfall simulator [20] were carried out in pastures and in areas under crop cultivation in the upper, middle, and lower regions of the two watersheds (Figure 5). The simulator operates with two Veejet 80.150 nozzles parallel to Field trials with the InfiAsper rainfall simulator [20] were carried out in pastures and in areas under crop cultivation in the upper, middle, and lower regions of the two watersheds (Figure 5). The simulator operates with two Veejet 80.150 nozzles parallel to

each other, positioned 2.3 m from the ground surface, with an average service pressure of 35.6 kPa. The diameter of drops applied by the simulator, considering the different pressure settings and rotation of the obturator disk, was exhaustively measured by

each other, positioned 2.3 m from the ground surface, with an average service pressure of 35.6 kPa. The diameter of drops applied by the simulator, considering the different pressure settings and rotation of the obturator disk, was exhaustively measured by Macedo et al. (2021) [22] using the flour method. Alves Sobrinho et al. 2008 [20] also made this assessment, confirming a mean drop diameter of 2.0 mm. Calibration tests were carried out in the laboratory, using a grid with 25 collectors over an area of 50 square centimeters, and uniformity coefficients above 80% were obtained. Renato River basin, the crops were maize at the V4 vegetative stage after soybean succession. In the two basins, the pasture areas were occupied by *Brachiaria* spp., with an average height of 50 cm. The simulated rainfall was carried out in different areas in the two sub-basins, namely, for soybean (the Caiabi sub-basin) and for maize (the Renato subbasin). Due to the cultural practices adopted in the region, the amount of remaining soybean straw in the maize crop was very small, when compared with the remaining maize straw, which can influence the timing of soybean planting. For this reason, no simulated rainfall was carried out in those plots with only straw in the Renato sub-basin.

this assessment, confirming a mean drop diameter of 2.0 mm. Calibration tests were carried out in the laboratory, using a grid with 25 collectors over an area of 50 square

Due to the regional agricultural calendar, the evaluations in cultivation areas in the Caiabi River basin occurred in soybean plantations at the V7 vegetative stage, with the plants cultivated in maize straw (no-tillage system), whereas in cultivation areas in the

*Sustainability* **2023**, *15*, x FOR PEER REVIEW 8 of 23

centimeters, and uniformity coefficients above 80% were obtained.

**Figure 5.** Installation of the *InfiAsper* rainfall simulator (**a**–**d**) and its operation (**e**), with plastic inside the support structure of the *InfAsper* to reduce the effects of wind; scheme of the components of the *InfiAsper* rainfall simulator (**f**) showing: (1) metallic structure, (2) water application unit, (3) control panel, (4) reservoir and water pump, and (5) runoff collector [22]. **Figure 5.** Installation of the *InfiAsper* rainfall simulator (**a**–**d**) and its operation (**e**), with plastic inside the support structure of the *InfAsper* to reduce the effects of wind; scheme of the components of the *InfiAsper* rainfall simulator (**f**) showing: (1) metallic structure, (2) water application unit, (3) control panel, (4) reservoir and water pump, and (5) runoff collector [22].

In the cultivation areas of the Caiabi River basin, simulated rainfall was performed considering the following treatments (Figure 6): covered with soybean (plant + straw), only straw, and without crops and scarified at 0.1 m soil depth (4). In the Renato sub-Due to the regional agricultural calendar, the evaluations in cultivation areas in the Caiabi River basin occurred in soybean plantations at the V7 vegetative stage, with the plants cultivated in maize straw (no-tillage system), whereas in cultivation areas in the Renato River basin, the crops were maize at the V4 vegetative stage after soybean succession. In the two basins, the pasture areas were occupied by *Brachiaria* spp., with an average height of 50 cm. The simulated rainfall was carried out in different areas in the two sub-basins, namely, for soybean (the Caiabi sub-basin) and for maize (the Renato sub-basin). Due to the cultural practices adopted in the region, the amount of remaining soybean straw in the maize crop was very small, when compared with the remaining maize straw, which can influence the timing of soybean planting. For this reason, no simulated rainfall was carried out in those plots with only straw in the Renato sub-basin.

In the cultivation areas of the Caiabi River basin, simulated rainfall was performed considering the following treatments (Figure 6): covered with soybean (plant + straw), only

straw, and without crops and scarified at 0.1 m soil depth (4). In the Renato sub-basin, the treatments evaluated were covered with maize, without crops, and scarified. In the pasture areas for both basins, rainfall was assessed with the conditions of soils covered with pasture, uncovered soils, and soils scarified at 10 cm in depth. Simulated rainfall was replicated 4 times per basin region and treatment, totaling 156 tests. The useful area of the simulator plot was 0.7 square meters (m<sup>2</sup> ) and the average slope of the surface in the field was 3 degrees (Figure 6h). To standardize soil moisture, the plots were dampened before the beginning of the simulated rainfall, according to the methodology described in [12]. *Sustainability* **2023**, *15*, x FOR PEER REVIEW 9 of 23 basin, the treatments evaluated were covered with maize, without crops, and scarified. In the pasture areas for both basins, rainfall was assessed with the conditions of soils covered with pasture, uncovered soils, and soils scarified at 10 cm in depth. Simulated rainfall was replicated 4 times per basin region and treatment, totaling 156 tests. The useful area of the simulator plot was 0.7 square meters (m2) and the average slope of the surface in the field was 3 degrees (Figure 6h). To standardize soil moisture, the plots were dampened before the beginning of the simulated rainfall, according to the methodology described in [12].

**Figure 6.** Soil cover treatments with (**a**) soybean, (**b**,**e**) pasture, and (**c**,**f**) maize, soybean + straw (**d**), and straw only (**g**), without crops (**h**,**i**), and scarified (**j**,**k**). The red arrows represent the flowchart of simulated events, as follows: (i) soybean + straw, straw only, without crops and with scarified soil; (ii) maize, without crops and with scarified soil; (iii) pasture, without crops and with scarified soil. In simulated rainfall tests in the different cropping systems, surface runoff was collected every minute (**l**,**m**), measuring the volume of water loss (**n**). Subsequently, the samples were dried in an oven (**o**,**q**); the differences between soil losses in pasture conditions, uncovered soil, and scarified soil for the same sampling point and time after the start of runoff (6 min) can be observed in (**p**).

The rainfall intensity (RI) of the simulated rainfall was defined, based on the intensity– duration–frequency (IDF) relationship created for the study region, according to the authors of [30]. The RI value was approximately 75 mm per hour, considering a return period of 10 years and an average duration of 42 min. After the beginning of the runoff, the material collection was performed at intervals of 1 min using plastic containers, and the runoff volume was measured with a graduated cylinder and then transferred to 0.5-liter (L) jars. The volume of runoff water was sent to the laboratory for the quantification of the water and soils lost through erosion [12]. To identify the amount of sediment in the water, the decantation and subsequent evaporation process of the water was conducted. To achieve this, the jars with the collected material were subjected to a temperature of 55 ◦C, in a forced circulation ion oven, until reaching a constant dry mass (up to 96 h). After drying, the samples were weighed on an analytical balance for the quantification of sediment.

#### *2.3. Analyses of the Soil Characteristics and Vegetative Cover Dry Matter*

For the physical-hydric characterization of the soil, near each point where the simulated rainfall was applied, mini-trenches of 0.4 × 0.4 m were dug for the collection of disturbed and undisturbed soil samples in the 0- to 0.1- and 0.1- to 0.2-m layers in the three regions of each sub-basin. The attributes analyzed in the physical-hydric characterization were granulometry (sand, silt, and clay), bulk density, particle density, total porosity, macroand microporosity, and hydraulic conductivity. Granulometry was determined by the pipette method, using a sodium hydroxide (NaOH) solution with mechanical agitation for 16 h, based on the principle of Stokes' law. Bulk density was obtained by the graduated cylinder method, using undisturbed samples. In the laboratory, the samples were dried in an oven at 105 ◦C and then weighed 48 h later [31]. Particle density was determined by the volumetric flask method. Total porosity was obtained via the relationship between the bulk density and particle density in Equation 1 [32]. Macroporosity was obtained by the tension table, with a tension of 10 kilopascals (kPa), and microporosity was obtained by taking the difference between total porosity and macroporosity [31]:

$$\text{TPo} = 1 - \text{Bd} / \text{Pd} \tag{1}$$

where TPo equals total porosity, Bd is bulk density, and Pd equals particle density. The soil class of all studied areas is latosol [25]. Soil textural distribution particles in the two watersheds are shown in Table 1.


