Remote sensing monitoring of mangrove growth rate at selected planted sites in Mauritius

sites (Le Morne and Grande Rivière Noire) on a southern African island – Mauritius – using Google Earth Pro historical Landsat 7 and Landsat 8 images. Data were processed using ImageJ software. To our knowledge, this technique has not yet been applied for monitoring mangrove growth. The mangrove sites were classified into four zones based on water level and tidal variations. On average, the rate of increase of canopy coverage expressed by a coefficient ‘ b’ at Le Morne ( b = 1.901) was higher than that at Grande Rivière Noire ( b = 1.823). The coefficient ‘ b ’ positively correlated with the zonations (r ~ 0.8). Higher ‘ b ’ values (2.319–2.886) were observed in Zone 1, where the substrate is always covered with water at low tide. The use of remote sensing data along with image processing analysis proved to be an effective tool to obtain relevant information, not only for mapping mangroves but also for monitoring the canopy growth rates of planted mangroves.


Introduction
Mangroves are trees and shrubs that thrive in the harsh conditions between the land and sea. Mangrove forests form part of the most productive and unique ecosystems on earth. They are ecotone ecosystems occurring mostly along the tropical and subtropical coastlines. 1 They adapt well to inter-tidal conditions, and play a vital role in the aquatic food web by providing a plethora of ecosystem services, particularly as a breeding ground for several fish and prawns species and as a food source for aquatic organisms. They also provide goods and services to people in fisheries; coastal protection against storm surges, rough waves and erosion; pollution abatement; and forest products. 2 Mangrove forests are also important ecosystems for carbon sequestration, allowing carbon to be stored in their biomass and sediment. 3 Even though mangroves are of prime importance to coastal ecosystems, the mangrove population worldwide is being threatened by anthropogenic activities and climate change. 4,5 Mangroves are found in 123 countries and territories, and, as of 2016, their global coverage 6 was around 136 000 km 2 although mangrove forests in the southern African region account for around 7% of this area [7][8][9][10][11][12][13][14][15] . The largest southern African mangrove forests are found along the Indian Ocean coasts. 9 Mozambique and Madagascar each harbours more than 3000 km 2 (20% of African mangroves), making together 4% of the global distribution. 10,11 In South Africa, the mangrove cover is estimated to be around 2000 km 2 , 9,14 while for Tanzania mainland, Wang et al. 12 estimated that there is about 1083 km 2 of mangrove cover. 13 Mauritius, a small island state off the southeast coast of the African continent, has a mangrove cover of 1.45 km 2 . 15 Worldwide, a net loss of around 4.3% of mangroves was noted in the 20 years preceding 2016, although the average rate of mangrove loss is now reported to be slowing. 6 Given the significance of mangrove forests, there is a need for continuous monitoring of their dynamics. However, precise, dependable and timely data on the world's mangrove forests are not readily available. 16 In southern African countries [17][18][19][20][21] , and other regions with wide expanses of mangroves (e.g. Indonesia 22 , Malaysia 23 and Thailand 24 ), geographic information systems (GIS), based on digital satellite and aerial photographs, are most commonly used to create maps showing mangrove forests. 25 Remote sensing techniques are ideal for inaccessible areas where in situ field data cannot be undertaken. Development in remote sensing with high spatial, spectral and temporal resolution, and historical remote sensing data provide the opportunity for better characterisation, mapping and monitoring of mangrove forests from local to global scales. 26 Additionally, it allows 'indirect' access to mangrove habitats located in remote areas and areas that are usually temporarily swamped [25][26][27][28] , thus allowing scientists to focus their research on specific levels of ecological details 26,29,30 . Light detection and ranging (LiDAR), hyperspectral and multispectral optical images, and synthetic aperture radar (SAR), are among the satellite image data studied in addition to aerial imagery. The three types of digital image classification algorithms utilised for mangrove mapping and monitoring in diverse research are object-based, pixel-based, and knowledge-based classifiers. 31,32 By the determination of the percentage canopy closure of mangrove forests, further investigations based on the density of the mangrove area can be undertaken. 30 The main objective of this study was to use remote sensing and image processing techniques to determine the extent of canopy coverage of mangrove forests and their distributions as part of our efforts to understand the ecosystem and ecophysiology of mangals. In this work, canopy coverage was established using image classification (pixel counting) with data acquired through the Google Earth engine Google Earth Pro TM . This novel method developed here can be readily extended to study the temporal and spatial evolution of mangrove areas globally.

