The study was conducted at Copo National Park (CNP) which has an area of 1,181 km², and is located in Copo Department, northwest of the Province of Santiago del Estero, Argentina (Parques Nacionales, 2022) (Figure 1 A, B). According to the National Weather Service (2021), the rainy season starts in October (44.7 mm) and finishes in April (35.9 mm); the accumulated annual rainfall reaches 629 mm and its maximum values occur in January (134.8 mm). The average monthly temperature in January is 26.7 °C (summer) and 12.7 ° in July (winter).

Chaco Semi-arid Forest in CNP is xerophytic and open, it is considered the climax vegetation of the region (SAyDS, 2007). The study site represents the best preserved and most extensive tract of Chaco Forest in Argentina (Tálamo and Caziani, 2003). The landscape is flat (nearly horizontal) and homogeneous, with no surface water bodies, and is vegetated by thorny, semi-deciduous forests dominated by Schinopsis quebracho-colorado, Aspidosperma quebracho-blanco and Ziziphus mistol, which reach heights of 18m to 20 m. The dense stratum of 1-6m tall shrubs is dominated by Capparis retusa, Acacia praecox, Celtis pallida, Achatocarpus praecox and Schinus polygamus (Tálamo and Caziani, 2003; Talamo et al., 2012), among others. Broadleaves and herbaceous predominate in the lower strata (SAyDS, 2007).

FIGURE 1
A - North of Argentina, Province of Santiago del Estero (grey) and Copo National Park. B - Eight sampling points of Chaco Semi-arid Forest in Copo National Park (red polygon).
Sources A: Instituto Geográfico Nacional (2022) and B: Earth Observing System Project (2022) Sentinel-2 L2A.

The time series of the Normalized Difference Vegetation Index (NDVI) (Rouse et al., 1973), it was derived from the Temporary Vegetation Analysis System (available on the http://www.satveg.cnptia.embrapa.br/satveg/). The system provides NDVI time series profiles (derived from MOD13Q1 (satellite Terra) and MYD13Q1 (satellite Aqua) images belonging to the LP-DAAC/EOS-NASA project, with Google Maps interface. MODIS NDVI, produced on 16-day intervals and at spatial resolution 250m x 250m, derived from atmospherically-corrected reflectance in the red, near-infrared, and blue wavebands. The VI's use a MODIS-specific compositing method based on product quality assurance metrics to remove low quality pixels.

Eight time series of original NDVI_MODIS were derived, one for each sampling point, and each time series ranged from 12.08.2000 to 26.06.2022 consisting in total 970 NDVI values for each sampling point (Figure 1 B and Figure 2).

The number of the forest samples (n = 8 pixels) was determined by the Fitzpatrick-Lins methodology (1981). Sampling points were located by simple random sampling with Idrisi TerrSet 2020sample module (Figure 1 B). Covers that did not correspond to FCHS were masked with binary masks monitored with high-resolution images available on Earth Observing System landviewer (2022).

Each sampling point is a NDVI_MODIS covering 5.4 pure hectares of CNP Chaco Forest throughout the period analyzed. The Lat/Long coordinates of the sampling points are: M1 (Lat -25.757; Long -61,967); M2 (Lat -25,718; Long -61,926); M3 (Lat -25,706; Long -61,833); M4 (Lat 25,937; Long -61,844); M5 (Lat -25,678; Long -61,752); M6 (Lat -25,811; Long -61,951); M7 (Lat -25,771; Long -61,854) and M8 (Lat -25,825; Long -61,862) (Figure 1 B).

The FCHS growing seasons (GS) were delimited by transforming the original NDVI time series into NDVI_RATIO time series (Figure 2); the method it was developed by White et al. (1997) and was performed using equation 1.

(1)

The NDVI_RATIO oscillates in the range of zero to one. In equation 1, NDVI is the 16 day NDVI_MODIS, NDVI_MAX is the maximum 16 day NDVI_MODIS and NDVI_MIN is the minimum 16 day NDVI_MODIS. The beginning and end of each biannual growing season were determined by the NDVI_RATIO 0.5 threshold (White et al., 1997) (Figure 2). According to Myneni and Williams (1994), the NDVI_RATIO 0.50 threshold minimizes background effects on fAPAR/NDVI_RATIO.