**Table 1.** Soil textural distribution in the hydrographic sub-basins of the Renato and Caiabi Rivers.

\* Means that are followed by equal uppercase letters in the same column and equal lowercase letters in the same row do not differ significantly from each other, as established by the Kruskal-Wallis test at a 5% confidence level.

Vegetation cover was removed after rainfall on the covered ground and before rainfall on the bare ground from each plot. To measure the dry matter content of the vegetation cover, plant materials were collected, identified, and taken to the Hydraulics and Hydrology Laboratory at the Federal University of Mato Grosso. Subsequently, they were dried in an oven with forced air circulation at a temperature of 65 ◦C for 72 h, until reaching a constant dry mass of around less than 5% moisture. The dry mass was quantified using a thousandth of an analytical scale, while for the straw dry mass, the separation of the remaining soil particles was carried out.

#### *2.4. Experimental and Statistical Design*

In cultivated areas in the Caiabi River, the experimental design was conducted in randomized blocks (RBD), in a 3 × 4 factorial scheme, with 3 regions in the basin (upper, middle, and lower) and 4 soil cover/management treatments (soybean + straw, straw, without crops, without crops and scarified). In the Renato River basin, a similar experimental design was used (RDB), but in a 3 × 3 factorial scheme, with 3 regions in the basin and 3 soil cover/management treatments (maize, without crops, without crops and scarified). In pasture areas, regardless of the basin, a 3 × 3 factorial scheme was used, with 3 regions in the basin and 3 soil cover/management treatments (*Brachiaria* spp., without crops, and without crops and scarified). In all the conditions mentioned above, simulated rainfall was replicated 4 times. The repetitions were spaced 50 m apart, ensuring the same level in the toposequence and land use. The slope of the field area in our experiments ranged from 3 to 4 degrees. All variables were subjected to the Kruskal-Wallis test at a 5% probability (test for nonparametric data) using the Statistica program, version 10.0.

#### **3. Results**

#### *3.1. Vegetative Cover Dry Matter*

The sampled dry-matter contents of soybean and maize (crops), pasture, and straw are contrasted in Table 2. The dry mass of vegetative cover in plots that were subjected to simulated rainfall ranged from 7.26 to 11.91 metric tons/hectare in the Caiabi sub-basin and did not show significant differences in the Kruskal–Wallis test between the hydro-graphic sub-basin regions. The straw represents at least 70% of the dry mass of the vegetative cover (soybean + straw). This vegetative cover condition makes it possible to understand the soil and water losses in the absence of these covers and their relationships with the physical and water characteristics of the soil. No comparisons were made between the two sub-basins because the crops (soybean and maize) and the phenological stages were different.

and pasture in the experimental plots of simulated rainfall in the hydrographic basins of the Caiabi and Renato rivers.

**Table 2.** Dry matter (metric tons/hectare) measured for soil-cover treatments of vegetation, straw,


\* "With crops" indicates the presence of soybean and straw in the experimental plots of the Caiabi sub-basin (with only maize for the Renato sub-basin). Differences between means were compared using the Kruskal–Wallis test at 5% probability, where (i) capital letters represent the analysis of the sub-basin regions for the same soil cover, while (ii) lowercase letters represent the analysis of land cover for the same region of the sub-basin.

The Renato River sub-basin has phytophysiological, geological, and pedological characteristics that are different from those found in the Caiabi River basin. Those areas

occupied by cultivation had maize at the V4 vegetative stage, while the plants in pasture areas (*Brachiaria* spp.) had an average height of 45 cm. The dry weight obtained for maize varied along the sub-basin, due to the planting density; in both regions, the spacing adopted was 0.5 m between rows. However, the plant populations were 55,000 and 80,000 plants per hectare in the upper/middle and lower regions, respectively. In this sub-basin, there were also no significant differences that were verified by the Kruskal-Wallis test for the dry weight of plants in cultivation, with vegetation (e.g., maize) and pasture areas.

#### *3.2. Caiabi River Sub-Basin*

#### 3.2.1. Soil Characteristics in the Caiabi River Sub-Basin

The physical attributes of the soil in the Caiabi river basin varied according to the different regions (Table 3). There was a greater presence of clay in the upper region, while the lower region was characterized by a higher sand content. Although the soil classes are the same, with a predominance of oxisol [25], the lower part of the Caiabi River has a lower altitude, favoring the deposition of sand. The same occurred with bulk density and particle density, with higher values in the lower Caiabi River, since minerals present in the sand fraction have higher densities than clay minerals.

**Table 3.** Physical and hydric characterization of the Caiabi River sub-basin soils in the pasture and cultivation areas.


\* The soil characteristics measured included microporosity (Micro), macroporosity (Macro), total porosity (TPo), particle density (Pd), bulk density (Bd), and hydraulic conductivity (K0). Means followed by equal uppercase letters in the same column and equal lowercase letters in the same row do not differ significantly from each other, as shown by the Kruskal-Wallis test at a 5% confidence level.

3.2.2. Soil Losses and Surface Runoff in the Caiabi River Sub-Basin

The dry weight of vegetative cover in plots that were subjected to simulated rainfall in the Caiabi River basin ranged from 7.26 to 11.91 metric tons (t) per hectare (Table 1). There were no significant differences between the treatments, using the Kruskal-Wallis test to compare sub-basin regions and land use. This condition allows us to understand the soil and water losses in the absence of these crops and their relationship with the physical and hydric characteristics of the soil (Tables 4 and 5), as well as the surface runoff variable resulting from the simulated rainfall in the Caiabi River sub-basin. There were significant

differences in soil scarification for the other soil cover/management conditions in pasture areas. However, there were no significant differences between the positions in the sub-basin for this same use (Table 5).

**Table 4.** Average values of soil loss (grams/square meter) under different uses, soil cover/management, and regions of the Caiabi River sub-basin.


\* Means followed by equal uppercase letters in the same column and equal lowercase letters in the same row do not differ significantly from each other, according to the Kruskal-Wallis test at a 5% confidence level.



\* Means followed by equal uppercase letters in the same column and equal lowercase letters in the same row do not differ significantly from each other, according to the Kruskal-Wallis test at a 5% confidence level.

#### *3.3. Renato River Sub-Basin*

#### 3.3.1. Soil Characteristics in Renato River Sub-Basin

Unlike the Caiabi river basin, the soil attributes of the Renato River basin did not vary by region (Table 6). A similar presence of clay and sand was observed in the upper, middle, and lower parts of the basin. The same occurred with bulk density and particle density, porosity, and hydraulic conductivity, which showed a similar distribution in the different regions. The hydrographic basin of the Renato River is located in the Amazon Forest biome, unlike the Caiabi River basin, which is located predominantly in the Cerrado biome. These two biomes have different geological formations, although the soil classes are the same [25].


**Table 6.** Physical and hydric characterization of the Renato River sub-basin soils in pasture and cultivation areas.

\* Soil characteristics measured included microporosity (Micro), macroporosity (Macro), total porosity (TPo), particle density (Pd), bulk density (Bd), and hydraulic conductivity (K0). Means followed by equal uppercase letters in the same column and equal lowercase letters in the same row do not differ significantly from each other, according to the Kruskal-Wallis test at a 5% confidence level.

