Maize Cultivation Three Hundred Years Ago Triggered Severe Rocky Desertification in Southwest China

Understanding the forest evolution is vital to answering the reforestation potential in karst areas. Here, we present the first‐ever pollen record in karst depression sediment, combined with comprehensive dating methods (137Cs, 210Pbex, and 14C) and historical documents, to reveal plant change history in southwest Guangxi, a severe rocky‐desertification region. We inferred three stages of “virgin forest–deforestation–sparse tree planting” over the past three centuries. Before the 1780s, the barren mountains used to be a lush mixed broadleaf forest probably. However, maize cultivation, along with explosive population growth and migration, accelerated mountain reclamation and deforestation, leading to severe rocky desertification since the 1780s, featured by the co‐occurrence of Zea pollen appearance and Dicranopteris spore surge from 0.92% to 12.18%. Since the 1930s, sparse tree planting began, as Cupressaceae/Taxodiaceae pollen abruptly increased by 32%. Our study is significant in understanding the rocky desertification causes and guiding ecological restoration in the region.


Introduction
Rocky desertification refers to transforming karst areas covered with vegetation and soil into rock-baer landscapes.Its occurrence, accompanied by land productivity plunging, is driven by intense human activities such as deforestation on vulnerable dissolvable carbonate bedrocks (Wang, 2002).Not only is it an extreme land degradation issue, but rocky desertification also induces severe economic consequences.Owing to minimal useable resources, including soil and forest, people living in rocky desertification areas are usually struggling with poverty and stuck in a "Poverty-Excessive Exploitation-Ecological Degradation-Less Resource-Poverty Intensification" circle (Jiang et al., 2014).
Southwest China is a typical region in the world suffering from rocky desertification and poverty (Figure 1a).To combat rocky desertification, Chinese governments have implemented a series of large-scale ecological projects such as "Grain to Green" since around 2000, and invested more than 130 billion RMB.As an encouraging result, southwest China is "greening" with rocky desertification relief in terms of total area and degree (Brandt et al., 2018).However, the sustainability of large-scale reforestation is questioned (Zhang et al., 2022), and whether natural restoration can gradually form forests under geological constraint remains uncertain (Yue et al., 2023).
• The first-ever pollen record of a karst depression in southwest China is presented • Three stages of "virgin forestdeforestation-sparse tree planting" over the past three centuries were inferred • Maize cultivation since the 1780s triggered severe rocky desertification in the region

Supporting Information:
Supporting Information may be found in the online version of this article.
Understanding the forest evolution and rocky desertification history is critical to answering the reforestation potential and sustainability.While deforestation after the 1950s such as "Great Leap Forward" (1958)(1959)(1960) was recognized as an essential factor contributing to deforestation and rocky desertification (Jiang et al., 2014;Wang et al., 2004), historical records revealed that since the Ming dynasty (1368-1644), increasing population reclaimed more land for cultivation, destroyed most of the native evergreen forests, and led to severe soil erosion in the region (Yuan, 2008).Suitable crop varieties introduction and adoption promoted the reclamation process.Maize, a drought-tolerated and high-yield crop, was widely cultivated on steep slopes of mountains, aggravated deforestation and soil erosion (Chen & Kung, 2016).Though historical documents sparsely recorded the increasing population and deforestation activities, the specific process of rocky desertification and human impacts is still unclear.Recently, karst depression deposits and isotopic dating methods were used to calculate rocky desertification-related soil erosion rates in successive stages over past decades to centuries (Bai et al., 2010;Zhang et al., 2020).However, the divided stages could be too long to tell all the information about human impacts in the past centuries (Zhang et al., 2020), though the modern processes since the 1950s were well monitored (e.g., Wang et al., 2008;Zhao et al., 2022).Besides, the soil erosion rates failed to explain the vegetation type changes, hindering our understanding of forest evolution.It is required to analyze depression deposits with higher time resolution and different methods.
Here, we analyze the pollen percentage changes with high resolution (2 cm) in a depression of Northwest Guangxi, a region experiencing severe rocky desertification (Figure 1a).Combining the pollen data with multiple isotopic dating methods (i.e., 137 Cs, 210 Pb ex , and 14 C) and historical documents, we attempt to get a more comprehensive view of the forest evolution and rocky desertification history.In the confined environment, the depression sediment mainly came from the surrounding mountain slopes during rain wash and soil erosion (Bai et al., 2010).Correspondingly, the pollen in the sediment contains rich information about the plant evolution of the surrounding mountains.Our research questions are: (a) Was the rocky desertification area used to be a forest?(b) When the deforestation and rocky desertification started?(c) Which specific human activities led to plant-type changes and rocky desertification?This study presents the first-ever pollen record in the Karst region and forest/ deforestation evolution history over the past three centuries, which is vital to support rocky desertification control and reforestation sustainability in the geological constraint region.

