Earthworms, Soil Porosity, and Infiltration Rates in Pine Plantation Forests in Java, Indonesia

Abstract

Pine plantations on volcanic slopes in Indonesia are considered to be forests and are managed for wood production and slope protection. Logging practices followed by replanting may affect soil health. Existing agroforestry management contracts allow farmers to intercrop with vegetables in young plantations and grow fodder grasses in older ones. However, critical data on hydrological functions in such systems are scarce, while concerns over heavy rainfall and floods increase. We explored the relationships between soil cover, soil carbon, earthworms, soil porosity and infiltration rates in relation to slope class in second-rotation pine plantations around two years of age (intercropped) and at ten-year old pine-grass stages. Five slope classes (0%–8%, 8%–15%, 15%–25%, 25%–45%, and >45%) were compared with three measurement points each. Basic soil chemical and physical characteristics were measured for the 0–10, 10–20 and 20–30 cm layers. Remnant natural forest was available as a historical reference only on the steepest slope class. Organic soil carbon (C_Org) divided by a texture-based reference level was 1.12, 0.32 and 0.49 for natural forest, young and old agroforestry on very steep slopes, respectively. Within pine-based agroforestry relative decline with slope class (1–5) was pronounced in earthworms (biomass −3.46, population −4.18) and infiltration rates (−2.35) while bulk density increased (0.49); for soil carbon (C_Org), nitrogen, available phosphorus and exchangeable Mg effects in the −1.26 to −1.68 range indicated a loss of functional topsoil. Differences with age of the agroforestry systems were much smaller but included a decreasing earthworm population but an increase in mean earthworm weight and partial recovery of the C_Org/C_Ref ratio. Pine-based agroforestry on very steep soils had only 10%–14% of the earthworm biomass and 35% of the infiltration rate of reference natural forest. Understory vegetation biomass and litter layer necromass were more than five-fold higher in the natural forest. Across all samples a higher C_Org and higher earthworm biomass were associated with complementary positive changes in infiltration rates and soil porosity. Regression analysis suggests equal skill of tree cover, soil C_Org, porosity, aggregate stability and earthworms to predict infiltration rates while explanatory variables were strongly correlated. Management of the pine plantations may have to achieve a closer approximation of the conditions in natural forests to effectively protect upper watersheds.

1. Introduction

Two decades after a major challenge to the prevailing forest-flood debate [1,2,3,4], public debate after every flood event in a country like Indonesia still tends to focus on the fate and condition of forests [5]. The common but overgeneralized expectation that all types of forests take care of all hydrological functions ignores the wide variation in above- and below-ground conditions in what can be called ‘forest’ [6]. Hydrological functions of tropical forests depend on the soil on which the remaining forests grow and on the impact management practices have on soil health [7]. Although research has shown that terraced fields used for rice cultivation can provide downstream flood prevention at a level comparable with existing forests on Java [8], infiltration rates at the field scale remain the single most distinguishing parameter between land cover types. However, given the strong dependence of infiltration rates on soil texture and slope, quantification of effects of land use requires statistical controls before differences in measured values can be attributed to land use, rather than to pre-existing variation [9,10]. Specific questions remain on the impacts of conversion of natural forests to plantation crops such as oil palm [11,12], rubber [13] or industrial timber [14]. Continued empirical studies have shown difficulties in dealing with the various confounding factors in more aggregated scales [15].

Deforestation and forest degradation are politically and emotionally loaded terms that refer to changes across and within the broad categories of a forest-nonforest dichotomy [16], interacting with a range of watershed functions, tradeoffs and associated metrics [17,18]. Tree planting is still in public discussions, expected to be a universal remedy to any lack of watershed functions, but evidence is mixed [19]. Reduced soil health in plantation forests compared to natural forests justifies the use of the term ‘forest degradation’ [14], despite increased wood production. As landscapes are shaped by the specific history of land uses across topography and soils, functionally relevant soil properties vary in space and time, but their association with land cover may be based on a two-way causality: land use impacts on key soil ecological processes can be hard to distinguish (in survey-type studies) from soil properties that (socially) determine land use patterns. Experiments with truly randomized allocations of land use systems are scarce. Where the deepest and most productive soils are usually prioritized for agriculture, forests remain in less fertile locations but may yet be the preferred land cover on sloping land [20]. Survey-type empirical data address the specific context in which they were collected, making overgeneralization of forest-water relations a risk. Legal categories and designations may be too coarse for the nuanced ecological relationships involved.

