Conservation Policy Changes in Protected Areas on Hilltops: Effects on Hydrological Response in A Small Watershed


 Public policies affecting land use/land cover also have an impact on water resource availability, and hilltop protected areas are a relevant factor in ensuring continued availability of water resources. The legislation ruling the delimitation of protected areas on hilltops has changed at the national level in 2012 and in Rio de Janeiro state in 2014. However, these environmental legislation changes did not take into account the feedback effects of restricting protected areas to hilltops on the regularity of hydrological responses in watersheds. As such, this manuscript sought to analyze the contribution of hilltop-only protected areas to continued water availability. We analyzed hydrological responses in the São João river watershed, which provides water for domestic, industrial, and agricultural uses in the Região dos Lagos municipalities of Rio de Janeiro state. Our results show that designating only hilltops as protected areas, as prescribed under the new pieces of legislation, does not prevent abrupt changes in hydrological responses that can lead to changes in streamflow volume and regularity as well as increases in sediment flows, which may compromise drainage systems and continued water supply due to reservoir silting. Therefore, we conclude that protecting hilltops only, as established under current Brazilian legislation, is not sufficient to safeguard the environmental function of maintaining water resource availability.


Introduction
Rainforests cover about 7% of the Earth's surface, with estimates pointing to them being home to two-thirds of global biodiversity Rainforests also perform functions that are strongly correlated with biodiversity, as evidenced by their direct in uence on ecosystem services (e.g., water supply) and regulation (e.g., water availability and quality and the mitigation of extreme events). To some extent, ecosystem services and biodiversity can be considered synonyms (Andrade et al., 2020;Schröter et al., 2019). This strong overlap allows for more synergies in environmental protection strategy and public policy making, as the aims of both are to have landscapes whose functions or processes contribute, directly or indirectly, to human resources associated with fauna and ora variability (Costanza et al., 2017(Costanza et al., , 1997. the hydrological responses of watersheds, given that 95% of the water that supplies drainage systems comes from hills in mountainous areas, with possible consequences for availability to the general population (Hallema et al., 2016;Shakya and Chander, 1998; Vieira et al., 2018).
Brazil's legal framework is based on the concept of the Kelsen Pyramid, where hierarchically higher standards institute and regulate the creation of methods used in hierarchically lower standards (Kelsen, 1967). As such, the Brazilian Constitution gives the states the right to enact their own environmental standards, provided those are in compliance with the standards established by the Federal Government (Brasil, 2016). In 2014, the Rio de Janeiro State Environmental Institute (Inea) published Resolution no. 93 ("Inea Resolution") to prevent gaps in the delimitation of PPAs on hilltops and to preserve the landscapes of the Rio de Janeiro state, in compliance with the criteria presented by the Forest Code (Frega, 2014). ). As such, we sought, in this manuscript, to assess the contributions of hilltop protected areas to hydrological responses and sediment control in the São João river watershed, located upstream of the Juturnaíba Reservoir in the state of Rio de Janeiro.