Study area
This research study was focused on the two largest planted mangrove areas on the southwestern coast of Mauritius. The Republic of Mauritius is a small island developing state with the mainland, Mauritius, centred around 20°34'84" S and 57°55'22" E whilst other islands of the republic are scattered in the South-West Indian Ocean. Mauritius has a coastal zone inhabited by two mangrove species, namely Bruguiera gymnorrhiza (L.) Lam. and Rhizophora mucronata Lam., covering ~1.45 km 2 . 15 Mauritius has been experiencing a wide range of changes in its coastal zone over the last few decades due to anthropogenic activities (such as global warming and deforestation) and invasive plants, resulting in a loss of mangrove biodiversity. Since 1995, the Albion Fisheries Research Centre, under the aegis of the Ministry of Blue Economy, Marine Resources, Fisheries and Shipping, embarked on a Mangrove Propagation Programme to protect and restore denuded areas. As of 2019, the total area covered under mangrove propagation on mainland Mauritius was ~0.217 km 2  with their plot delineated by yellow boundaries; Le Morne (S1-S17) and Grande Rivière Noire (S1-S3). The mangrove areas were classified into plot sites based on patches under mangrove cover and labelled LM (S1-S17) and GRN (S1-S3) for Le Morne and Grande Rivière Noire, respectively ( Figure 1).
The planted mangrove plots were then divided into zones based on the tidal water-level variations as follows: • Zone 1: Substrate always submerged with water at low tide • Zone 2: Substrate exposed at low tide only • Zone 3: Substrate exposed at intermediate times between high and low tide only • Zone 4: Substrate exposed at high tide

Image and field data sets
The sites were studied using satellite imagery from Landsat 7 and 8 Google Earth Pro TM (available free to the public) on spatial and temporal scales. Landsat 7 imagery, of 15-m spatial resolution, was used for data before 2013, and Landsat 8 imagery, of the same resolution, was used for data as of 2013. The imagery was selected in reference to the time periods from the year of propagation for each study area, 2003 for GRN and 2012 for LM, up to 2021.

Image processing techniques and data analysis
Google Earth Pro images throughout the years were selected based on the image resolution, cloud cover, time period and colour correction. From 2004 to 2009, no satellite images were available through Google Earth Pro. Image classification (pixel counting) techniques were then applied by using ImageJ (National Institutes of Health and the Laboratory for Optical and Computational Instrumentation), an open-source software for image processing. The satellite images were adjusted to a black and white threshold colour, and a number was attributed to each pixel (0 and 255). The pixels covering the mangrove plots were then counted to retrieve the percentage canopy cover. Figure 2 represents a typical plot in 2013 and 2021, after image processing techniques, for estimating the percentage canopy cover. The percentage canopy cover was then plotted in graphs using the function, y = ae bt , where a and b are constants. The equation was linearised to extract 'b', the coefficient representing the rate of increase of canopy coverage. This function represents the onset of the expected sigmoid-shaped growth curves corresponding to three phases. 33 The first initial phase or lag phase represents the initial growth stage, the second phase or log phase represents the exponential period of growth, and the third phase or the stationary phase represents the steady growth stage.