Sampling Site Information
The sampling site Jiayu depression (107.92°E,24.22°N) is located in southwest Guangxi province, remaining a severe rocky-desertification region in 2021 investigation (Figure 1a).According to the closest meteorological record  at Du'an station, the average annual temperature is 21.4°C.The lowest temperature of 12.2°C occurs in January, while the highest temperature reaches 28.2°C in July.The total annual precipitation is 1,723 mm.
The depression occupied an area of approximately 1.4 ha, the catchment area is about 60.03 ha, and the basin elevation ranges around 394-671 m.The slopes at an average incline of 36°are covered by shrubland and needleleaf trees, with limestone rocks also exposed (Figures 1b and 1c).Because it is located in a closed environment (Figure 1b), the main source of sediment in the depression is from soil erosion of surrounding slopes.Different from slowly and-evenly-deposited marine or lake sediments, rates of sediment deposition in depressions vary greatly over time.When the vegetation in mountains was dense, the soil loss in slopes was slight (e.g., 10 t/ (km 2 •yr)) and corresponding deposition rate in depressions was low (Zhang et al., 2017).However, after the destruction of forest vegetation, erosion rates could rise sharply to thousands of t/(km 2 •yr).And when the easily lost topsoil was depleted, soil erosion rate became slight again, generally less than 50 t/(km 2 •yr) (Zhang et al., 2017).In 2022, a sediment core (diameter 5 cm) was sampled using drills at the depression's center.

Isotopic Dating Methods and Measurements
We used three commonly used isotopic dating methods in sediments, namely 139 Cs, 210 Pb ex , and 14 C.The dating principles of the three methods were described in Text S1 in Supporting Information S1 and also found in Zhang et al. (2020).Among them, 139 Cs and 210 Pb ex (abbreviation of excess 210 Pb) dating can determine the reliable time markers of 1963 and ∼100 years before the sample year (in our case, the year 1921), respectively (Mabit et al., 2008).Specifically, the bottom layer with higher 137 Cs activity than lower layers can be identified as the 1963 horizon, and the upper layer with 210 Pb ex value of zero indicates the 1,921 horizon. 14C dating can be used wherever organic materials (e.g., charcoals, teeth) in sediments are available.
Dry-weight raw samples were measured for 137 Cs and 210 Pb ex dating by gamma spectrometry with a highresolution coaxial germanium detector (GMX40P4, ORTEC) in Key Laboratory of Mountain Hazards and Surface Processes, Institute of Mountain Hazards and Environment, CAS.The activities of 137 Cs, 210 Pb, and 226 Ra were determined by peak areas of the 662 , 46.5, and 351.9 keV gamma-ray of 214 Pb, respectively. 210Pb ex was then calculated using the difference between 210 Pb and 226 Ra activities.The measuring uncertainties were generally less than 10%.
Charcoals were picked in the sediment for 14 C dating.A total of six charcoal samples in the sediment up to a depth of 238 cm were measured in the BETA laboratory in Miami, USA (https://www.radiocarbon.cn/).The calibration ages provided by BETA laboratory were adopted for the main analysis.

Pollen Analysis
All relevant sediment samples were smashed before the chemical treatments.Palynological extraction followed standard pollen analytical technique (e.g., Brown, 2008), including treatment with cold 10% HCl to dissolve calcareous minerals, and 40% HF to digest silica, respectively.Using a 7-μm nylon sieve to remove tiny particles then mounted in glycerin jelly.A pill of tablet containing a known quantity (10,315 grains/tablet) of Lycopodium spores was added to each sample to determine absolute pollen concentrations (pollen grains per unit of sediment weight, Maher, 1981).The percentages of pollen were calculated based on the total terrestrial pollen, but the percentages of spores were based on the sum of pollen and spores (Hao et al., 2021;Yang et al., 2019).Identification of major types of palynomorphs (pollen, spores and freshwater algae) mainly followed descriptions of Wang et al. (1995) and Tang et al. (2018).Particularly, the diameter of Zea pollen is significantly larger than general Poaceae pollen, which was adopted as a main identification of Zea pollen.