NDVI_RATIO 0.50 threshold enabled the definition of the FCHS growing season (GS) of the entire period by removing values less than 0.5 NDVI from the time series. They corresponded to the NDVI_RATIO of the months of June, July, August and September (dry season). NDVI_RATIO of each growing season and sample point was integrated to generate Seasonally Integrated NDVI_RATIO (SI-NDVI_RATIO) (Equation 3) (Zhang et al., 2018; Zhang et al., 2020).

FIGURE 2
NDVI_RATIO time series of Chaco Semi-arid Forest, Copo National Park of the eight sampling points (M1 to M8).

Aboveground Net Primary Productivity (ANPP) of the FCHS was determined by the radiation use efficiency model proposed by Monteith (1972). The model ascertains that the ANPP of a plant cover is proportional to the incident photosynthetically active radiation (PAR), that is, to the fraction of that radiation intercepted by green tissues (fAPAR), and to the conversion efficiency (ε). According to Fensholt (2004), due to the strong relationship between NDVI and fAPAR, the Monteith model can be written as follows (Equation 2):

(2)

In equation 2, the efficient use of the energy (ε) was determined by Ruimy et al. (1994) for dry forest ε = 0.50 g DM. Mj^-1 m². SI-NDVI_RATIO is the NDVI_RATIO growing season integrated, defined by the area under the curve delimited by start and end of the season, was determined mean (equation 3) (Zhang et al., 2018; Zhang et al., 2020).

(3)

The fraction of photosynthetically active radiation (fAPAR) absorbed by the Semi-arid Chaco Forest was determined by equation 4 (Seaquist et al., 2003). The parameters α = 0.504 and β = 0.01 are the constants that arise from the linear relationship between NDVI_RATIO Maximum (0.996) and NDVI_RATIO Minimum (0.50) of the 2000-2022 time series (Myneni & Williams 1994, Fensholt 2004, Seaquist et al., 2003).

(4)

The average global seasonal incident radiation in the area under study (average of 22 years), was derived from the NASA Prediction of Worldwide Energy Resourse Project (NASA, 2021). Incident photosynthetically active radiation (PAR) in foliage is considered to be a constant fraction of 48% of the incident radiation at the atmosphere limit (Seaquist et al., 2003, Fensholt, 2004). The ANPP of FCHS was quantified and expressed in kg DM. ha^-1, for each GS and sampling point, starting from the seasonally integrated productivity (SI-ANPP) of the seasonally integrated NDVI_RATIO (SI-NDVI_RATIO) were quantified.

Rainfall data for the period 2000-2022 were derived from the Weather Station of Sociedad Rural del Noreste Santiagueño (SRNS) (coordinates lat -27.65; long -62.40). Rainfall produced in the months corresponding to the GS (October to May) were integrated to generate the integrated seasonal rainfall (SR) for each of the 22 growing seasons (Figure 2).

The anomalies of the seasonal rainfall were determined as follows (equation 5 and figure 3):

(5)

where X_i is the integrated seasonal rainfall (SR) (2000-2022) and X_h is the historical integrated seasonal rainfall (1918-2022).

FIGURE 3
Anomalies of seasonal rainfall corresponding to period 2000-2022
Source: Rural Society of Northeastern Santiago (com. pers., 2022).

Rain use efficiency (RUE) could be a critical indicator for evaluating the response of primary productivity to variability of rainfall in arid and semi-arid ecosystems (Le Houérou et al., 1988; Sala et al., 1988). The RUE was determined by equation 6 (Bai et al., 2008; Zhang et al., 2020):

(6)

where RUE is the rainfall use efficiency (kg DM. ha⁻¹ mm⁻¹), SI-ANPP is the seasonally integrated aboveground net primary productivity (kg DM. ha^-1 season⁻¹), and RS is the seasonal integrated rainfall (mm).

The SI-ANPP of FCHS (kg DM. ha^-1 ·season−1) (expectations) were compared, using the t-test for independent samples (Di Rienzo et al., 2022). The classification variables (distributions) were two cluster: One with seasonal negative rainfall anomalies (AN) (Figure 3), which including: 2000-2001, 2001-2002, 2003-2004, 2005-2006, 2007-2008, 2008-2009, 2012-2013, 2014-2015, 2017-2018, 2021-2022, and another with seasonal positive rainfall anomalies (AP) (Figure 3) which including: 2002-2003, 2004-2005, 2006-2007, 2009-2010, 2010-2011, 2011-2012, 2013-2014, 2015-2016, 2016-2017, 2018-2019, 2019-2020, 2020-2021.

