FARMERS’ PERSPECTIVE AND A GEOSPATIAL APPROACH ON THE DISTRIBUTION OF Mimosa pigra: A STUDY IN MATARA DISTRICT, SRI LANKA

Mimosa pigra L. is an invasive prickly plant that has been scattered in agricultural lands in the Southern Province of Sri Lanka. This study was aimed to understand the farmers’ perspectives on the spread of M. pigra in Matara district, Sri Lanka while uplifting awareness and estimating the distribution of the weed to support control activities. Semi-structured interviews with a pretested sample questionnaire survey were carried out among 400 randomly selected farmers among 16 Divisional Secretariats Divisions (DSDs) in the district. The spread areas of M. pigra were estimated using modern geospatial techniques, where images of unmanned aerial vehicle (UAV) (DJI Macvi Pro) were taken for mapping the affected area. UAV was captured 26.8 ha (5.84% of the total land area) of M. pigra invaded land extent in Akuressa and Malimbada DSDs. Based on the survey, 91.5 % of the farmers had no previous knowledge of controlling the particular weed. Approximately, 22.8% of the farmers expressed interest to establish Cinnamomum sp. as a perennial crop, and 11.8% suggested cultivating vegetables as the annuals in the areas invaded by M. pigra. The geospatial techniques used in this study was correctly identified the spread of M. pigra, which is very useful information for controlling weeds. Further research activities are suggested within the adaptive management framework and suitability of Cinnamomum sp. cultivation in M. pigra affected areas.


INTRODUCTION
Mimosa pigra (Giant Mimosa) is a native plant of Central America and becomes a serious environmental, agricultural, and economical issue in many countries since it has been introduced accidentally or deliberately (Groves 2002). It was first recorded in Sri Lanka in 1997, and the species is mainly confined to the Central and North Western Provinces of Sri Lanka (Marambe et al. 2014). According to Marambe et al. (2014), M. pigra has spread approximately 200 ha along the Mahaweli River and its tributaries. However, in 2012, it was found that fallow paddy fields were In tropical climates, M. pigra germinates around the year and grows very fast up to 6 m tall (Karim and Mansor 2013;Ostermeyer and Grace 2007). It prefers wetland places in the humid (flood plain, swamps, shallow dams, and riparian zones) and subhumid tropics with a wide variety of soils from clay to sandy clay (Karim and Mansor 2013;Okonkwo et al. 2016;Wongsiriamnuay and Tippayawong 2012). Concerns to the unified framework of biological invasion, M. pigra in Sri Lanka can be categorized under group 'E' (Blackburn et al. 2011) which constitutes only invasive species, with characteristics such as dispersal, survival, and reproduction in multiple sites across a greater or lesser spectrum of habitats (Blackburn et al. 2011;Richardson et al. 2000;Richardson et al. 2011). Thus, it needs to give more attention to the spread of M. pigra as an invasive plant because it is a serious threat to biodiversity with increasing magnitude in Sri Lanka.

FARMERS' PERSPECTIVE AND A GEOSPATIAL APPROACH ON THE DISTRIBUTION OF
M. pigra has potential impacts on biodiversity, the sustainability of agriculture, tourism, restrict stock movement and access to waterways, irrigation projects, competing with pastures, and the traditional lifestyle of farmers (Karim and Mansor 2013;Ostermeyer and Grace 2007;Thi et al. 2011;Walden et al. 2004;Wongsiriamnuay and Tippayawong 2012). M. pigra incurs economic losses to paddy farmers, anglers, ranchers and greatly impacts wetland ecosystems following Australia, South East Asia, and Africa (Rijal and Cochard 2016) which can similarly be observed in Sri Lanka.
In the control of M. pigra, the living mimosa plants were difficult to burn and cut, because the stems re-sprouted quickly after treatment (Thi et al. 2004). Mimosa seeds can survive in fire (Paynter 2004b) and fire-triggered mimosa seed germination (Thi et al. 2004). Therefore, it is clear that the burning land is prone to rapid recolonization (Paynter 2004b). Moreover, it is also tolerant to flood (Wongsiriamnuay and Tippayawong 2012).
Control of M. pigra is possible using herbicides at a low cost (Beilfuss 2007;Rijal and Cochard 2016;Thi et al. 2004