Data Sources
The desertification degree data in Figure 1a was derived from the 2021 investigation of the State Forestry Administration and Grassland Bureau.Local climate data was obtained from the nearest meteorological station, Du'an.Forest area reconstruction in the discussion combined data from the Guangxi Statistical Yearbook and references (He et al., 2007;Yang et al., 2018).Other historical documents we referenced include "Qingyuan Prefecture Chronicles" in Daoguang reign (1821-1851) and "Guangxi Yearbook" in the Republic of China (1921China ( -1949)).

Isotopic Dating Results
The 137 Cs and 210 Pb ex dating identified that the year 1963 and 1921 were located at a depth of 25 and 60 cm of the sediment, respectively (Figure S1 in Supporting Information S1).There are several possible periods for the 14 C dating calibration of charcoals at a certain depth due to atmospheric 14 C fluctuations over recent centuries (see an example of 14 C dating result at 32 cm in Figure S2 in Supporting Information S1).According to principles of stratigraphy (i.e., from bottom to top, oldest to youngest) and taking the two reliable time markers (i.e., 1961 at 25 cm and 1921 at 60 cm) as references, the most rational periods were inferred at different depths (Figure 2).For instance, there are four possible periods at the depth of 200 cm (Figure 2).Firstly, the age is expected to be older than 1921 at 60 cm, so the period of 1940 ± 10 is ruled out.Secondly, the age should be younger than the age at 238 cm, that is, at least younger than 1725 ± 54.Consequently, an inferred period of 1769 ± 34 is deduced from the remaining three periods (Figure 2).Except the dating result at a depth of 72 cm remains uncertain, ages at other depths can be determined reasonably.At two near depths of 100 and 104 cm, their calibration dating ages belong to the same periods and obey stratigraphy principles, indicating that the charcoal 14 C dating was reliable.

Pollen Type
One hundred two samples in the sediment at every 2 cm depth (total depth: 2.04 m) were used for pollen analysis.Pollen grains were presented in all samples, and the pollen counts ranged between 150 and 360. Figure S3 in Supporting Information S1 presents the photomicrographs of the main pollen and spores from the core.It is clear that the diameters of Zea pollen are several times longer than 20 μm, ranging from 50 to 100 μm, significantly larger than general Poaceae pollen.The results are similar to those of Palynological Database (https://www.paldat.org/pub/Zea_mays/304812).After identifying specific plant genera/families, we divided them into four main plant types categories: needle-leaf trees, broadleaf trees, herbs, and ferns.The average pollen percentage of needle-leaf trees (32.08%) is higher than that of broadleaf trees (11.96%) in the two-m core.Needle-leaf tree pollens mainly include Pinus (15.45%) and Cupressaceae/Taxodiaceae (7.97%).Broadleaf tree pollen mainly includes Quercus (3.6%), Aquifoliaceae (2.97%), and Cyclobalanopsis (2.31%).Average herbs' pollen percentage is 56.04%, with Zea (the genus to which maize belong) accounting for 3.67%.Ferns' spore percentage is 47.39%, to which the common pioneer species Dicranopteris contributed 7.50%.
Based on the temporal variation features of the main pollen types we focused on, combined with dating results, we divided three stages in the drilling core from the bottom to the top 20 cm (Figure 3).The top 20 cm was determined as local long-term tillage depth, according to the practices of cattle harrows and treadle plows (Revision Committee of "Series of Social and Historical Survey Materials of China's Ethnic Minorities", 2009).Therefore, the sediment in upper 20 cm experienced strong human disturbances, possibly affecting the accuracy of pollen analysis.Additionally, there are more documents revealing the modern forest changes, which was not our main focus.Thus, we did not take the tillage depth into our analysis.
Stage I (before the 1780s, at depths of 204-128 cm): It is featured by high pollen percentages of broadleaf (16.69%) and needleleaf (39.05%) trees.The main needleleaf trees are Pinus (23.05%).No Zea pollen is found in the stage.
Stage II (from the 1780s to the 1930s, at depths of 128-42 cm): Broadleaf and needleleaf tree pollens in the stage decreased to 10.97% and 18.80%, respectively, while Zea pollen appeared and Dicranopteris spore content surged from 0.92% to 12.18%.Stage III (from the 1930s to the 1960s, 42-20 cm): Broadleaf tree pollen further declined (6.12%).However, needle-leaf tree pollen increased sharply to 47.78%, among which Cupressaceae/Taxodiaceae experienced an apparent rise, from 0.70% in stage II to 33.38% in stage III.Zea pollen content was higher than the previous two stages.