3.3.2. Soil Losses and Surface Runoff in the Renato River Sub-Basin

Significant increases in soil loss were observed under the scarified soil treatment when compared to plots with vegetation, regardless of the sub-basin region and land use (Table 7). This was similar to what was reported in the Caiabi River sub-basin, with higher soil losses for uncovered and scarified soils. Therefore, soil cover and limiting the soil surface turnover and soil exposure in areas of farming expansion are important for soil conservation, regardless of the biome. This sub-basin is characterized by the abundant presence of the Amazon rainforest (Figure 2), unlike the Caiabi River sub-basin, which is located in the Amazon-Cerrado ecotone [33]. Differences in soil losses in pasture areas with cover (vegetation) in different regions of the Renato River sub-basin are related to soil granulometry, since there is, on average, 80% of sand in the lower region (Table 4). In soils with higher sand contents, infiltration tends to be greater than in clayey soils.

**Table 7.** Average values of soil loss (grams/square meter) under different uses, soil cover/management, and the regions of the Renato River sub-basin region.


\* Means followed by equal uppercase letters in the same column and equal lowercase letters in the same row do not differ significantly from each other, according to the Kruskal-Wallis test at a 5% confidence level.

Regarding surface runoff, a behavioral inversion was observed under scarified soil treatments when compared to other soil cover types, since there was a reduction in the runoff in the uncovered and scarified soils of pasture areas (Table 8). This inversion is the result of the rupture of the surface layers, which are usually compacted in areas of farming and pasture, thus increasing the roughness that limits runoff and promotes water infiltration into the soil. The differences between the average values of surface runoff, especially in the lower regions between the Caiabi and Renato River basins, result from differences in the saturated hydraulic conductivity in the soils (Tables 3 and 6).

**Table 8.** Surface runoff (millimeters/hour) according to the different uses, soil cover/management, and regions of the Renato River sub-basin.


\* Means followed by equal uppercase letters in the same column and equal lowercase letters in the same row do not differ significantly from each other, according to the Kruskal-Wallis test at a 5% confidence level.

#### **4. Discussion**

#### *4.1. Erosion Drivers and Implications*

Agricultural crops (soybean + straw, only straw, maize, and *Brachiaria* grass species) are responsible for minimizing the direct impact of raindrops, acting as rain droplet buffers, preventing the disaggregation of particles, and reducing the sediment load in surface runoff [34]. According to the authors of [7], well-managed pastures can be considered sustainable, as they maintain soil quality in terms of the physical, chemical, and biological aspects and prevent erosive processes. In this sense, the pastures of the three sub-basin regions, with an average height of 50 cm, achieved satisfactory phytomass productivity (Table 2). Similar to soybean under a no-tillage scheme, the data are in accordance with the authors of [35], who observed similar results in the Cerrado latosols.

When studying the different levels of cultivated crops, such as soybean, maize, and pasture, prior researchers [12,16] concluded that soil losses increase with the reduction and removal of vegetative cover. Similar results were observed in our study, as soil losses in the Caiabi River sub-basin indicated significant differences between soil treatments with and without vegetative cover. In general, areas covered with vegetation (including straw) provided lower soil losses, revealing the importance of vegetative cover for reducing soil degradation (Table 5).

The occurrence of differences in soil losses for the types of crops and soil scarification demonstrates the need to maintain vegetation cover or straw, regardless of land use (cultivation or pasture), with minimal soil turnover. Due to the distinct physical and hydric characteristics in the sub-basin regions (Tables 3 and 6), a reduction in soil loss was observed from the upper to the lower region for those soils without crops and that were scarified, regardless of land use. The upper region has soils with a predominantly clayey texture, while the middle and lower regions have soils with a more sandy texture (Table 2). In this case, surface sealing may occur in clayey soils, which makes infiltration difficult and promotes an increase in runoff and consequent soil losses, due to the reduction in roughness [36,37]. Moreover, this sub-basin is located in an ecotone area (Cerrado/Amazon rainforest). In this transition area, there are geological, morphological, and pedological variabilities, as well as in the phytophysiognomy of the region [33], affecting the erosive processes along the sub-basin.

In terms of pastures, even though the pasture evaluation did not include soil covered only with straw, soil losses were still reduced with such live vegetative cover. These results are in accordance with [12,14,15], who, while studying simulated rainfall in uncovered soil treatments in different regions of Brazil, observed an intensification of losses due to erosion compared to covered soils. In general, unprotected soils tend to be lost due to the direct impacts caused by raindrops, which provoke detachment and the consequent transport of particles [38].

In addition to vegetative cover, water infiltration into the soil depends on other intrinsic factors, such as texture, porosity, bulk density, and compaction levels, which may compromise hydraulic conductivity (Table 6). Therefore, even in covered soil treatments that reduce soil losses, surface runoff may be high, as observed in this study. In this sense, the lowest values of runoff flow observed in the scarified plots resulted from the roughness formed in the layer that was turned over, along with soil aggregate rupture, which not only facilitates infiltration in areas of low slopes in the short term but also intensifies greater soil losses. Soil turnover in pasture and cultivated areas (e.g., plowing, sub-soiling) disrupts the aggregates, which facilitates this rupture due to rainfall. This facilitates the erosive transport of soil after tillage breaks up the soil layer compacted by animal trampling and wheel slippage from the tractors, planters, sprayers, harvesters, and trucks used during harvesting. Consequently, some researchers [13,39,40] recommend minimal turnover, combined with leaving the straw from crops in-field or using alternate turnovers (e.g., the planting of crops such as maize after the soybean harvest) as ways to minimize soil and water losses by erosion.

The increased values of water loss due to surface runoff in pasture areas with vegetation and without crops (Tables 5 and 8) may be related to precipitation falling directly on the straw, which covers the soil in no-tillage soybean cultivation. This condition may favor water runoff by its running off directly onto the upper surface of the straw. Nevertheless, the water loss depends on fragment sizes, layer height, and straw density, which can reduce infiltration [41]. To reduce soil and water losses, adopting crops with the purpose of protecting the soils and providing better conditions for the use and sustainability of production systems, including good water infiltration, is recommended [42]. Therefore, soil and water conservation management practices should be adopted, such as the use and incorporation of straw to increase infiltration, level terraces, no-tillage systems, minimum tillage, conserved pastures with rotational grazing, and crop rotation. By adopting these practices, it is possible to reduce the exposure and consequent soil loss, in addition to avoiding nutrient and carbon losses, as well as river aggradation [13,38–40,43,44].

According to our results observed in the scarified plots, the vulnerability of soil particles is evident with regard to transport. This confirms greater soil losses in those areas where conventional planting involves traditional management using soil tillage, typically involving one plowing stage and two harrowing stages, in addition to the minimum use of soil cover. Several authors obtained similar results in different regions of Brazil in studies with natural rainfall [39,40,45] and also in studies with simulated rainfall [13].

Raindrops and the surface runoff of water during rain events can lead to soil erosion, with the amount and type of vegetation covering the soil being a significant factor in reducing soil erosion [46]. Soil losses in the two sub-basins that we studied with vegetative cover (pasture, soybean, soybean with straw, and maize) can happen, especially during the early stages of plant development when there is more soil exposure. After the rainfall interception, infiltration, and saturation of the soil surface layers, the water surplus moves depending on the topographic gradients. Therefore, vegetative cover does not entirely eliminate erosion in those areas used by farming production systems. However, it drastically decreases erosion when compared with badly managed and/or unprotected areas, as was shown in the understory of olive orchards, with both the lower-cost natural regeneration of early successional weeds and intentionally planted cover crops in Minas Gerais State, Brazil [47].

Vegetative cover is naturally responsible for protecting the soil from the direct action of rainfall, and it might not necessarily eliminate all losses, except for some cases in native forests. In this context, several studies show the absence of soil losses in areas of preserved native forest or their drastic reduction in comparison with agricultural land use, such as pasture and cultivated crops [17,42,48,49]. In other words, the soil losses observed in this study may be directly related to the conversion of native forests into farming land. Furthermore, they indicate the need for new studies on simulated and natural rainfall in the Teles Pires River sub-basin region and other rivers in the Cerrado and Amazon biomes, with their transitory ecotones.

