3.1 Classification and Change Detection
We observed a remarkable increase in montane vegetation, while paramo vegetation decreased. Agriculture, urban developments, soil, and barren terrains showed subtler shifts (Figs. 2 and 3). Suesca and Fúquene lagoons experienced a noticeable reduction in their surface area (Fig. 1), whereas the reservoirs of Tominé, Neusa and Sisga, as well as other non-lagoon water bodies within the paramos, were surprisingly resistant to change.
Water bodies, soil and barren land did not show any major changes in cover area over time, and no urban development was detected inside the paramos since the beginning of the study period. However, when assessing each paramo separately, we found strong trends in the changes between LULC classes, portrayed in the corresponding thematic maps of the classification (Figs. 2 and 3). In the case of Rabanal (Fig. 2), montane vegetation exhibited a persistent increase, steadily rising by 19.5% from 1984 to 2021, which translates to an average 4.5% annual rate of change and a land cover expansion of 21.7 km2. Likewise, agriculture inside the paramo increased by 12.5% during this period, with an average annual rate of change of 4.9%. In contrast, paramo vegetation showed an astonishing 33.8% decrease since 1984, corresponding to an average annual rate of change of 1.3%.
Parallel trends were observed for the major LULC classes in paramo Guerrero (Fig. 3), though on a larger scale since Guerrero is about 3 times larger than Rabanal. An expansion of montane vegetation by 20.2% took place, transforming a landscape area that spanned 73.7 km2 (an average annual rate of change of 3.2%), with the largest increase observed between 1988 and 1990. At the same time a staggering 110.4 km2 or 30.3% of paramo vegetation was lost, with the most significant reduction occurring between 1989 and 1991, equivalent to an average annual rate of change of -1.5%. Agriculture grew by 41.2 km2 (1.6% annual change rate), with most expansion occurring between 1988 and 2000. The average annual rates of change varied between periods, with 1984 to 1991 showing significantly greater changes compared to later years.
Sentinel-2 classifications confirmed these tendencies: between 2016 to 2021, while agricultural land use showed a slight change, montane vegetation increased by 43.89% in paramo Rabanal and by 40.64% in paramo Guerrero. These changes represent striking average annual growth rates of 8.77% and 8.12%, respectively. Meanwhile, paramo vegetation showed a 11.5% reduction (average annual decline of -4.1%), while paramo Guerrero saw a 4.7% decrease (average annual rate of change of -2.4%). Consequently, declining paramo vegetation mirrored a surge in montane vegetation and agricultural lands (Fig. 4). The remainder of the LULC classes remained largely unaltered throughout the entire study period.
Landsat and Sentinel-2 classifications clearly distinguished paramo vegetation from agriculture and montane vegetation. Nonetheless, Landsat classifications revealed a greater extent of paramo vegetation and less montane vegetation compared to Sentinel-2 classifications, probably because of the heterogeneity of classes [90], and the detection of small-scale features in the higher resolution data [91]. Sentinel-2 classifications clearly indicated a larger paramo vegetation loss by 2021 compared to the Landsat classifications, showing an additional 5.3% and 5% paramo vegetation loss inside paramos Rabanal and Guerrero areas, respectively. Consistent with the continuous loss trend of paramo vegetation and gain in montane vegetation observed in both paramos over the entire study period.
It should be noted that a considerable portion of the study regions — 56.9% (210 km2) in Guerrero, and 60.9% (68.7 km2) in Rabanal — remained unaltered since 1984. Nonetheless, when considering the paramo vegetation cover alone, 47.2% in Guerrero and 39.2% in Rabanal were converted into forest and agricultural land (Fig. 5). These shifts corresponded to 73.3% and 83.5% of the total quantified change since 1984 in Guerrero and Rabanal, respectively. There was also a transition of paramo soils to agricultural use; this shift accounted for 2.54% and 5.29% of the total quantified change in Guerrero, and 5.1% and 20.1% in Rabanal, based on the respective classifications derived from Landsat and Sentinel-2 (Fig. 4). Built-up areas did not show significant change into other LULC classes, and as for soil and barren land, the major changes observed were between each other and to-and-from agricultural use. A comprehensive compilation of change detection statistics for both paramos and both sensors is presented in the Appendix, in the form of percentages and areas of each "from – to" transition (see Tables 5A, 6A and 7A for details).
Landsat-based change detection analysis from 1984 to 2021 in paramo Guerrero showed that a total of 158.9 km2 underwent transformation during this time frame, mostly loss of paramo vegetation, which accounted for 116 km2 (Fig. 5). Nearly 42% (66.1 km2) of the original paramo vegetation was replaced by montane vegetation. Another 31.7% (50.3 km2) of the paramo vegetation was replaced by agricultural activities. Other changes included the shift from agriculture to montane vegetation (6.9%) and the transformation from soil to agriculture (5.1%) (see Appendix; Tables 5A and 6A). The remaining “from – to” change categories, each representing less than 5% of the total change, did not yield significant insights.
Rabanal experienced substantial LULC transformations between 1984 and 2021. A significant proportion of the transformed area consisted of the conversion of paramo vegetation to montane vegetation and agriculture, accounting for 51.5% (22.6 km2) and 32% (14.1 km2) of the total change (43.9 km2), respectively. An additional 4.1% of the total change encompassed the conversion of paramo vegetation to soil, while the remaining “from – to” change categories individually represented less than 3% of the total change. A large portion of the land within the paramo area, comprising 68.7 km2 or 61%, remained unchanged throughout the study period. The percent change from paramo vegetation to other LULC classes is illustrated in Fig. 6 for every period within both paramos. Each period corresponds to the time interval between two different supervised classifications, as denoted in the Appendix (Table 2A).
Change detection analysis from Sentinel-2 showed the same tendencies within the paramo areas of Rabanal and Guerrero during the five-year interval from 2016 and 2021. Again, the most substantial transformation observed was paramo vegetation being replaced by montane vegetation, constituting 43.6% (11.1 km2) and 29.5% (23 km2) in Rabanal and Guerrero, respectively. However, there were nuanced variations in secondary changes: for Rabanal, the runner-up was the shift from paramo vegetation to agriculture, accounting for 16.3% of the altered landscape (Fig. 7), while in Guerrero, the transformation of soil into agricultural use represented the second-most substantial alteration at 20.1%. It is worth emphasizing that despite these notable changes, the majority of the paramo areas in both Rabanal and Guerrero remained unaltered during this briefer period, with 77.2% and 78.6% of these sensitive landscapes retaining their LULC classification between 2016 and 2021. Figure 7 provides a year-by-year breakdown of percent change from paramo vegetation to other LULC classes.