The 14 C Dating Uncertainty in the Depression
With isotopic dating and specific plant type identification at the study site, we obtained more specific forest evolution and timing information over the past three centuries.However, it should be noted that the 14 C dating ages possibly contains great uncertainty in the depression.Older organic matters on the hillside might be deposited into the depression, leading to older-than-reality estimates of 14 C (Zhang et al., 2020).Moreover, because the depression has been affected by plowing for a long time, the 14 C dating results could even be earlier than the real ages.Additionally, the depression sediment experienced greatly varying deposition rates over time (Zhang et al., 2017), so we did not build an age-depth model like other stable deposits did.As an alternative approach, we only took the isotopic results as references and provided historical documents and local field evidence to support our inferences in subsequent discussion.

Forest Evolution and Rocky Desertification Over the Past Three Centuries
The pollen changes of different plant types and the three stages we divided for the sediment revealed a forest evolution process of "virgin forest-deforestation-reforestation" (Figure 4a).It is generally consistent with the historical forest area changes in the whole province, despite slight differences in the turning points of forest decreasing and increasing (Figure 4b).
In stage I before the 1780s, the pollen percentages of both broadleaf (16.69%) and needleleaf trees (39.05%) were high (Figure 3), indicating that the current barren mountains surrounding the depression probably used to a lush mixed broadleaf forest.More importantly, we noticed that this stage distributed the pine genus (Pinus) in large quantities.Considering that local pine species (mainly Pinus massoniana) favor acid soil (Ali et al., 2019), the result likely implies that alkaline carbonate bedrock properties did not constrain vegetation growth.However, given that Pinus pollen can transport with air masses far away from their source (Benkman, 1995), the possibility that the high Pinus pollen content was resulted from far-distance vegetation changes could not be excluded.
In stage II (1780s-1920s), a significant feature was Zea (the genus to which maize belongs) pollen appearance and Dicranopteris surge, indicating that maize cultivation triggered deforestation and rocky desertification.Maize was an important agricultural product introduced from North America into China in the mid-16th century.However, its diffusion was slow until the 18th century due to less population migration and geographic limitations (Chen & Kung, 2016).After the mid-18th century, particularly after the Yongzheng-Qianlong reign (1723-1735-1796), the maize diffusion accelerated (Chen & Kung, 2016).Thanks to national policies such as "Tax Assessment According to Farmland rather than Family Numbers" ("摊丁入亩" in Chinese), the Chinese population exploded by nearly 200 million (Lee et al., 2009).During our field trips in southwest Guangxi, we also found evidence of emigration to Guangxi during Qianlong reign in tombstone and family genealogy (Figures S4 and S5 in Supporting Information S1).Tremendous population pressure forced people to cultivate more land in the mountains, as there was no plain land for cultivation.Considering maize is tolerant of harsh environments and even suitable for growing on steep slopes, it was quickly adopted in mountain reclamation.The earliest Zea pollens appeared at 124 cm in the sediment, whose age should be earlier than 1780-1798 at 104 cm (Figure 3).Therefore, it is reasonable to infer that maize has been transmitted to southwest Guangxi before the 1800s, earlier than the cognition in the previous study (Chen & Kung, 2016).Because of the mountain reclamation, woody plant pollens were much less than before (Figure 3).Moreover, Dicranopteris spores surged at stage II, coinciding with maize pollen occurrence.Dicranopteris is a pioneering species that first grows in a barren land (Pang et al., 2018).The co-occurrence of Dicranopteris and maize implies that maize cultivation destroyed forests and accelerated ecological succession.This kind of deforestation and land reclamation continued until the 19th century.For example, local historical documents "Qingyuan Prefecture Chronicles" Volume 2 "Climate in Geography" in Daoguang reign (1821-1851) recorded the situation: "In the past, there were few people in the vast land, vegetation was lush, and the forest miasma was serious.Nowadays, the wild caves in the mountains are populated by Han people immigrated from Hunan, Guangdong, Guizhou, and Fujian.Wherever they go, slash-and-burn is performed, and the wood is all cleaned, so the miasma gradually disappears."The historical record supported our inference of land reclamation and deforestation of mountains at the stage II.
In stage III (1930s-1960s), the pollen percentage of Cupressaceae/Taxodiaceae increased abruptly (Figure 3), indicating the happen of reforestation.Statistical yearbook recorded that forestry was an essential economic industry in Guangxi in the Republic of China (1912China ( -1949)).According to "Guangxi General Chronicles • Foreign Economic andTrade Chronicles" from 1912 to 1931, the export volume of forest products accounted for at least 31.6% of the total value of Guangxi's primary export commodities, and the most reached 61.4% in 1 year.In the 37th year of the Republic of China (1948), the "Guangxi Yearbook" recorded: "The total number of forest trees in the province was 380 million, including 40.6% of natural trees and 59.4% of planted trees.According to the main tree species, the fir forest is 586,500 acres, the area of masson pine is 2,080,500 acres…" Until now, forestry remains a vital income generation in the region (Lu et al., 2021).Relying on ecological restoration methods such as reforestation is an effective way to eliminate the cycle of "deforestation-ecological degradation and intensified rocky desertification-reduced resources-intensified poverty."Also we noticed that deforestation during the Great Leap Forward (1958)(1959)(1960) after the founding of New China seemed to have little effect at the sampling site, as there was no significant change in the pollen percentage of trees during this period (32-25 cm).It possibly indicates that there were already few trees in that period and historical deforestation is the main cause of rocky desertification.