In agricultural frontiers such as the Teles Pires River basin and, consequently, in the drainage sub-basins studied, soil and water losses lead to numerous environmental problems. These problems include the pollution and contamination of rivers and streams by the transport of chemical products, the deposition of particles that cause aggradation, the exposure of stocked carbon, removal of the surface layer responsible for farming production, damage to cart roads by the formation of gullies, dam bursts, and the destruction of local biodiversity [2]. Thus, an alternative method to circumvent and/or mitigate erosive processes is to adopt conservationist practices, especially those involving vegetative cover, or the combination of such practices with edaphic or mechanical practices, such as terracing, catchment basins, drainage channels, and the construction of dams on the sides of plantations.

#### *4.2. Policy Recommendations*

Environmental conservation policies that are already implemented in Brazil have contributed to reducing soil and water losses, including those that encourage the direct planting system (i.e., no-till farming), the recovery of springs and degraded pasture areas, carbon sequestration, and the adoption of agroforestry systems included in the Low Carbon Agriculture Plan, present in Law 12,187 [50] and regulated by Decree 7390 of 9 December 2010 [51]. The protection of native forests is also regulated by federal law (12,651, of 25 May 2012), known as the "Forest Code," which establishes general rules on the protection of native vegetation, including permanent preservation areas, legal reserves, and areas of restricted use [52].

With regard to water resources, the National Policy on Water Resources (Law 9433 of January 8, 1997) has the following main objectives. The first objective is to ensure the necessary availability of water for current and future generations, with adequate quality standards for the respective uses. The second objective is the rational and integrated use of water resources. The third and final objective is to prevent and defend against critical hydrological events, either of natural origin or arising from the inappropriate use of natural resources, and to encourage and promote the capture, preservation, and use of rainwater [53].

In addition to minimizing soil losses, these initiatives contribute to increasing carbon stocks, conserving the biodiversity of biomes, and preserving rivers and lakes from the silting up caused by constant erosion processes [2]. The soil's ground cover, in addition to protecting against the direct impact of raindrops, also protects agroecosystems from wind erosion and solar rays, which affect the soil microbiology [54,55]. From 2009 to 2020, Brazil made progress in achieving the goals of establishing policies for the conservation of natural resources, with a focus on reducing climate change [56]. However, due to recent increases in deforestation, these environmental challenges are omnipresent.

In the present study, the impacts of agricultural land cover were evaluated on soybeans (in the Caiabi River basin) and maize (in the Renato River basin). The maintenance of bare soils, combined with scarification, promotes greater soil loss regardless of the crop and the region of the watershed. According to Borrelli et al. 2017 [2], the presence of cover and the absence of soil disturbance are the quickest ways to conserve pedological and edaphic resources. In this sense, the correct management of the soybean crop with the direct planting system and contour planting are alternative methods capable of stopping

or mitigating erosion [10]. The no-till system has been used in Brazil since the 1960s in the southern region of the country; the results point to an increase in the capacity of water infiltration into the soil, with a reduction in surface runoff, in addition to favoring the microbial community, improving the soil structure, and nutrient cycling [2,12,40,44]. Contours or contour planting avoids the formation of preferential lines for surface water flow, minimizing sediment transport.

For maize cultivation, soil and water conservation practices are also important, although, in the state of Mato Grosso, most areas use this crop immediately following soybean cultivation as a second planting (*safrinha*), when the rains are less frequent. Even so, in the months of March and April, rainfall can still be enough (288 and 121 mm/month, respectively [27]) to require attention when it comes to soil conservation. Technical assistance, combined with rural extension, can encourage rural producers to keep residual straw (e.g., stover) on the ground after the maize harvest, protecting the soil during the fallow period of the dry season (June through September). Maize stover can anchor soil prior to planting the soybean crop again at the start of the wet season in October.

Most of the cultivated areas with pastures in Brazil are still degraded or are in the process of degradation. This is due to misuse, such as exceeding the pasture-carrying capacity, a lack of pH correction and fertilization practices, and a lack of planning efforts to avoid erosion [14,15]. Another important factor that must be taken into account in the conservation of pastures is the criterion for animals entering the pasture, which can favor the loss of surface protection, increasing the vulnerability of the soil to erosion [7].

In addition to the existing sustainable agricultural development policies, new policies are needed, with targets that reach all agricultural and livestock producers. In studies in southern Brazil, previous researchers [57] concluded that one of the biggest limitations in combating water erosion is the lack of information on the subject for rural producers. These new policies can be specific for each crop or can be integrated as a sustainability plan for the production systems of soybeans, maize, and pasture. Incentives can be included for crop succession and rotation, intercropping grasses with legumes, and combating degradation with rotational grazing. In an evaluation of erosion processes under different agricultural management scenarios, integrated soil conservation practices were found to have a greater effect on combating soil erosion [57].

Although all the landscapes evaluated in the present study are considered relatively flat, in the state of Mato Grosso, there are agricultural areas on sloping land [14,55]; therefore, new initiatives should pay special attention to those areas with steep slopes above a gradient of 15%. The greater the slope of an area, the greater the potential for soil and water losses [1,58]. One study measured the effect of slopes of 15%, 25%, 35%, and 45% on soil and water losses and concluded that the greater the local slope, the greater the losses of water and soil resources [14]. Consequently, greater soil losses can accelerate the silting up of watercourses, as well as increase the exposure of stored carbon in soils, which can increase CO<sup>2</sup> emissions, thus compromising the biodiversity of biomes [59].

Soils are providers of ecosystem services; when these are compromised, civilizations can be in imminent danger of existential instability. Soil degradation affects the hydrological cycle, compromising food security in the countryside and in urban centers. Therefore, it is necessary to adopt management practices that ensure sustainability, especially in biomes with high levels of deforestation, such as the Cerrado and the Amazon [60]. In this regard, article 225 of the Federal Constitution of Brazil states that "everyone has the right to an ecologically balanced environment, an asset for common use by the people and essential to a healthy quality of life, imposing on the public authorities and the community the duty to defend and preserve it for present and future generations" [61]. The sustainability of natural ecosystems and agricultural production systems is necessary to optimize the use of natural resources linked to soil and water, preserving them for current and future generations.

#### **5. Conclusions**

In this study, we evaluate the erosion process in agricultural production systems in the Cerrado-Amazon transition region, using the InfiAsper rainfall simulator. The erosive process is independent of soil texture and is closely related to management and use systems associated with vegetation cover and soil scarification, regardless of the area's position within the sub-basins. The removal of the vegetation cover formed by soybean, maize, and pasture negatively affects soil–water dynamics, with a significant increase in soil losses in all regions of both sub-basins. Areas subjected to soil management with surface scarification (soil turnover) experience greater soil losses, regardless of the sub-basin region and land use. The results indicate the need for agricultural producers and farmers to make use of management practices that prioritize the maximum vegetation cover for crop cultivation and animal husbandry areas, as well as minimal soil turnover, as a path to the sustainability of production systems by reducing erosive processes. Future rainfall simulator studies can quantify changes in soil erosion by (1) using cover crops (e.g., *Crotalara juncea*) following soybeans–maize, (2) growing cotton (*Gossypium* spp.) as a second crop after soybeans in Brazil, and (3) in sugarcane (*Saccharum* spp.) and other large-scale commodity crops in Brazil and around the world, to support more sustainable agricultural systems and development.

**Author Contributions:** Data collection, writing, methodology, formal analysis, and figures—M.A.B.A. and D.F.d.C.; data collection, review, editing, supervision, and financial support—A.P.d.S. and F.T.d.A.; review, editing, supervision, and financial support—A.K.H.; supervision and data collection, H.B.A. and A.F.d.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was financed by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior—Brasil (CAPES) and the Agência Nacional de Águas e Saneamento Básico (ANA), Finance Code—001 and Process 88887.144957/2017-00. The authors wish to thank the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) for their support with scientific initiation grants and a productivity grant.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Study data can be obtained upon request to the corresponding author or the second author, via e-mail. The data are not available on the website as the research project is still under development.