Study area
The São João river watershed ( Fig. 1) is responsible for the water supply of the Região dos Lagos municipalities in Rio de Janeiro state. The study area is upstream of the Juturnaíba Lake, which became a reservoir in the 1970s. The watershed has an area coverage of 577 km², and its main channel has been widened and straightened as a result of agricultural activity (Ikemoto and Napoleão, 2018). Moreover, it features a gauge station, called Correntezas Nova (catchment area cover of 400 km²), from which one can obtain stream ow data over long-term data time series.
The Juturnaíba reservoir (Fig. 1), apart from accumulating the largest water volume, is also used for ood control purposes in the São João river oodplain and to ensure water availability for farmland irrigation downstream of the river. The reservoir is used in two water supply systems. The rst system serves the Araruama, Silva Jardim, and Saquarema municipalities, with a combined population of ~ 340,000 people, while the second system supplies coastal municipalities with large tourist in ows in summer -Armação dos Búzios, Arraial do Cabo, Cabo Frio, São Pedro da Aldeia and Iguaba Grande -and a total population of ~ 715,000 (including permanent residents and seasonal in ows) (Cunha, 1995;Ikemoto and Napoleão, 2018).
The watershed's relief is characterized by steep hillsides and an abrupt transition thereof to the oodplain, as well as the predominance of dissected and isolated gently sloping hills. Natural vegetation is the Atlantic Rainforest, a forest complex with high species richness and endemism rates. However, only ~ 15% of its original cover in the Brazilian territory remain, dispersed in fragments of which 80% are smaller than 0.5 km² in area. Furthermore, only 9% of the remaining forest and 1% of the original forest areas are in protected areas (Ribeiro et al., 2009;Wagner et al., 2020).
The anthropic interventions carried out in the São João river modi ed the meandering channel patterns to rectilinear, reducing river size by 40% (Cunha, 1991). This, combined with the construction of the Juturnaíba reservoir, changed local morphology from meandering channels to decantation basins and removed the natural vegetation of the riparian zones. Moreover, the area saw increased erosion in tributary rivers, a gradual rise in peak stream ow, new ood zones (because of slower water velocity) and siltation of the main channel (Cunha, 1995(Cunha, , 1991  For this end, hilltop protected areas were characterized as comprising the upper two-thirds of hills with a minimum height of 100 meters and average slope greater than 25 degrees. Per the Forestry Code, hill baselines in wavy reliefs are de ned as the lowest point between two hills. Meanwhile, for the Inea Resolution, hill baselines are de ned by the nearby lowland or a nearby body of water. This difference in parameters for the de nition of hill baselines results in an 1500% difference in the total size of protected areas (~ 78.67 km²) distributed and connected along the watershed when using Inea Resolution guidelines compared to the parameters de ned in Forest Code, which leads to protected areas that are isolated and concentrated in water divisors totaling ~ 5.53 km 2 in area (Magdalena et al., 2018).
In total, we modeled seven scenarios using the de nitions for hilltop protected areas from the Forest Code and the Inea Resolution (Table 1) (Fig. 2C, Fig. 2F, Fig. 2D and Fig. 2G), two of which were extreme scenarios and one a control scenario. We created the control scenario ( Fig. 2A) based on the 2010 LU/LC maps for the state of Rio de Janeiro (Bastos and Napoleão, 2011), while, for the extreme case scenarios, one assumed the watershed is fully preserved (Fig. 2E) and the other assumed it has been fully deforested (Fig. 2B).

Model Calibration
The MHD-INPE model was calibrated for the daily stream ow data set offered by the Correntezas Nova gauge station for the period from 1984 to 2003 (Fig. 1). Calibration was conducted via the automatic method described in Duan et al. (1992), with Nash-Sutcliffe (Nash) as the e ciency coe cient applied to stream ow (Nash) and logarithmic stream ow (logNash) calculations. When Nash shows values ≤ 0, the model simulation does not result in better values than the mean observed stream ow numbers (Gupta et al., 2009). Simulations were validated using the method described in Moriasi et al. (2007), with Nash > 0.50 considered acceptable.

Framework of numerical experiments
We developed numerical experiments applying different LU/LC scenarios ( Table 2). In addition to the control experiment, which simulates the actual watershed status, we developed two extreme case experiments for comparison purposes.

Analysis of the modeled hydrological responses
The MHD-INPE is a deterministic distributed hydrological model (Mohor et al., 2015) that subdivides the watershed into a regular grid of cells that can consist of one or more types of Hydrological Response Unit (HRU), understood here to be a homogeneous unit that yields the same response to the same input. The HRU is represented by parameters that determine water balance and the ow generation process resulting from the interaction between LU/LC and different soil types.
We estimated evapotranspiration from the Penman-Monteith equation (Allen et al., 1998). Water loss by canopy interception was estimated with the model proposed by Gash et al. (1995). Water transpiration captured by plant roots was estimated per the model in Jarvis (1989). Water routing between cells was calculated per the Muskingum-Cunge method (TUCCI and CLARKE, 1997). And nally, the third segment includes ows with exceedance probabilities of over 0.7, classi ed as the low-ow segment volume of the FDC. In addition to these, a seasonality index was also applied that described inter-annual average stream ow variations (Table 3).    However, the watershed shows silting tendencies because of the straightening process and lack of maintenance along the main river course. This, combined with the watershed's relief, results in over ow over the river's banks, which changes the stream ow dynamics.
Nevertheless, the hydrological model performed in a similar pattern to that of the observed data in the dry season (Fig. 3).