Implications for Reforestation and Rocky Desertification Control in Karst Area
Though the modeling study questioned the reforestation sustainability (Zhang et al., 2020), the pollen results imply that the karst limestone area possibly still has reforestation potential, given that the study site probably used to be a dense forest without bedrock constraining for a long time in history.It increased our confidence in controlling present rocky desertification in limestone areas.There are also successful cases worldwide in controlling rocky desertification.For example, the Kalas Plateau forest was destroyed and turned into a desert in the 18th century.Through reforestation, the forest covered more than 51% in 1989 (Jiang et al., 2014).The vital issue is how we rebuild historical soil and water environments to support vegetation richness and sustainability.

Limitation and Prospect
As a case study, the results can only reflect some regional environmental changes and rocky desertification history in karst areas.For example, the population migration and reclamation in Guangxi province had an "east to west" spatial pattern (Zheng, 2007), which our single-site analysis cannot reveal.Moreover, it remains to be seen whether the areas with dolomite bedrock (another carbonate rock type) used to be forests in history, as dolomites usually support grass rather than tree growth (Liu et al., 2019).
As a pioneering and positive start, we hope our study can encourage more sediment and paleoecology studies in different regions with different bedrock types to get a full-scale view of the forest evolution and human impacts in geological constraint areas temporally and spatially.We also underscore the significance of incorporating historical documents into the established palynological method.Similar endeavors have been undertaken in other regions, as exemplified by the work of Sluyter and Duvall (2016).This integrative approach holds the potential to foster increased collaboration between paleoecologists and historians.

Conclusion
We revealed the forest evolution processes of "virgin forest-deforestation-reforestation" over the past three centuries with sediment evidences.It used to be a mixed broadleaf-needleleaf forest.However, maize cultivation and tree cutting in the mountains started around the 1780s, inducing deforestation and rocky desertification.After the 1930s, sparse tree planting happened and increased economic benefits.Our study revealing the forest evolution and human disturbances in history can provide crucial evidences for answering reforestation potential and addressing rocky desertification control in the geological constraint region.

Figure 1 .
Figure 1.Geographical and environmental surroundings of the sampling site.(a) The sampling site is located in a severe rocky desertification area in Guangxi province, southwest China.(b) and (c) are pictures of the sampling site taken during two field trips in different seasons.The main vegetation types are grass and shrubs, with a small proportion of short trees.

Figure 2 .
Figure 2. The time markers identified by 137 Cs, 210 Pb ex and 14 C dating in the sediment.All 14 C calibration periods are presented, and the most reasonable ages are connected by dotted lines.At 72 cm, two 14 C calibration periods are possible.

Figure 3 .
Figure 3.The pollen records in the sediment.Three stages were divided according to the critical pollen percentage variations.The isotopic dating ages are provided on the left for reference.

Figure 4 .
Figure 4.The forest evolution history over the past three centuries.(a) A diagram of tree covers and soil environments at three stages at our sampling site.(b) Forest area variations in Guangxi province during 1730-1970.