**Acknowledgments:** We thank the Federal Rural University of Rio de Janeiro, specifically the PPGA-CS and GPASSA. The authors also thank all the students and professors of the Tecnologia em Recursos Hídricos no Centro-Oeste" research group (dgp.cnpq.br/dgp/espelhogrupo/2399343537529589, accessed on 15 December 2022). We are thankful for the comments and edits from two anonymous reviewers and one MDPI Special Issue editor, which improved the quality of this work.

**Conflicts of Interest:** The authors declare no conflict of interest. Supporting entities had no role in the design of the study; in the collection, analysis, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


**Disclaimer/Publisher's Note:** The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

**Jiliang Ma 1,†, Jiajia Qu 2,†, Nawab Khan 3,\* and Huijie Zhang 4,\***


**Abstract:** Minor beans other than soybeans or peanuts are edible beans (EBs) that significantly contribute to the Chinese agricultural sector and play a vital role in the sustainability of agricultural production, diversification of food consumption, and income generation for producers. These beans are an important source of protein in a healthy diet, helping to improve national food security. In addition, adjusting and optimizing the industrial structure promotes the sustainable development of agriculture and diversifies staple food crops and introduction of new revenue streams for EB products. The current study examines the responses of mung bean and broad bean producers to environmental and internal input constraints. This study uses the production function with a multilevel mixed-effects method and is based on 848 households from two major EB-producing provinces of China in 2018 and 2019. The results show that local climatic conditions influence planting behavior. These types of beans are considered as a supplement and backup crop to the staple crop. Commercialization encourages cultivation. Producers show variable price responses to output prices, but very strong responses to product costs. Minor bean production is favored by small households because of its low labor intensity. For households growing these beans for consumption, soil fertility and environmental outcomes are improved. Findings from research on planting behavior have strong policy implications for guiding research and development for drought and pest resistance, market monitoring for price stabilization, promoting EB production through low-cost technologies, and encouraging sustainable agriculture.

**Keywords:** edible beans; planting behavior; high quality; multilevel model; determinants; China

#### **1. Introduction**

It has been noted that the national development goal for the next decade is to achieve harmony and unity between humans and nature through eco-friendly development [1]. This requires a multipronged approach: in addition to increasing the supply of safe and high-quality agricultural products, the focus needs to shift from quantity to quality [2], while increasing agricultural incomes by establishing sustainable and efficient agricultural production structures compatible with existing resources [3]. Although edible beans (EBs) account for a small share of agricultural production in China, EBs play an integral role in the sustainability of agricultural production, diversification of food consumption, and income generation for producers [4]. EBs refer to more than 20 kinds of legume crops such as mung bean (MB), adzuki beans, common beans, broad bean (BB), peas, etc., except soybeans and peanuts. They are an important source of fiber and protein for a healthy diet [5]. At present, the per capita per year consumption of EBs is about 1.7 kg, which is

**Citation:** Ma, J.; Qu, J.; Khan, N.; Zhang, H. Towards Sustainable Agricultural Development for Edible Beans in China: Evidence from 848 Households. *Sustainability* **2022**, *14*, 9328. https://doi.org/10.3390/ su14159328

Academic Editor: Aaron K. Hoshide

Received: 22 June 2022 Accepted: 27 July 2022 Published: 29 July 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

expected to expand rapidly with the increase in income and the improvement of nutrition awareness under the strategy of high-quality agricultural development [5].

EBs are also important crops for ecological protection, utilization of idle land, and reduction in disaster damage, which can continuously promote the development of green, low-carbon, and circular agriculture [6]. Therefore, EBs are widely recommended for intercropping or rotation with cereals, root crops, cotton, and fruit saplings. In addition, they are grown on small marginal plots that are less fertile for other crops. These beans are also heat and drought tolerant and have a short growing period, making them popular in all weather conditions. Consequently, during extreme weather events, farmers are drawn to when they quickly replant them to mitigate income losses. EBs are important tools for optimizing crop structure, developing rural areas, reducing poverty, and revitalizing rural areas. Although EBs are widely grown in most parts of China, their production is concentrated in less-developed provinces with large ethnic minority populations and widespread poverty. The provinces with the most EBs are Yunnan, Inner Mongolia, Heilongjiang, Jilin, Sichuan, Guizhou, Chongqing, Shanxi, and Shaanxi, with a planting area of more than 100,000 hectares. In 2018, together this accounted for 65% of the country's total arable land.

According to the investigation by China Agriculture Research System, smallholder production remains the main format of EB production and an important source of household income. The central government has been encouraging adjustments and optimizing the agricultural structure and local brand certification and management to increase highquality agricultural supply. Examples of EBs serving as a major sector for local poverty reduction include MB from Baicheng City, Jilin Province, the BB from Dali Prefecture, Yunnan Province, and adzuki bean from Kelan County, Shanxi Province. In the case of Kelan County, Shanxi Province, a local noncereal production base was established with adzuki as the leading crop. The production base covers 6400 hectares and 141 administrative villages (of which 90 villages are defined as poor), which benefits 20,485 residents in 7923 households. It was estimated that the average household income increased by 244.98 US\$ (1 US\$ = 6.8985 yuan on 2019) for 3412 poor households (8854 residents). Fresh BB from Dali Prefecture, Yunnan Province also proves to be a feasible approach for erasing poverty among households living at high altitudes. Despite the strategic importance and growing demand, the domestic cultivation area of EBs had more than halved between 2001 and 2018, from 3.8 million hectares to 1.8 million hectares. The share of EBs in crop cultivation areas also dropped from 3.5 to 1.5%. As a result, bean imports climbed sharply and exacerbated the declining competitiveness of domestic production. There is a rich body of literature on the planting behaviors of cereal farmers, but very few on EB farmers.

The main objective of this study was to analyze EB plantations and their impact on China's sustainable agricultural development (the ratio of EB planted area to total planted area). Based on a survey of 848 households in 2018 and 2019, in two producing areas (MB in Baicheng City, and BB in Dali Prefecture), this study provides an empirical analysis of producers' planting behavior and its determinants. Results from this analysis will contribute to the understanding of farmers' planting behavior and provide evidence in policy recommendations on sectoral development to enhance competitiveness and profit under a high-quality agricultural development strategy. The unique market structure of EBs, different from cereal crops, can also shed insights into other low-volume agricultural products.

#### **2. Literature Review**

The development of cash crop production (such as mung and broad beans) to improve household welfare has been at the heart of food policy debates in many developing countries [7,8]. Several studies have shown that cash crop production can be an effective way to improve the household's economic well-being and agricultural development [9,10]. For example, Christiansen et al. [11] studied coffee farmers in Tanzania and found that in the presence of health and drought shocks, coffee farmers remained economically resilient compared to other major crop farmers, suggesting a positive impact on agricultural development and economic well-being of farmers. Similar studies have been found in other

developing countries. For example, Kennedy et al. [12] found that participation in a cash crop program resulted in increased household income in six African and Southeast Asian countries (including Gambia, Guatemala, Kenya, Malawi, the Philippines, and Rwanda); later Finnis [13] showed that in southern India, growing cash crops of farmers achieve higher economic and social benefits, and in Malawi, households that choose to grow cash crops also show significantly higher incomes than those that do not use or adopt [14].