The RUE (kg·DM. ha⁻¹·mm⁻¹) (dependent variable y) by means of the linear or nonlinear regression method (α = 0.05) were related to integrated season rainfall (SR) (independent variable x). The models were based on two goodness-of-fit tests: 1) the best adjustment of the adjusted coefficient of determination (R²) and 2) by the lack of fit test which is an estimation regardless s² of the model or pure error (Di Rienzo et al., 2022).

Analysis trend of SI-ANPP was performed using a linear best-fit regression line fitted to SI-ANPP of growing season of the period 2000-2022. The slope of the resulting best-fit line was tested using a t-test for a parameter to see if it was significantly different (p < 0.01) from a zero slope (Di Rienzo et al., 2022).

The analyses were performed with the Infostat package (Di Rienzo et al., 2022). Only the relationships found statistically significant (p < 0.05) will be displayed.

The geographic information system used was Idrisi TerrSet 2020.

The NDVI_MODIS time series and the 0.5 NDVI_RATIO thresholds make it possible to delimit the start and end of the growing stations for the FCHS with high efficiency. The growing stations comprised the NDVI_RATIO integrated of October, November, December, January, February, March, April and May. According to Gao et al. (2020), NDVI time series analysis that considers seasonality has been tested with both low and medium spatial resolution images and is capable of detecting forest disturbances in historical time series. However, while Gao et al. (2020) report that remote sensing thresholds are currently still a challenge, the NDVI_RATIO 0.5 threshold used in this study reduces uncertainty in delimiting growing seasons, and thus increases accuracy in quantifying seasonal ANPP of Chaco Semi-arid Forest at local and regional level. According to White et al. (1997) the transformation is attractive because it is consistent. Thus, a single NDVI_RATIO threshold may be used, obviating the need to establish absolute NDVI thresholds or landcover-specific thresholds.

The results of this study are promising since the average SI-ANPP obtained for the entire period (twenty tow season) was 43741.49 kg DM. ha^-1, (a minimum of 14294.88 kg DM. ha^-1 and a maximum of 64938.18 kg DM. ha^-1) and it is similar to the average obtained by Baccini et al. (2012). They determined an ANPP mean of Chaco forests of 49200 kg DM. ha^-1 which varies between 28600 and 69800 kg DM. ha^-1.

Significant differences were found between SI-ANPP of growing season with seasonal negative rainfall anomalies negative and seasonal positive rainfall anomalies (T = -3.49; p = 0.0006) (Table 1). The growing seasons with seasonal negative rainfall anomalies had an SI-ANPP mean of 41390.25 kg DM. ha^-1, a minimum of 14294.88 kg DM. ha^-1 and a maximum of 62563.00 kg DM. ha^-1. In this sense, Phillips et al. (2009) used records from multiple long-term monitoring plots across Amazonia to assess forest responses to the intense 2005 drought. The authors determined forest lost biomass due to unusually intense dry season, and conclude that Amazon forests appear vulnerable to increasing moisture stress.

The growing seasons with seasonal positive rainfall anomalies had an SI-ANPP mean of 46092.74 kg DM. ha^-1, minimum of 25684.06 kg DM. ha^-1 and maximum of 64938.18 kg DM. ha-1. SI-ANPP mean values of the FCHS determined (Table 1) are higher than the ANPP values determined by Baldassini (2018) and Baldassini & Paruelo (2020), in the FCHS in Anta, Salta, Argentina, in a precipitation gradient, similar to that of the study area, which oscillates between 600 and 800 mm yr^-1. The authors determined, respectively, that the average annual ANPP ranges from 10875 kg DM. ha^-1 to 11414 kg DM. ha^-1, and 11063 kg DM. ha-1 yr^-1. According to Baldassini (2018) and Baldassini & Paruelo (2020), the FCHS in Anta, Salta, evidences processes of forest degradation due to unsustainable anthropic activities, specifically selective logging and overgrazing.