In Indonesia’s forestry law, designated forests (‘kawasan hutan’) are categorized by primary function as production, watershed protection or biodiversity conservation forests. In the production forest category, diverse natural forests can be replaced by monocultural plantations, despite the impacts this conversion has on aboveground biodiversity and soil functions [4]. However, production forests still have to maintain minimum levels of environmental functionality, such as infiltration of rainfall, avoidance of surface runoff and erosion. Hydrological properties of planted forests in a first or subsequent production cycle depend strongly on the practices used in clear-felling and replanting, with possible recovery during the next cycle. Monocultural plantations of Pinus merkusii Junghuhn et de Vriese for 30–40 year rotations occupy around 0.16 M ha on Java, with about half the area used for resin tapping to provide income to surrounding villages [21]. Given the complex history of plantation forests on the volcanic slopes of densely populated Java (Indonesia) [22], some of them have been terraced before trees were planted, others not. While Indonesia has many examples of farmer-managed forest analogs known as ‘agroforests’ that structurally and functionally resemble natural forests [23], other forms of ‘agroforestry’ have monocultural plantations as stating point, where farmers are allowed to manage agricultural crops (or fodder grasses) as long as these do not interfere with the trees managed by a forest authority [24]. In mosaics of remnant forests, agroforestry, open field agriculture and settlements, agricultural intensification reduces both aboveground biomass and indicators of soil quality [25]. Managing the hydrological impacts of these systems depends on the way structure and function are understood, rather than reliance on a simple forest/nonforest dichotomy.

Tall trees in monocultural tree stands without a protective litter layer were found to face serious erosion risks as water drips falling from the canopy had more energy than the rainfall [26,27]. The brown leaves decaying on the forest floor as a litter layer may be as important for ‘forest functions’ as the green leaves in the canopy [28]. The standing litter (necromass) layer can be estimated from annual input and mean residence time, with the latter depending on nitrogen content, lignin and polyphenolics [29]. Litter layers interact with earthworms, providing both food and habitat (shelter) for these ‘ecosystem engineers’ [30]. Functionally, different earthworm species and ecological groups may interact differently with soil conditions, calling for taxonomic identification in empirical studies [31,32,33]. Compared to natural forests, earthworms in agroforestry systems may involve shifts in species composition, population densities, sizes, pore formation and soil mixing activities [34,35]. Soil organisms such as earthworms can both reduce soil loss by improving porosity and increase it by diminishing soil stability as a result of their mixing activities [36]. Where the quantitative relationship between point-level infiltration measurements and subcatchment level stormflow depends on spatial heterogeneity within a field [37], changes in mean infiltration rates can be interpreted qualitatively for different slope classes, where slope decreases the time available for infiltration before runoff occurs. ‘Infiltration-friendly’ forms of agroforestry were identified on the basis of small runoff plots, as well as soil porosity measurements [38]. Elsewhere, grasslands were found to have large earthworm populations and high infiltration rates with an initial decline after conversion to agroforestry [39]. On volcanic soils, earthworm activity has been linked to coffee production in pine-based agroforestry [40], while episodes of volcanic ash deposition influence vegetation, litterfall, earthworm activity and the regeneration of soil porosity and infiltration rates [41]. As volcanic ash can have hydrophobic properties, infiltration rates can be affected [42].

As specific research questions for pine plantations in the production forest zone of volcanoes on Java, we targeted:

• How close can pine plantations with agroforestry management approximate a reference natural forest in terms of infiltration rates?
• How do slope class and plantation age under prevailing agroforestry management influence soil health indicators?
• Are earthworms active agents in restoring water infiltration that deserve to be promoted by specific interventions?

2. Materials and Methods

2.1. Study Area

The research was conducted from January 2023 to August 2023 in and around the Bromo-Semeru massif, which includes Java’s highest volcano, with a national park (National Park Bromo Tengger Semeru or TNBTS) core surrounded by ‘watershed protection’ and ‘production’ forests, including P. merkusii plantations (introduced from Sumatra to Java in the 1920s during the Dutch colonial era) managed by Perum Perhutani (State Forestry Public Corporation in Java and Madura Island). Sites were selected in the Poncokusumo District, Malang Regency (Figure 1), Latitude 8°01′–8°03′ S and Longitude 112°82′–112°87′ E. The average annual rainfall over the past 10 years measured at the study site is 2170 mm, with an average air temperature of 24 °C (Figure 2). Soil sampling and earthworm observations were conducted towards the end of the rainy season (April–May 2023).