Several recognized channels for promoting crop production can improve the economic well-being of households and agricultural development. First, cash crop production can be an effective way to increase family farming income [15–17]. Specialization in cash crop production (relative to staple crops) typically results in higher economic returns per unit of land, including land, water, technology, and, to some extent, labor input. Second, boosting cash crop production can help diversify household livelihoods, thereby further improving household resilience to economic shocks (such as market price shocks) and other climate-related shocks (such as droughts, extreme heat, and cold temperatures). For example, several studies have found that crop diversification, such as intercropping and crop rotation, can increase the resilience of family farming production [18,19]. Third, the benefits of cash crop production also benefit other non-cash-crop farmers through the impact on employment because most cash crop production is labor-intensive [20,21]. Increased labor demand for high-value cash crops is likely to increase average wages for non-cash-crop farmers. In addition, the introduction of cash crop opportunities has shown that households can reduce cash constraints and be able to purchase improved crop production inputs [22]. As a result, their ability to adapt to yield enhancement techniques and agronomic practices is enhanced [23]. This cash income ultimately provides farmers with the opportunity to invest and improve farm management, thereby stimulating agricultural innovation and increasing yields [23–25].

However, there are reasons to question the positive impact of cash crop production on household economic well-being [14,26]. For example, some recent studies have shown that cash crop production has failed to increase the economic well-being of households in some developing countries, especially the poorest households, due to high barriers to entry [27]. They found that promoting cash crop production did little to improve the living standards of the poorest and that these poorest households were often ignored or barred from participating in the production of these cash crops [28]. In China, previous studies on cash crop production have focused on two aspects. A set of the literature focuses on the conceptual and theoretical basis of farmers' economic crop production choices and their operating mechanisms [29]. Other research on cash crop production has generally focused on how cash crop production affects household labor distribution, their subsequent household relocation decisions, and other noneconomic outcomes such as environmental and ecological consequences [30,31]. Although there is ample evidence that cash crop production can be an effective approach to improving household economic well-being [28,32], these observations are mostly correlated. Evidence on its causality is rather limited. It is unclear to what extent and under what conditions cash crop production can achieve desirable results at the micro-household level [14,33].

In addition, farmers' decisions on the production of cash crops (relative to staple crops) are increasingly influenced by perceived risks from climate change [34,35]. In the context of soybean production in China, farmers' perceptions are enhanced due to repeated changes in climate (such as excessive rainfall and flooding in different regions), which directly affects farmers' expected soybean yields and their economic benefits [36]. In response to this heightened perception of climate-related risks and the observed adverse effects on soy production, farmers may consider other alternatives and more resilient crops to address this potential negative impact [37]. As hypothesized by Asrat and Simane [38] and Ojo and Baiyegunhi [39], adaptation to climate change involves a multistep process in which strong perceptions (or strongly perceived changes in climate conditions) must be established, and subsequently, appropriate on-site responses may be initiated for these changes. Several studies have examined this relationship [34,35,40–43] and concluded that

household adaptation to climate change behavior is directly related to its perception [44]. However, there is limited research on the combined effects of edible bean (EB) cultivation and its impact on commercial and agriculture development [45] and economic welfare [39]. It can be investigated from the above studies that EBs and crops planting play an important role in increasing farmers' income and agricultural development in China.

#### **3. Data and Methodology**

#### *3.1. Survey Design and Data Collection*

Mung bean (MB) in Baicheng City, Jilin Province, and broad bean (BB) in Dali Prefecture, Yunnan Province are selected for this study for their significant role in edible bean (EB) production and representativeness. Both locations are important players in national production. Baicheng City MB is a certified geography signature product with an average annual planting area of 80,000 hectares. Annual MB production in Baicheng City reaches 100,000 tons, accounting for 11% of national total production. Approximately half of Baicheng MB is exported, representing above 30% of total exports. A major noncereal wholesale market is located in Baicheng City, facilitating the commercialization of local MB. Yunnan province is the top producer of EBs in the country, representing 16.5% of the national planting area. In Yunnan province, Dali Prefecture is the largest producer of BB, with more than 20% of provincial production coming from this location.

The household-level of data is obtained from households in two consecutive years, 2018 and 2019. A cluster sampling method was used in the survey [46]. Household-level data was obtained from 5 counties/cities in Baicheng City, Jilin Province, and 4 counties/cities in Dali Prefecture, Yunnan Province, covering 32 townships and 66 villages. Among 938 questionnaires distributed, 848 provide valid responses (90%) (see Table 1 for details). Dali Prefecture represents 45.8% of the total sample, while Baicheng City captures the remainder. The distribution of households reflects the geographic concentration of production, with wide cultivation of BB in Dali Prefecture, but a high concentration of MB in Baicheng City (mainly in Tiaonan and Tongyu counties).


**Table 1.** Samples Location.

The county-level data on the weather was obtained at the county level from the Chinabric database. The highest and lowest temperature and an average rainfall of Chinabric database at county/city are collected. Weather information in April is collected for MB as they are usually planted in April in Baicheng City and harvested in late September and early October. Dali producers usually plant BB in late September and harvest in April and May of the following year. Hence, weather information in October was collected for Dali Prefecture.

#### *3.2. Empirical Model*

This paper uses regression analysis, which is a statistical technique for estimating the relationship that independent variables have on a dependent variable [47]. Regression analysis has many estimation methods to correct for errors. Specifically, a multilevel mixedeffects model was chosen for this study [48,49] because it recognizes the hierarchical nature, clustering, and the structure of the data in this study. Households in the same county tend to be more exposed to the same environmental and social characteristics than households chosen at random from the population. The county difference refers to the resource and environmental endowments difference. Multilevel mixed-effects models accommodate the existence of data hierarchy by introducing separate residual components for each level of the hierarchy. In this study, residuals are assumed to exist at both household and county levels.

The residual variance is partitioned into a between-group component (the variance of county-level residuals) and a within-group component (the variance of the household-level residuals within the same county). Between-group residuals, or county effects, signify unobserved county characteristics that affect household production behavior [50]. County effects are unobservable but can lead to a correlation between EB planting behaviors for households within the same county. Because individual households at the county level are not independently distributed, results from classic Ordinary Least Squares (OLS) are no longer valid [51].

Multilevel mixed-effects models allow both fixed and random effects, especially in the case of non-independence in a hierarchical data structure; they are also called multilevel models. Multilevel mixed-effect models offer several advantages. First, they correct inferences based on observations that are not independent. Ignoring the intercorrelation among households can result in underestimated standard errors of regression coefficients and thus an overstatement of statistical significance. Standard errors for the coefficients of higher-level predictor variables will be the most affected by ignoring the grouping of the county.

Second, multilevel mixed-effects models address individual county-level effects by separating county unobserved conditions from local agronomical practices in obtaining household planting behavior. Moreover, county-level effects are estimated simultaneously with the effects of county-level coefficients. Instead of multiple county dummy variables in a fixed-effects model, this approach allows the estimation of county-level variables of interest (weather indicators in this study), while separating observed (weather) and unobserved (county random effects) county characteristics.

Third, results from the multilevel mixed-effects model can be generalized to a wider population of counties. Unlike a fixed-effects model, inferences are limited to the counties in the sample and cannot be made beyond these groups. By assuming that the random county effects come from a common distribution, planting behaviors obtained from this study can be extrapolated to other counties.

This analysis uses a linear multilevel mixed model with random intercepts. Regression models with one dependent variable and more than one independent variable are called multilinear regression [47]. The regression model can be expressed as

$$y\_{ij} = \beta\_{0j} + X\_{ij}\beta\_1 + e\_{0ij} \tag{1}$$

$$
\beta\_{0j} = \beta\_0 + \mu\_{0j} \tag{2}
$$

where *j* = 1, 2, . . . , *n* refers to county, *I* = 1, 2, . . . , *m* refers to households, *yij* is the share of EB planting area in total planting area for, *xij* are variables affecting EB producer behavior for *i*-th household of *j*-th county, *β*0*<sup>j</sup>* is the sum of fixed intercept *β*<sup>0</sup> and a random intercept for the *i*-th county *µ*0*<sup>j</sup>* , *β*<sup>1</sup> is the fixed slope, *e*0*ij* is a zero-mean Gaussian error term.