Distribution of M. pigra Study area and collection of UAV data
According to the questionnaire survey findings, it was understood that studying the distribution of M. pigra was important in the study area. Therefore, three Grama Niladhari Divisions (GND) (Paraduwa South, Paraduwa East, and Pahala Kiyaduwa) in Akuressa and Malimbada DSDs that showed high infestation was selected. The geographical setting of the study area according to the projection system WGS_1984_UTM_Zone_44N as shown in Fig. 1. The total land extent of the study area is around 1,134 ac.
DJI Mavic Pro was used as the UAV platform to collect remote sensing data of the study area. The UAV has a stabilized camera mounted on it. The focal length of the camera is 2.2 mm and it can produce imageries in visible spectral (RGB) bands. The UAV platform is specifically suitable for studying vegetation. The resolution of the imageries at the typical flying height of 50 m is around 2 cm (DJI 2019).
The first step of UAV data acquisition was the preparation of the flight plan to cover the whole study area. For this purpose, a mobile application of Pix4D Capture was used. A predefined flight plan with a constant shutter speed of the UAV camera was used to ensure the best coverage of the study area and without affecting the motion of the UAV for the quality of the images (Drone Deploy 2019). The GPS receiver of Mavic Pro UAV supports the Global Navigation Satellite System (GNSS), which can access the best combination of satellites and provide positional accuracy of around 3 m (Suzuki et al. 2016). UAV data was acquired to cover key growing areas of M. pigra. The UAV data were acquired between 9-11 a.m. with the flight parameters, 50 m flight altitude, 90o camera angle (orthogonal view or nadir position), 80% front overlap, and 60% side overlap. The attitude helps maintain the spatial resolution of the UAV images around 2 cm. The overlapping (front and side) allows for correcting mosaicking by matching key points of the images and compensating wind disturbance and GPS errors to generate a complete mosaicked orthophoto (Drone Deploy 2019).

Data description and data processing
The list of data sources used to estimate the spread of M. pigra in the study area is shown in Table 1.
Agisoft PhotoScan was used to process the collected UAV images. The software allows generating mosaicked orthophoto of the study area by using multiple images captured by UAV. The collected UAV images were required to import into the software for preprocessing.
The software allows

Extraction of land cover
Four types of machine learning (ML) techniques, namely, support vector machine (SVM), k-nearest neighbour (KNN), random forest (RF), and neural networks (NN), were used for the extraction of the land cover of the study area with the use of R software (Priyankara et al. 2019). Both overall accuracy and the kappa statistic were higher (over 90%) with the SVM, and the method was selected to extract the land cover of the study area. Mosaicked-orthophoto of the study area was classified into two land cover types, namely, M. pigra (MP) and other lands (OL) for this study. The accuracy of the classified land cover map was evaluated using 100 reference points. Stratified random sampling techniques were used to generate the reference points (Stehman 2009). Mosaickedorthophoto and Google Earth were used as a source of reference information for the land cover map.