The history of plantation management in large blocks constrained our choice of sampling sites for an age × slope interaction test. Possible confounding of age with inherent soil properties was tested by comparing basic soil properties. The soil in all locations is considered to be geologically young with limited horizon development and classified as Inceptisol according to Soil Taxonomy criteria [43,44], probably as Andic Eutrudepts [45]. It has layers with andic (volcanic ash soil) properties with a strong influence of volcanic amorphous material, so that its physical structure is crumbly loose, low soil bulk density (typically <1.0 g cm⁻³), and highly porous, which has implications for high water-holding capacity. With standard dispersion methods, the texture is mostly sand, through strong bonding of microaggregates. Soil chemistry of andisols is characterized by a relatively high organic C content (around 7%, [46]), high exchangeable cation concentrations and a high P retention (P fixation potential).

2.2. Research Design and Sampling Strategy

After a survey of the area and the forest management data, sites were purposely selected to match the survey design. A factorial design was used for the survey with land use type as first factor at three levels: (YAF) = Young agroforestry system (Pine-horticulture: age around 2 years); (OAF) = Old Agroforestry System (Pine-Grass: age around 10 years); (PFM) = Protection Forest Management, and within YAF and OAF slope class as second factor with five levels: 0%–8%, ‘Flat’; 8%–15%, ‘Sloping’; 15%–25%, ‘Rather Steep’; 25%–45%, ‘Steep’; >45%, ’Very Steep’. Given the land-use history, PFM was only observable in the very steep class at slope > 45%. Ten-year-old agroforestry was sampled in Wringinanom village, and two-year-old systems in Gubugklakah Village, both part of the Ngadas Forest Management Unit (Figure 3). The protected forest was measured inside the TNBTS in the Ngadas village. Within the constraints of what was available in the landscape, three replicate sampling points were selected for each factorial combination (

Table A1. Coordinates of field observation and soil sampling locations.

Land Use	Slope (%)	Replication
		1	2	3
		Latitude (S)–Longitude (E)
PFM	>40	8.012183–112.865643	8.012739–112.865112	8.012909–112.863519
YAF	0–8	8.024245–112.840852	8.029940–112.841878	8.029904–112.840970
	8–15	8.011655–112.856664	8.012572–112.855474	8.012827–112.854841
	15–25	8.025211–112.841339	8.025588–112.841682	8.025718–112.841810
	25–45	8.025050–112.841273	8.025309–112.841675	8.025586–112.841476
	>45	8.025791–112.841475	8.026906–112.841292	8.026790–112.839951
OAF	0–8	8.032908–112.827567	8.033420–112.827656	8.031129–112.827596
	8–15	8.030663–112.827390	8.030913–112.827711	8.031490–112.827411
	15–25	8.029752–112.826668	8.030903–112.827752	8.033359–112.827898
	25–45	8.030334–112.827043	8.030648–112.826099	8.030738–112.826619
	>45	8.032381–112.828402	8.032647–112.828225	8.032488–112.827479

). All sites appeared in the field to be on the same soil type, and intrinsic similarity was confirmed by soil sampling of the 0–30 cm layer and analysis in the laboratory.

At each designated land location, the determination of the observation plot began by randomly throwing a token to determine the starting point for a plot measuring 100 m to the north and 20 m to the east (Figure 4A; [47]). In this plot, diameters at 1.3 m above the ground of all trees with a diameter of more than 5 cm were recorded for calculation of tree population (ha⁻¹), total basal area (m² ha⁻¹), and basal area of the dominant tree species.

The large plot was subdivided into 20 m × 20 m subplots, with 4 quadrants each. Within each subplot, five points were established to assess tree canopy density, understory vegetation with less than 5 cm stem diameter and the litter layer (Section 2.3). Soil infiltration measurements were carried out at one point in the middle of each quadrant (Section 2.5), avoiding prior compaction. Next, earthworm populations (Section 2.6) were sampled, before soil samples were collected for further soil physical (Section 2.4) and soil chemical (Section 2.5) analysis of disturbed soil samples (five per subplot) were taken of the 0–30 cm layer and composited.