Equation (2) assumes that *µ*0*<sup>j</sup>* and *e*0*ij* are unobserved random variables that are independent of each other. Independence of between-group can be measured by ICC (Intra-Class Correlation Coefficient) as the ratio of between-group variance to the total variance. STATA (16th version) software is employed to estimate the results, and Standard Maximum Likelihood (ML) and Restricted Maximum Likelihood (REML) estimation are applied, with the latter providing consistent estimates.

$$\text{ICC} = \frac{\sigma\_{\mu\_0}^2}{\sigma\_{\mu\_0}^2 + \sigma\_{\varepsilon\_0}^2}$$

,

where *σ* 2 *µ*0 is between-group variance and *σ* 2 *e*0 is within-group variance. ICC takes a value between 0 and 1. ICC is close to 1 when between-group variances are large. If ICC is close to 0, within-group variance is large, or observations are independently distributed. The rule of thumb is that mixed-effects models are appropriate when ICC is greater than 0.1.

#### **4. Results and Discussion**

Planting behavior is defined as decisions by producers to maximize profit under certain constraints in resources. The existing literature on agricultural planting behavior covers planting behaviors about crop selection and input. Crop planting behaviors involve the willingness to grow [52,53], and planting area and input decision involves the choice of variety [54], application of chemical fertilizer [55], and pesticide and herbicide [56–59]. Referring to the literature, the dependent variable is defined as the share of edible beans (EBs) planning area in household total planting area instead of planting area, which controls the variations of land size across households. The explanatory variables address weather, market and household, and individual characteristics.

Weather is proven to have affected crop production [60–62], planting system, and crop allocation [60,63], as well as producer decision [64,65], and we noticed that farmer's willingness to plant maize is negatively related to average temperate during the growing season but positively related to average rainfall. Interviews with EB farmers in our research regions reveal that farmers pay close attention to temperature and rainfall before EB planting. Hence, average rainfall and the highest and lowest temperature at the beginning of the planting season are included in the analysis.

Most research indicated that the higher the market price, the stronger the willingness to plant [66,67]. On the contrary, other researchers pointed out that to a large extent, the planting behavior did not depend on the price level, and the farmers' response to price fluctuation was more diverse [68]. In addition to prices, the production costs of different crops also affected the behavior of farmers. The higher the production costs, the lower the cost benefit of crops, which directly reduced the production benefits of farmers and the planting behavior [69]. Hence, output prices of last year and production costs are included in the analysis.

In terms of household and individual characteristics, this study selected household size, the share of wage income in household total income, household head age, and education. Larger households are more likely to choose multiple crops to diversify production risks. Wage income is treated as a proxy for non-agricultural income. Household head demographics identify the producer's access to new technology and information [70–73].

#### *4.1. Descriptive Statistics of the Sample*

There is substantial variation in planting area: the average mung bean (MB) producer reported a planting area of 2.67 hectares (ha), while the average broad bean (BB) producer only reported 0.14 ha. This regional difference is due to the availability of arable land: the average Baicheng household cultivates 9.63 ha, far greater than the average land size of 0.39 ha in Dali Prefecture. BB production is an integral part of livelihood as its average share in total planting area reached 59% (Table 2).


**Table 2.** Descriptive statistics of the sample.

Note: Weather information in Tiaobei was not available and replaced with Da'an in a similar latitude. Education level: 0 = illiterate, 6 = primary school, 9 = middle high school, 12 = high school graduate and above. Source: Weather data from Chinabric database.

Interviews show that farmers adjust their planting decision based on weather conditions. For instance, farmers in Baicheng City will choose to plant maize if the average rainfall is favorable; otherwise, farmers will choose to either delay maize planting or switch to MB for its drought resistance. There was little difference in temperature between 2018 and 2019 in Baicheng, China, with a slightly warmer temperature in July 2019. The highest temperature was 35.9–38.7 ◦C in 2018 and 36.1–39.7 ◦C in 2019. Similarly, the lowest temperature was close to −5.2–−4.6 ◦C in 2018 and −8.5–−6.3 ◦C in 2019. However, average rainfall differed remarkably across China in Baicheng City.

Market factors include last year's output prices and average cost. Output prices are the prices sold from the farm gate. The average output price is 0.94 US\$/kg for MB and 0.57 US\$/kg for BB. The average production cost is 42.40 US\$ for MB, and the BB cost almost doubles that of MB at 75.54 US\$. Input costs, such as the purchases of seeds, fertilizer, pesticides, and herbicides, accounted for 50.5% of total MB cost, while land rent accounted for another 41.5%. As a result, labor costs only represented 5.1% of the production cost for MB. Input costs were a higher portion of BB production cost at 71.4%. Labor cost is also higher at 12.1%, but land rent was a much smaller part of the total cost at 4.6%. The difference in cost structure can be attributable to the land market and the adoption of mechanization. The land market is very active in Baicheng City, with an average rent of 434.88 US\$ per hectare. This leads to a larger land size that supports the use of machines in every stage of MB production, resulting in a much lower dependence on human labor. Dali is the opposite case of smaller plot size from the inactive land market and low adoption of agricultural machinery.

The average size of EB-growing households was four to five persons, with middle school education. The average household head was 48 years old among MB growers. BB growers were older, with the household head at 52 years old. The share of household land that was irrigated was 37.6% of Baicheng because MB is planted in lands without irrigation. BB plots are more likely to be irrigated as 78.3% of household land was irrigated in Dali. MB producers also tend to be more specialized in agriculture with only 5.3% of household income derived from wages. Almost half (46%) of household income came from wages and salaries, indicating that BB-growing households tend to have more diversified income sources and are less dependent on crops for income earning.

Table 3 reports the values of ICC for MB are 0.132 under ML and 0.142 under REML, suggesting that 13.2% of variance comes from between-group components. ICC values are even higher for BB at 0.304 under ML and 0.327 under REML [74]. The results confirm the appropriateness of mixed-effect models in this analysis to avoid inconsistent estimation. The superiority of mixed-effect models is further confirmed by chi-square values when compared with OLS. Akaike information criterion (AIC) values indicate that ML is the preferred model [75]. Coefficients in Table 3 are used to estimate elasticities of planting behaviors in Table 4, evaluated at the sample mean.


**Table 3.** Estimation results from multilevel models.

Note: \*\*\*, \*\* and \* indicate 1%, 5% and 10% statistical significance level. Standard errors in parentheses. Estimated parameters are listed above the stand errors in the parentheses.



Note: \*\*\*, \*\* and \* indicate 1%, 5% and 10% statistical significance level. Elasticity is an economics concept that measures the responsiveness of one variable to changes in another variable, Elasticity = *dy dx* ∆*x* ∆*y* , in which *y* is the independent variable and *x* is a dependent variable.

#### *4.2. Impacts of Weather on Farmers' Planting Behavior*

Weather plays an important role in EB producers' decisions. The highest temperature and average rainfall before planting season could substantially affect planting decisions.

A heat wave before planting season is likely to increase MB planting (at a 5% significant level). When the highest temperature in April increases by 1%, the share of MB cultivation would increase by 0.53%. A field study suggests that MB production serves as a strategy to mitigate loss. When it becomes too warm (higher temperature) for maize production, farmers would choose MB as a substitute to decrease crop cultivation loss.

The average rainfall is negatively associated with MB production. A 1% increase in average rainfall in April could lead to a 0.09% drop in MB planting share. This is due to the substitutability between maize and MB. The average profit for MB is about 72.48 US\$, slightly higher than that of maize at 65.23 US\$. However, farmers prefer maize production because it requires less labor input due to a higher level of mechanization and comes with subsidies from the government. When soil contains more moisture in April, farmers would choose to plant maize for a potential good harvest. However, when average rainfall is low, farmers exhibit a higher propensity for MB, highlighting EBs' role in loss mitigation. The results of Pataczek et al. [76] show that mung bean (Vigna radiata) is gaining attention as a short-season crop that can tolerate dryland conditions (with less rainfall). MB is such a minor crop that dryland smallholder farmers can use to break the downward spiral and increase the profitability and sustainability of their farms. Integration of mung bean in cropping systems may increase the sustainability of dryland production systems. Diversification of local production systems through the inclusion of mung bean as a catch crop provides additional income to farmers and has the potential to improve soil fertility [77].