RESULTS AND DISCUSSION Farmers' perspective on the spread of M. pigra
The majority of the farmers (48%) were between the ages of 41 and 60, while 43.3% were between the ages of 21 and 40. The remaining farmers were less than 20. Moreover, 55.2% of the farmers were recorded as male. There was no correlation between gender and age in the awareness and identification of M. pigra (p> 0.05; r = 0.21029).
According to the results of the questionnaire survey, the level of awareness of M. pigra among the participants was 53.2%, while more than 70% had no idea about this plant's invasiveness (Table 2). It was noted that more than half of the respondents could identify M. pigra. The majority of the participants (47.6%) stated that the place of distribution of M. pigra was not known while 32.2% stated that they had seen the growth of the plant in Athuaraliya (4.7%) and Akuressa (32.2%) areas and it was very lower level (< 1%) in other areas ( Table 2). The findings of the study proved that the flood plains are the most favourable habitat for M. pigra in this area as reported by many researchers (Beilfuss 2007;Walden et al. 2000;Okonkwo et al. 2016).
The participants of 55.2% said that they do not know the exact date of the first observation of the plant in the area. However, 29% of the participants stated that they have encountered M. pigra during the period 2009 -2014, which seems like an outbreak. 7.2% and 6% of participants said they observed the plant in the area during the periods 2003 -2008 and 2015-2018, respectively (Table 2). According to the responses from the participants, the number of encounters was recorded in the period 1997-2002, which was about 2.5%. Results were given the idea about the spreading magnitude of M. pigra in the surveyed areas.
In response to the farmers' awareness of the extent of the distribution of the plant in the area, about 74.6% of them do not have any idea about the extent of the distribution ( Table 2). The participants of 13.7% stated that the land extent of the spread is around 26 -50 acres (Table 2). Furthermore, 48% of them stated that the status of the land in which the highest occurrence of M. pigra plants is not known to them (Table 2). However, 21% stated that the occurrence is highly seen on low land while 7.7% stated that it is seen on uplands (Table 2). Additionally, 8% of the participants stated that the occurrence is similar in both uplands and lowlands.
The contribution from the media to increase the farmers' awareness of the spread of the plant was minimal. Many farmers (46.8%) said that they did not hear anything from the media related to the spread of the plant, but some farmers acquire information from various sources including newspapers/ magazines (14.4%), television/radio (13.4%) ( Table 2). In addition, mostly, they got to know about the spread of the plant from their day-to-day working experiences in the fields and other experienced farmers in the area. Despite the fact that the internet makes people aware of a wide range of issues, the information gained through the internet was minimal (0.5%) in this case. It might be because of the barriers related to accessibility to the internet, such as lack of devices, connectivity, and knowledge. However, the farmer's community has a very low interest to observe alien species (Nghiem et al. 2013). Thus, this reason will be caused to spread of several invasive species like M. pigra belonging to other countries.
M. pigra reproduces primarily by prolific seed production and vegetatively through cut stems (Okonkwo et al. 2016). However, many farmers did not know about the exact propagation method (47%), while 44.8%, 2% and 6.2% of farmers respectively stated seeds, vegetative parts and both as main propagation methods of M. pigra ( Table 2). The results also revealed that the awareness of propagation of M. pigra is important since farmers are using M. pigra sticks to support the vine crops in their fields.
In response to the mode of distribution, farmers reported various methods such as flood (31.3%), wind (21.3%), animals (3.8%), animal faeces (5.6%), and human activities (7.5%). Whereas the rest (29%) did not have an idea about the method of propagation (Table 2). M. pigra produces copious light seeds which can float on the water and are mostly carried from one place to another by floating on water (Lonsdale 1993; Rijal and Cochard 2016;Karim and Mansor 2013). The shreds of evidence that plants have been colonized along the riverbanks were reported in floating plains in Adelaide River in Northern Australia, Mekong River in Cambodia, and Mahaweli River in Sri Lanka and (Marambe et al. 2014;Rijal and Cochard 2016). Because of the annual flood of the Nilwala River, farmers are not willing to collide with M. pigra that is similarly observed by Rijal and Cochard (2016) in Cambodia and Thi et al. (2011) in Vietnam. The segments of M. pigra pod are enclosed with bristles that facilitate them to adhere to animals and clothing. The seeds are also dispersed adhering to vehicles and other machinery (Walden et al. 2000). Table 2, among the croplands, 33.3% were paddy, and paddy lands were highly affected than tea (0.7%), rubber (0.7%), and cinnamon (0.3%) lands. The farmers said that the most prominent types of plants are vines (20%) followed by bushes (14.4%). Other than that, the calculated percentages show that the occurrence of trees, grasses, and bushes accounts for 7.7%, 6.5%, and 3.7%, respectively (Table 2). Therefore, it is clear that any type of plant can be seen along with M. pigra, especially vines and bushes. Nearly half of the participants (49.3%) said that there is an impact to grow other plant species with M. pigra and the rest did not have any clear idea about that.