TABLE 1
T-test for sample independents of SI-ANPP (kg DM. ha^-1) of growing season of Chaco Semi-arid Forest corresponding to the period 2000-2022

Cluster A = seasonal negative rainfall anomalies, Cluster B = seasonal positive rainfall anomalies, T = T-test statistic and p-valor = probability.

No significant relationships were found between la SI-ANPP and seasonal integrated rainfall (N = 22; R² = 0.09; p> 0.05).

Significant and nonlinear relationships were found between rain use efficiency of forest systems and seasonal integrated rainfall (Figure 4): N = 22; R² = 0.73; p = 0.0001. The model did not show a lack of fit. Seasonal integrated rainfall accounted for 73% of the variation of the RUE in FCHS CNP (Figure 4). According to Gamoun (2016), the RUE is also highly dependent on soil and vegetation type and environmental conditions and, therefore, may not necessarily respond linearly with rainfall. Furthermore, Hsu et al. (2012) state that, in many cases, ANPP during relatively wet years determines the nonlinearity of precipitation-ANPP relationship and the frequency of these “outlier” years is increasing with climate change.

FIGURE 4
Rain use efficiency of growing season of FCHS and seasonal integrated rainfall relationship corresponding to period 2000-2022.

The RUE means of the twenty tow growing season of FCHS period 2000-2022 was 68 kg DM. ha^-1 mm^-1, being the maximum 124 kg DM. ha^-1 mm^-1 and the minimum 39 kg DM. ha^-1 mm^-1 (Figure 4).

The three growing season of forest system with respective maximum RUE (124, 108 and 98) had rainfall of the 400 mm (Figure 4, dash line). In this sense, Huxman et al. (2004) state that, at sites with the lowest annual average rainfall, high efficiency of water use associated with individual growth rate plant, is translated to water use high efficiency at the ecosystem level. Furthermore, Huxman et al. (2004), consider that there is convergence to a common maximum RUE that is typical of arid in years when water is most limiting, deserts, grasslands and forests all exhibit the same rate of biomass production per unit rainfall, despite differences in physiognomy and site-level RUE ecosystems.

The RUE of twelve growing season of FCHS oscillates between 39 kg DM. ha^-1 mm^-1 and 93 kg DM. ha^-1 mm^-1 all with rainfall ranging from 500 mm to 700 mm (normal rainfall for the study area) (Figure 4, solid line). Lower effectiveness of rain in growing season (RUE below average) with seasonal negative rainfall anomalies is consistent with higher evaporation losses (Noy-Meir, 1973). Increase in direct evaporation reduces ecosystem water use efficiency (lower PPN with the same precipitation) (Huxman et al. 2004; Newman et al. 2006). According to Noy Meir (1973) very common are drought-evading behavioral adaptations in which the plant uses stomatal control on productivity and water input so that both activities occur mostly (or exclusively) in season days, or hours of low transpiration.

In this sense, Reichstein et al. (2007) inferred a slight decrease of RUE during severe drought in European forests. Both authors, using remote sensing data from MODIS and AVHRR, indicate a strong negative anomaly of the fraction of absorbed photosynthetically active radiation in summer 2003, at more than five standard deviations of the previous years. Similar results were obtained by Zhang et al. (2014) who investigated the temporal patterns of NPP (MOD17) and RUE and their key driving climate factors, during the early 21st century drought for four biomes in China: Needleleaf forest, Broadleaf forest, Woody savannas, and Grassland. The authors supported that the RUE responded strongly to variations in current and previous year drought for all the four biomes.

The growing seasons of FCHS that had seasonal rainfall >700 mm, all with seasonal positive rainfall anomalies, are the ones that had the lowest RUE, which ranged between 40 kg DM. ha^-1 mm^-1 and 55 kg DM. ha^-1 mm^-1 (Figure 4). The widely accepted view is that RUE should increase with annual rainfall until other environmental factors limit ANPP (Le Houérou, 1984). They are major determinants of soil water availability; consequently, large amounts of rainfall, according to Noy-Meir (1973) ''ineffective precipitation'', are not utilized for plant growth, and have important effects on the site-level RUE (Noy-Meir, 1973; Le Houérou, 1984; Sala et al., 1988). Moreover, increases in annual rainfall amounts usually reduce RUE due to runoff and drainage (Bai et al., 2008; Vermeire et al., 2009; Dardel et al., 2014, Zhang et al., 2014; Gamoun, 2016; Zhang et al., 2018; Zhang et al., 2020; Li et al., 2020). In this sense, Paruelo et al. (1999), hypothesized that vegetation constraints decrease and biogeochemical constraints increase along a mean annual rainfall gradient. Thus, at each site, a combination of vegetation and biogeochemical constraints define a boundary within which ANPP varies in response to changes in water availability.

