Monitoring effects of land cover change on biophysical drivers in rangelands using albedo

: This paper explores the relationship between land cover change and albedo, recognized 12 as a regulating ecosystems service. Trends and relationships between land cover change and surface 13 albedo were quantified to characterise catchment water and carbon fluxes, through respectively 14 evapotranspiration (ET) and net primary production (NPP). Moderate resolution imaging 15 spectroradiometer (MODIS) and Landsat satellite data were used to describe trends at catchment 16 and land cover change trajectory level. Peak season albedo was computed to reduce seasonal effects. 17 Different trends were found depending on catchment land management practices, and satellite data 18 used. Although not statistically significant, albedo, NPP, ET and normalised difference vegetation 19 index (NDVI) were all correlated with rainfall. In both catchments, NPP, ET and NDVI showed a 20 weak negative trend, while albedo showed a weak positive trend. Modelled land cover change was 21 used to calculate future carbon storage and water use, with a decrease in catchment carbon storage 22 and water use computed. Grassland, a dominant dormant land cover class, was targeted for land 23 cover change by woody encroachment and afforestation, causing a decrease in albedo, while 24 urbanisation and cultivation caused an increase in albedo. Land cover map error of fragmented 25 transition classes and the mixed pixel effect, affected results, suggesting use of higher resolution 26 imagery for NPP and ET and albedo as proxy for land cover. 27


132
[20] at 30m pixel resolution were selected for land cover change analysis. Land cover classes included 133 grasslands (UG), shrublands, indigenous as well as invasive trees and bushes (FB), bare soils (BR), 134 water bodies (WB), wetlands (WL), croplands (CL), forests (FP) and urban, built-up (UB). As 135 described in [4,20], the existing South African National Land Cover map for 2000 [44] was adapted 136 to these eight classes through aggregation to conceptually broader classes [45] and manual editing 137 [4,33]. Supervised object-based image analysis using a rule-based decision tree classification of 138 Landsat 8 imagery was implemented to generate the 2014 land cover maps [4,33]. The overall 139 accuracy achieved for these maps was 84 ± 1% and 85 ± 1% for 2000 and 2014 respectively. Land cover 140 changes between T1 and T2 were analysed along with explanatory variables to generate transition 141 potential maps. Markov chain analysis was used to assign probabilities to potential changes to derive 2030 land cover maps, amounted to 23% and 16% of the catchment for S50E and T35B respectively 150 [20]. Nine land cover change trajectory labels were assigned to specific land cover transitions to relate 151 land cover change to specific landscape processes [4]. Landscape changes in the study area were 152 grouped into three land change categories [46,47]. Table 1 shows the land cover class transitions identified by trajectory labels with expected albedo change direction for each class transition, based 154 on literature values [36,48,49] for similar land cover classes, provided in brackets: (↑) to signify 155 increase, (↓) decrease or (-) no change. The land change category is also specified as abrupt      (Table 1Table 1

175
The land cover trajectory labels (Table 1Table 1

321
PSA generally followed an increasing trend in response to drop in rainfall, and a decreasing 322 trend in response to increased rainfall, when comparing Figure 4A and B with Figure 4C.

333
The correlation between mean PSA, NPP, NDVI and ET is reported in    Published albedo values are compared to similar land covers as those found in the study area 379 (Table 4Table 4).    Table S2 and Table S3.

423
Transition classes (   Table S2 and S3), however the affected area covers less than 2% of the two catchments. In

465
Structural changes occurred at all three points in 2007.

489
In the higher albedo scenario, the total modelled NEE in 2030 for persistent classes in T35B could      (Table 4), conversion from shrubland presenting a lower mean albedo than grassland,

536
should cause a gradual decrease in albedo of ~0.03 (Table 4). Contrary to expectation, the grassland 537 to cropland transition shows an increase in albedo. This increase in albedo may be related to the land

Catchment differences
Correlation analysis between PSA and the variables NPP, ET and NDVI at catchment scale (Table 3), showed similar trends with negative correlations between PSA and NDVI and PSA and ET.
material, Table S1). Intensification of wooded areas revealed different patterns in the two catchments:  (Table 4Table 4).

598
However, persistence of shrubland may be accompanied by densification of woody vegetation, Land cover change brought about by woody encroachment of grassland and particularly can potentially act as a carbon sink [13] due to increase in woody biomass [79]. Invasion of grassland 631 by IAPs can also reduce productivity due to loss of rangeland productivity for livestock production.

632
Acacia spp. are effective in utilising available resources more efficiently and may therefore outcompete

693
It is probable that a decrease in precipitation leads to desiccation of vegetation and soil, thus

712
Supplementary Materials: The following are available online at www.mdpi.com/xxx/s1,