As mentioned in
Few questions were raised on the importance or usages of the plant, where, 91% of the participants said that there is no importance while the rest highlighted the following importance such as compost making (6.5%), mulching (1%), firewood (1%), animal feed, and fencing (0.4%) ( Table 2). The literature says that M. pigra is used in folk medicine, green manure, poles, hedges, as ornamental plant, erosion control, animal feed, timber, temporary fences, and fuel wood (Shorinwa et al. 2015;Wongsiriamnuay and Tippayawong 2012;Okonkwo et al. 2016). Here it is convinced that M. pigra utilization as an energy source may be a useful option.
Animal husbandry is also an important livelihood in the Southern Province of the country, hence, the impact of this invasive weed on animal husbandry was evaluated. According to the participants, the majority (55.5%) did not have a clear idea about the effect of the plant on animal husbandry. However, 39.1% stated that there was an impact on animal husbandry. Accordingly, the main impact on animal husbandry was the reduction of grassland (34.4%), ultimately causing feed stress on grazing animals (Table  2). Further, the distribution of this plant is causing injury to the animals ( Table 2). The impact is not only limited to farm animals. 15.2% of the participants mentioned that other animals are also affected by this plant species and 26.1% said there is no impact on other animals. Wongsiriamnuay and Tippayawong (2012) observed that the thorny M. pigra replaces grasslands and reduces available habitat for animals. The effect of M. pigra on the reduction of waterbird population was also observed by Shanungu (2009) and Walden et al. (2004), and an unsatisfactory microhabitat for lizards and amphibians were also reported by Walden et al. 2004. The main impacts on other animal species are loss of habitat (6.2%), feed shortage (3.7%), injuries (3.7%), and reduction of fish (1.7%) ( Table 2).
The awareness of different available control methods is vital to control this invasive plant species in the region. The participants of 70% were not aware of the available control methods. The most known controlling methods were burning (5.1%), uprooting (7.3%), and both uprooting and burning (8.1%) ( Table 2). The effectiveness and success of the controlling methods were least known by most of the participants (91.5%). The participants of 7.2% stated that the controlling methods they were applied are moderately successful. The results revealed that there is a low concern about control methods of M. pigra.
Cutting and/or digging out of mimosa plants was commonly practised in fields (Rijal and Cochard 2016), including in Sri Lanka (Personal observation, Perera PCD). However, cutting and burning is not effective (Thi et al. 2004). M. pigra controls using herbicides (e.g. paraquat, metsulfuron methyl, triclopyr butoxy ethyl ester, and glyphosate) (Son et al. 2004), however, it is not an environmentally friendly method and it will also affect the native plants. The invasion of mimosa in rice fields has increased expenses for soil preparation and for labour to remove mimosa before cultivation (Thi et al. 2011;Rijal and Cochard 2016) which is similar to the context in Sri Lanka. Most of the participants (76.9%) did not have any awareness and idea about the control measures taken by the government. The participants of 23.1% knew about there are some government projects to control this plant but with less success. Awareness programs were suggested to conduct among the communities in the affected areas. Marambe et al. (2014) mentioned that they have resulted in several community participatory activities to eradicate small patches of M. pigra from the Central Province.
Considering the management practices of M. pigra, ecological research in Australia showed that fire may burn the vegetation, but the plant easily reestablishes from soil seed banks (Rijal and Cochard 2016). Only a small number of potential biocontrol agents showed any measurable impacts on plant vigour and seed banks (Heard 2012). Therefore, biocontrol and fire, the use of herbicides, mechanical control, restoration, and management of native vegetation have been focused understudying integrated approaches (Paynter 2004;Paynter and Flanagan 2004;Thi et al. 2004). Thus, the respondents' suggestion of the establishment of cinnamon on M. pigra invaded land will be a new aspect and can be implemented after further research studies as the control of M. pigra.
According to the above results were revealed that there is no exact idea about the distribution of M. pigra in the affected area. It is essential to identify the exact distribution of M. pigra before going to plan the control, strategy, and control of the further spread of M. pigra for future administrative works. Table 3 shows the results of the accuracy assessment of the technique used to extract the land cover of the study area. The overall accuracy and the kappa statistic are 96.2% and 0.94, respectively, which shows how accurate the method is for identifying the land cover in the study area. Fig. 2 shows the results of the extraction of land cover and Table 4 shows the relevant descriptive statistics of the extraction of land cover, in the study area. The total land extent of the study area is 458.83 ha, and 5.84% of the land extent of the total extent already was invaded by M. pigra. The spread of M. pigra is more serious in the GNDs Paraduwa South and Paraduwa East than Pahala Kiyaduwa.

Spatial Distribution of M. pigra
The study was tried to apply some of the novel technologies to identify the spread of the plant as a case study, which is limited to a small area but can be applied to a large extent in a practical context with some limitations. The method was able to detect the spread of the plant spatially and also its extent preciously. The findings will ultimately help to be aware of the farmers in the study area about the spread of the plant in spatial means and the findings will help to take the necessary control measures to eradicate the plant from the study area for the continuation of the agriculture activities. The UAV-based remote sensing method applied to estimate the distribution of the spread of the plant is very accurate in comparison with other methods applied previously to identify the spread of this kind of invasive plant. In the case study, it was considered only two types of land cover, namely, land affected by the plant and other lands. It can be further improved by identifying other land use/land cover types to evaluate the impact of the plant on other activities and can be considered in future studies. The method is flexible to carry out in any field where accessibility is limited and possible to monitor the progress of the controlling activities of the plant frequently. The cost for the identification, monitoring and controlling activities of the plant with the application of the method is considerably lower than other methods. The method is   having light limitations such as shorter flight duration, difficulties in maintaining flight altitude, the stability of UAV, and manoeuvrability due to wind and turbulence (Sylvester, 2016).

CONCLUSION
Farmers' awareness of the spread of the plant, its distribution, and extent were very minimum in the area. Cultivation of cinnamon and vegetable in the affected areas is a viable option to control the spread of the plants. UAV was captured 26.8 ha (5.8 % of the total land area) of M. pigra invaded land extent in Akuressa and Malimbada DSDs among the total land extent of 458.83 ha. The novel UAV-based remote sensing methods can play a vital role in the estimation of the distribution of the invasive plant preciously. Further research studies are suggested within adaptive management frameworks and suitability of Cinnamomum sp. cultivation in M. pigra affected areas.

ACKNOWLEDGEMENT
The authors thank the staff of the Department of Agriculture, Southern Province, and the University of Ruhuna, Sri Lanka. This study was funded by the Department of Agriculture-Southern Province of Sri Lanka. We would also like to thank anonymous reviewers whose comments helped to improve the article.

AUTHOR CONTRIBUTION
PCDP conceptualized and designed the study. PCDP, PP, and APSM conducted the survey and collected the UAV data. PCDP and PP analyzed the data. PCDP and PP wrote the paper with input from all authors. All authors discussed the results and commented on the manuscript.