Hydrological responses of the São João river watershed
As expected by the water balance, in long-term average values, evaporation reduction (Fig. 4B) suggests an increase in stream ow generation (Fig. 4A). The hydrological responses simulated with the LU/LC scenarios wherein only the hilltop protected areas are preserved per the Forest Code (3CF) or the Inea Resolution (3CI) show similar variation to the extreme scenario of the completely deforested watershed. In contrast, given that most areas de ned as protected areas are currently forested, the hydrological response scenarios for Control, 2CF and 2RI are close to the response found when the watershed is fully forested. In these scenarios, despite the lower evapotranspiration at the dry season and the beginning of the rainy season, stream ow is still lower than in the Forest scenario because of the lower soil hydraulic conductivity capacity inputted in the model.
An analysis of long-term monthly mean amounts reveals a rise in the stream ow regime and a reduction in evapotranspiration processes in the seasonal variation, which shows the effects on maintaining forest cover only on the hilltop protected areas (Fig. 4A and Fig. 4B, 3CF and 3RI). In all other scenarios, among which the Forest scenario, mean hydro-climatic variables show behavior closer to that of the watershed's present state ( Fig. 4A and Fig. 4B, Control).
The FDC descriptors, when applied to the numerical modelling results considering the Pasture and Forest scenarios, show two extreme hydrological response patterns (Fig. 5). Dry and rainy season stream ow rates are higher in the Pasture scenario ( Fig. 5A and 5C), with a greater slope in the middle section of the FDC (Fig, 5B), suggesting loss of regularity with rapid responses to rainfall events. Seasonal variation stream ow rates are also higher in the Pasture scenario (Fig. 5D).
The Control scenario shows a pattern close to that of the Forest scenario. The MWL index, associated with dry season stream ow rates, is close to that of the Forest scenario (Fig. 5A), while the MWH index, associated with the rainy season, is higher (Fig. 5C). The slope of the middle section of the FDC, shown by the MS index, is greater in the Control scenario than in the Forest scenario (Fig. 5B), revealing a quicker, less regular hydrological response; also, seasonal variation is greater in the Control scenario (Fig. 5D).
The addition of forest to current LU/LC in the hilltop protected areas shown in the 2CF and 2RI scenarios resulted in a decline in dry and rainy season stream ow rates ( Fig. 5A and 5C), along with changes in soil hydraulic properties and better hydrological response in the watershed (Fig. 5B and 5D). On the other hand, when the model assumes that only the hilltop protected areas have native forest (scenarios 3CF and 3RI), the hydrological response of the watershed changes to a pattern similar to that of the Pasture scenario. In these scenarios, dry and rainy season volumes rise ( Fig. 5A and 5C), while hydrological response is more rapid and less regulated ( Fig. 5B and 5D).
Stream ow rate rises in the watershed under the native vegetation loss scenarios are associated with a direct response increase (Fig. 6A). As such, the São João river watershed shows direct response increases under scenarios 3CF and 3RI (Fig. 6A). Conversely, native vegetation loss is associated with a decline in response rates for base ow, which is the phenomenon that feeds channel runoff at dry season (Fig. 6B).
This variation in ow paths is associated with variations in saturated zones, where direct runoff into the watershed occurs (Fig. 6C).
The replacement of Atlantic Rainforest with pasture results in a decline in evapotranspiration (Fig. 4B), which, associated with a decline in soil water retention capacity, increases saturated zones. In the pasture scenarios (Pasture, 3CF, 3RI), the watershed keeps a higher saturated zone rate than in the Forest, Control, 2CF and 2RI scenarios.
Saturated zones had seasonal variation in size, covering a larger area of the watershed during the rainy season (January, February and March) than in the dry season (July, August and September) (Fig. 7). Under the Control scenario (Fig. 7A), saturated zones are clustered around the main channel where the oodplain is located.
In the Pasture scenario, saturated zones are the largest at both seasons, including regions with high elevation in the rainy season ( Fig. 7B). On the other extreme, i.e., the Forest scenario, saturated zones are smaller and show fewer variations between seasons (Fig. 7E). The biggest differences between the Forest and Control scenarios (Fig. 7A) are in the size of the oodplain occupied by pastures ( Fig. 2A).
In a LU/LC scenario wherein the Atlantic Rainforest is restricted only to the hilltop protected areas (Fig. 7D and Fig. 7G), saturated zone size is close that of the Pasture scenario in the watershed (Fig. 7B).
In the scenarios that keep LU/LC as currently existing in most of the watershed (2CF, 2RI) ( Fig. 7C and Fig. 7F), saturated zones are closer in size to those of the Forest scenario (Fig. 7B). This is because of the small area of hilltop classi ed as protected area under current legislation, which protects a smaller rate of natural vegetation than what is found in the current scenario (Magdalena et al, 2018). Nevertheless, we noted that the presence of forests on hilltops reduces saturated zones in the watershed (Fig. 7C, Fig. 7F,   Fig. 7G).
These responses suggest that, although liquid discharges increase in volume, reducing native vegetation coverage to the minimum prescribed by law would affects water supply in the watershed, which can cause problems in water resource management in the Juturnaíba reservoir and intensify con icts over water use. Besides, the greater stream ow rates in the rainy season suggest a rise in energy ows for sediment displacement, which may cause soil loss, silting of the main channel and deposit of sediments in the Juturnaíba reservoir (Chen et al., 2007;Cunha, 1995;Fohrer et al., 2001).