Field assessment
To estimate the percentage substrate underwater at high and low tides for each mangrove plot by Google Earth Pro, the GPS positioning at high and low tides was recorded using a GPS phone tracker. The pH, dissolved oxygen level and salinity were measured using portable instruments, namely a digital pH meter, dissolved oxygen meter and refractometer, respectively. The tidal water-level variations were monitored with a metre rod.

Statistical analysis
Tukey honest significant difference tests were done to detect differences between means. Correlations between the site assessment parameters and the linear equation function values were carried out using bivariate Pearson correlation analysis using the IBM SPSS Statistics 21 software.

Results
The rate of increase of canopy coverage, 'b', for each plot was determined from the second phase representing the exponential period of growth, and these were then correlated with chemical and physical parameters recorded at the mangrove sites to determine the factors affecting the growth and spread of the canopy cover.

Canopy cover pattern
The sigmoid-shaped growth curves for representative sites in the four tidal zones are illustrated in Figure 4. It is noted that, at LM, the plots started to reach their steady state (>95%) after ~6.4 years while at GRN it was achieved after ~15.4 years. The exponential canopy growth for LM, as reflected by the 'b' values (Table 1) and time taken to reach the steady growth stage, was as follows: t Zone 1 > t Zone 2 > t Zone 3 > t Zone 4 . At GRN, a similar pattern was obtained where t Zone 2 > t Zone 3 .

Chemical parameters
The variation of pH and dissolved oxygen across the sites studied was found to be insignificant (p>.01). The salinity values, recorded using the practical salinity scale at LM and GRN (Table 2), were found to vary based on the positioning of the plots (Figure 1). Salinity values for LM (S1, S2 and S3) in Zone 1 and LM (S4 and S5) in Zone 2, which were all on the seaward side, ranged from 35 to 36. The salinity in the rest of the LM mangrove plots, irrespective of their zones, varied from 31 to 36. The salinity at GRN (S1 and S2) ranged from 22 to 30, while that of S3, which was in close proximity to a river, ranged from 5 to 21. The 'b' values of all mangrove plots under study displayed a positive correlation with salinity (r=0.438, p<0.01).

Discussion
The mangrove canopy for the two sites, LM and GRN, was compared over time. Higher 'b' values were recorded at LM (1.392-2.886) than at GRN (1.752-1.957). Because the mangrove planted under the Mangrove Propagation Programme at LM (2 seedlings per m 2 ) was denser than that at GRN (1.68 seedlings per m 2 ), it is expected that GRN plots took more time to reach a percentage canopy cover of >95% (stationary stage). However, adjustments for the plant densities indicate that the 'b' values of GRN (~2.3 for Zone 2) are slightly higher than those of LM (~2.0 for Zone 2); which is expected given their proximity and hence similar climate conditions, but with the added advantages of slightly longer hours of sunshine and a wider salinity range for GRN sites.
The two sites were further investigated zone-wise. The 'b' values were as follows: b Zone 1 > b Zone 2 > b Zone 3 > b Zone 4 (Table 1; Figure 4). As GRN plots correspond only to Zones 2 and 3, they showed more or less the same tidal variation, thus explaining their similar 'b' values. These results suggest that higher 'b' values relate to regions with longer periods of tidal inundation. This finding is in line with studies carried out by He et al. 34 and Hoppe-Speer et al. 35 , who found that Rhizophora mucronata was healthier in inundated areas compared to non-inundation zones. Similarly, Jackson and Drew 36 and Adams 37 reported that, as a response to prolonged inundation, estuarine plants grow more rapidly to increase the biomass over the water surface.
The pH and and dissolved oxygen recorded at LM and GRN were in the ranges 7.

Conclusions
This study highlights the potential use of remote sensing techniques along with image processing for mapping and monitoring mangrove Table 1: Zonation, percentage substrate underwater at low and high tides, percentage of canopy cover as of 2021, area of mangrove plot and the coefficient of rate of increase of canopy coverage, 'b' at Le Morne (S1-S17) and Grande Rivière Noire (S1-S3)