The lowest temperature affects the planting decisions of BB planting in Dali. The share of BB planting area would drop by 1.78% if the lowest temperature in October falls by 1% (at a 10% significance level), as short episodes of extreme minimum temperature can essentially impede crop development and thus impact crop yield [78]. This is also associated with the growth pattern of BB. Experienced farmers will cut BB planting in a warm October because it could lead to rapid growth and the flowering season could be susceptible to frost and thus lower crop yield. Average rainfall does not appear to impact the planting behavior of BB, mainly due to reliable and sufficient rainfall patterns. Table 2 indicates that average rainfall was 210.2–473.4 mm in October 2018, and 188.4–541.6 mm in October 2019.

#### *4.3. Impacts of Market on Farmers' Planting Behavior*

MB farmers are very responsive to output price but not BB. Given that the output price is the price sold from the farm gate, the own-price elasticities we estimated represent the derived (farm gate) supply rather than the market supply [79]. If the output price goes up by 1%, the share of the MB planting area will increase by 1.18% (at a 5% significance level). This phenomenon reflects different levels of commercialization between these crops. MB farmers are active participants in markets and thus are more sensitive to price signals. Additionally, MB farms tend to be much larger at 2.6 hectares, allowing large-scale application of machinery to lower production costs and adjustment in the planting area. On the other hand, BB production is more stable and less responsive to market prices because most of the output is still used for own and local consumption. The lower level of machinery further limits farmers' ability to expand production quickly.

The average cost can negatively impact planting behavior for both EBs. The lower the cost, the higher the planting area. If cost decreases by 1%, the share of the planting area would increase by 0.14 and 0.08% for MB and BB, respectively. Because the majority of the total cost is spent on inputs like seeds, fertilizer, and chemicals, policies targeting lower input costs could promote the production of EBs. In 2018, Yunnan initiated a program to support local green food by providing free fertilizer to lower production costs.

#### *4.4. Impacts of Household and Individual Characteristics on Farmers' Planting Behavior*

Household size is negatively associated with EBs planting share, especially in the case of MB, implying that larger households with more arable labor are less inclined to plant EBs. In Baicheng City, larger households prefer maize for its higher labor productivity from mechanization. In Dali, the BB is less demanding in labor requirement; sometimes farmers choose to skip applying fertilizer or chemicals to cut labor input without sacrificing yield much.

The elasticities of the share of irrigated areas are significant for both MB and BB but of opposite signs due to local agronomical conditions. The share of MB planting area declines when a household has more irrigated plots, with a 1% increase of irrigated land share leading to a 0.37% decrease in MB plant area (at a 10% significance level). The land is largely non-irrigated in Baicheng City, with only 37.6% of the planting area irrigated. Farmers prefer to grow maize on irrigated plots while cultivating MBs in non-irrigated areas. This negative elasticity underscores the key role of MBs in loss mitigation. The impact of irrigation is positive in Dali, partly due to the high availability of irrigated plots (78.3% of the planting area is irrigated).

Wages and salaries are a good proxy of income diversification in rural households. A higher share of wages and salaries in household income indicates a diversified income base and fewer resources dedicated to agricultural production. The elasticity of MB is insignificant but significant for the BB. This is because there are fewer substitute crops in Dali and the flexibility offered by BB production through low labor demand and smaller plots. Diversification from agriculture introduces new income sources for EBs producers. Elasticity associated with household head age is positive and significant for MB with a value of 0.38 at a 5% significance level. Older household heads express a stronger preference for MB for its low input requirement compared to maize.

#### **5. Conclusions and Implications**

Few studies have focused on farmers' behavior in growing noncereal crops such as edible beans (EBs), despite their crucial role in the sustainability of agricultural production, the diversification of food consumption, and farmers' income. EB production also has its unique characteristics different from cereals: short growth period, low soil requirements, low yield, less machinery usage, but a high degree of commercialization. Therefore, our research will focus on improving the competitiveness of EBs and improving the allocation of agricultural resources under the eco-friendly development strategy. This paper estimates the planting behavior of two representative edible legumes: mung bean (MB) and broad bean (BB) by estimating a multilevel mixed-effects model. The empirical analysis uses the household-level data of Baicheng City in Jilin Province and Dali Prefecture in Yunnan Province in 2018 and 2019, combined with the county-level meteorological observation data in the month of planting.

This study determines whether or not it is a major factor in the decision to grow EBs. In the case of MB, producers choose to grow MB to mitigate the potential loss of income from unfavorable soils and temperatures. BB farmers are also temperature-sensitive when making planting decisions to ensure high yields. Although the commercialization and specialization of the two types of EB are different, and producers may be less responsive to output prices, they are generally responsive to production costs. Smaller households are showing a higher interest in growing edible pulses due to reduced demand for land and income generation. Although BB is grown in small pieces and eaten more often, its role in increasing nitrogen fertilizer cannot be ignored. Based on the above results, it is believed that under the background of high-quality agricultural development, the ecofriendly development and high-quality development level of the bean industry should start by improving the agricultural development capabilities of the bean industry in order to address climate change. To reduce carbon emissions to address climate change, China's lesser bean producers can improve production efficiency. They can accomplish this by market monitoring and relying on scientific and technological improvements to increase crop productivity per unit input of greenhouse gas emitting inputs (e.g., tractors, processing machinery, etc.).

First, in the process of high-quality agricultural development, to enhance the adaptability of EBs to climate change. It is necessary to increase investment in research and development, carry out the screening of main varieties and key technologies, and enhance stress resistance so that beans are resistant to drought, pests, and other risks. The ecological and cultural value of edible pulses should be further developed. In Dali Prefecture and other key ecological development zones, it is necessary to further explore the ecological service value of beans, improve the organic combination of the beans industry and local culture, and explore new forms of the beans industry. Second, for EBs with a high degree of marketization (such as MB in Baicheng City, Jilin Province), a market price monitoring mechanism should be established, and market supervision of small-scale agricultural products should be strengthened to prevent excessive consumption and reserve as well as stabilize market prices and farmers' production. At the same time, we should encourage the "Internet + Project" development model of EBs, further standardize and guide the EB sales and circulation market, and provide farmers with services such as prenatal markets. Information, quality control in the process of production and post-purchase, and sales through the internet or Internet of Things technology.

Third, it is important to improve quality and efficiency through science and technology, reduce the cost of bean production, and promote the upgrading of the bean industry. Strengthen the promotion and application of new varieties of high-yield and high-quality EBs suitable for mechanized production, and promote mature, high-yield, efficient, simplified, and integrated production technologies characterized by large-scale, standardized, and mechanized production, reducing the production cost of EBs and increasing farmers. The economic income of farmers has improved their enthusiasm to grow EBs.

Finally, it is recommended to establish a demonstration zone for the integration of the bean product industry, develop and improve the bean product industry chain, and closely integrate the bean product industry with the health industry. After starting with implementation of lesser bean industry integration demonstration areas, China can improve integration of bean production and processing, as well as facilitate marketing to the health industry. Future research can enable breakthroughs in the development of processing edible bean products using more eco-friendly processes.

**Author Contributions:** J.M., J.Q., N.K. and H.Z. developed and outlined this concept, including the method and approach to be used; J.M., J.Q., N.K. and H.Z. developed and outlined the manuscript; J.M., J.Q., N.K. and H.Z. contributed to the methodology and revision of this manuscript; J.M., J.Q. and N.K. wrote the article. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the China Agriculture Research System of MOF and MARA-Food Legumes (CARS-08) and the National Natural Science Foundation of China (No. 71904190).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare that they have no conflict of interest.

#### **Abbreviations**



### **References**