Responses to sediment deposit controls in the São João River watershed
Per our analysis of the spatial distribution of average annual soil loss in the watershed under the Control scenario, 76% of the catchment area shows values lower than 5 t ha − 1 year − 1 (Fig. 8A). Nevertheless, 14% of the catchment area has an average loss of 5-30 t ha − 1 year − 1 (slight and moderate). This soil loss is in lowland areas, a region currently covered with pasture areas because of the removal of oodplain vegetation and the riparian forest for the channeling of the São João river, as well as due to agricultural activity (Cunha, 1995(Cunha, , 1991. Areas showing severe and very severe soil loss amount to only 2% and 3% of the catchment area ( Fig. 8A).
As expected, under the Pasture scenario, we see a decline in the share of areas with soil loss at less than 5 t ha − 1 year − 1 and an increase in areas with slight (5-15 t ha − 1 year − 1 ) and moderate (15-30 t ha − 1 year − 1 ) soil loss, which represent 34% and 21% of the catchment area. The Pasture scenario also shows 13% of the catchment area with soil loss greater than 50 t ha − 1 year − 1 (very severe), concentrated in the water divisor regions (Fig. 8B), while, under the Forest scenario, this soil loss is less than 5 t ha − 1 year − 1 (Fig. 8E).
For the scenarios that consider only hilltop protected areas will retain vegetation coverage (3CF and 3RI), soil loss is close to that of the Pasture scenario; when compared to the Control scenario, forest decline results in intensi ed soil loss, especially in the watershed water divisors (Fig. 8D and Fig. 8G). Greater soil loss can intensify drainage system processes over time -an issue further compounded by increased surface runoff (Fig. 6A) in high-slope areas -and reduce useful storage volume in the Juturnaíba reservoir.
In the 2CF and 2RI scenarios, which are characterized by minimum vegetable coverage reduction or a small increase thereof when compared to the Control scenario, the spatial distribution of soil loss is close to that of the Control scenario (Fig. 8A, Fig. 8C and Fig. 8F). The differences between the scenarios that take into account Federal and State legislation (3CF and 3RI) are expressed in the number of hilltop protected area fragments when considering the Inea Resolution criteria, which considers the foot of the hill or the nearest body of water as the criteria for delimitation.

Discussion
Changes in hydrological responses caused by forest cover loss intensify sediment transport in the watershed. subsequently affecting the water supply system. The results obtained with the Control scenario, wherein forested area remains larger than that required by law in association with agricultural uses in the plains, suggest that preserving a greater forested area in the watershed will lead to better regulated water ow in the channel and reduce soil loss, with consequent declines in silting and sediment deposition, thus facilitating better water resource retention for sustainable management in the watershed.

Conclusions
The results obtained in this manuscript indicate that restricting the coverage and dynamics of native vegetation only to hilltops does not prevent drastic changes in hydrological response in the watershed. Forest loss in the watershed may cause increases in surface runoff and a decline in base ow, as well as soil loss. These changes can intensify sediment load in the São João river, causing its silting as well as sediment deposition in the reservoir used to protect water supply. Due to their small size, the hilltop protected areas