The results are consistent with the loss of large amounts of water by surface runoff when surpassing the threshold (boundary) of seasonal rainfall >700 mm and it would be related to geomorphology and climatic forces. Copo National Park is completely included in the sub-region of the paleo-causes or paleo-spillage of the Salado-Juramento rivers (Morello and Adamoli, 1974), an area that was shaped by ancient fluvial processes, where the forests grow on inter-fluvial homogeneous soil that occupies above 90% of the total area (Tálamo and Caziani, 2003, Talamo et al., 2012). It is a large silt-sandy plain, crossed by inactive dry channels filled with a finer sedimentation of sand. The soil profile is slightly concave with underground-type drainage with predominantly laminar runoff (Talamo et al., 2012). In rainy seasons, the water flows from high areas to lower areas or depressions; moreover, the presence of alluvial mini-fans (10-20 m) contributes to the formation of this drainage model (Caziani et al., 2003; Talamo et al., 2012).

In relation to climatic forces, and according to Noy-Meir (1973), Huxman et al. (2004), Vermeire et al. (2009), Hsu et al. (2012), Robinson et al. (2013), and Zhang et al. (2018), shifts in seasonal rainfall in intensity/magnitude, timing, frequency, and spatial and temporal variation of rainfall (persistent and random at all scales), which may independently affect communities, it will make it more important than ever to monitor how abiotic and biotic factors affect the transition from rainfall to plant-available soil moisture. In this sense, Figure 3 shows extreme seasonal positive anomalies of 500 mm and 700 mm. Further, in the period analyzed, average rainfall was recorded in the wettest quarter of great intensity/magnitude: October 63 mm (maximum 215 mm), November 92 mm (maximum 240 mm), December 127 mm (maximum 277 mm), January 130 mm (maximum 341 mm), February 99 mm (maximum 348 mm), March 93 mm (maximum 245 mm), April 69 mm (maximum 248 mm) and May 33 mm (maximum 290 mm).

The need for information on forest ANPP goes beyond that of REDD+, with countries wanting effective strategies to monitor the state of their forest resources (FAO, 2020). In this sense, the SI-ANPP of FCHS CNP showed a large positive trend in the period 2000-2022 (slope = 462.43; T = 25.64 p <0.0001) (Figure 5).

The stability of SI-ANPP of FCHS CNP throughout the period is due to two factors: the first and most important is the fundamental creation of legally established protected areas, such as national parks, for the preservation of the forest system from unsustainable anthropic activities (deforestation and degradation due to illegal extraction of high-value species), and the second and as a natural consequence of the first, the predominance of woody species of deep roots (Reichstein & Carvalhais, 2019) and the strategic displacements of its lower strata phenophases (broadleaves and herbaceous) in drought times (Noy-Meir, 1973; Reichstein & Carvalhais, 2019, Gavinet et al., 2019). Le Houérou (1984) states that the high level of synchronization between biomass productivity and the positive and negative trends of rainfall are due to the wide adaptation of plants to water-limited environments.

The stabilizing effect of species diversity results from the fact that different coexisting species are complementary in their responses to environmental fluctuations. Natural ecosystems display a range of adaptive responses to environmental changes differential reproduction of genotypes, and changes in species abundances (Lekevičius & Loreau, 2012), including differential gene activity and phenotypic plasticity.

FIGURE 5
Seasonally Integrated Aboveground Net Primary Productivity (SI-ANPP) of Chaco Semi-arid Forest in Copo National Park and trend line corresponding to period 2000-2022.

Improved knowledge on the response to the seasonal variability of rainfall of Chaco Semi-arid Forest in Copo National Park is thus crucial to better interpret the consequences of climate predictions of an altered global hydrological cycle and to implement appropriate adaptation measures to climate change in arid and semiarid lands similar to the study area.

tiedemannjl@gmail.com