2.3. Aboveground Vegetation Characteristics

In a 20 m × 100 m plot (2000 m² = 0.2 ha), trees were divided into two sizes: medium trees with a DBH (diameter at breast height) of 5–30 cm and large trees with a DBH of >30 cm. Tree population (tree density) was calculated as the number of individual trees recorded divided by plot size (0.2 ha). Within each large plot, DBH measurements were taken on each tree at a height of 1.3 m above ground level. DBH was measured using a diameter tape to determine the total Basal Area (BA) for trees with DBH > 5 cm using the following equation:

$$\mathrm{B}{\mathrm{A}}_{\mathit{total}}\left({\displaystyle \frac{{\mathrm{m}}^2}{\mathrm{h}\mathrm{a}}}\right)={\displaystyle \frac{\sum \pi {\left(\frac{DB{H}_i}{2}\right)}^2}{0.2}}$$

(1)

Tree canopy measurements were performed for each land use type to determine canopy density based on differences in land use to illustrate canopy cover and canopy openness on the land. Canopy observations were conducted using the Canopy App (University of New Hampshire version 1.0.3 downloadable at https://apps.apple.com/us/app/canopyapp/id926943048, accessed on 24 April 2026) looking upwards in a horizontally held cell phone at 1 m above the soil surface (or above understory vegetation). Canopy observations were made in five quadrants of small-plot locations (20 m × 20 m) to represent variations in stand conditions. The results for the percentage canopy cover extracted from the Canopy App were averaged to produce plot-level values.

Litter thickness, litter necromass, and understory biomass measurements were conducted at five sample points within a 20 m × 20 m subplot located as part of the 100 m × 20 m main plot following [31]. At each observation point, a 50 × 50 cm double frame was placed (Figure 4B). Litter thickness above the mineral soil was measured after mildly compressing it at five points in each frame. For litter necromass, all litter in the quadrat was collected and weighed (wet weight). Then, a representative subsample was taken, dried in an oven at 80 °C for 48 h (until the weight was constant), and used to convert wet weight to dry biomass. Dry litter necromass was calculated using the following equation:

$$\mathrm{N}\mathrm{e}\mathrm{c}\mathrm{r}\mathrm{o}\mathrm{m}\mathrm{a}\mathrm{s}\mathrm{s} \mathrm{T}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l} (\mathrm{M}\mathrm{g} {\mathrm{h}\mathrm{a}}^{-1})=\left[{\displaystyle \frac{B{B}_{total}\times \left(\frac{B{K}_{sub}}{B{B}_{sub}}\right)}{A}}\right]\times 10$$

(2)

where:

• $B{B}_{total}$ = total wet weight of sample in quadrant (kg = 10⁻³ Mg);
• $B{B}_{sub}$ = sub-sample wet weight (g);
• $B{K}_{sub}$ = sub-sample wet weight (g; dried till constant weight);
• A = square area (m²) is 0.5 m × 0.5 m → A = 0.25 × 10⁻⁴ ha.

Understory biomass was measured by destructively harvesting all understory vegetation within the quadrat (cut at ground level), weighing it (wet weight), taking subsamples, oven-drying to constant weight, and then calculating dry biomass using the same formula. Biomass values were expressed per quadrat area and converted to Mg ha⁻¹, then averaged across five points to obtain subplot values.

2.4. Soil Physical Parameters

Soil physical properties were analyzed to understand soil structure and water transport capacity within the forest system. Most characteristics were measured separately for the 0–10, 10–20 and 20–30 cm depth layers, but are reported here for the 0–30 cm layer only. Particle size distribution (particles < 2 mm) was determined using the pipette method. This method is based on Stokes’ Law regarding the sedimentation rate of particles in suspension, which allows for accurate separation of fractions based on particle size [48]. As standard ring samples are difficult to insert in the presence of tree roots, bulk density ($\rho b$) was measured using the block-sized (20 cm × 20 cm × 10 cm = 4000 cm) samples [49] collected in field moisture conditions. Particle density $\left(\rho p\right)$ was measured by the pycnometer method. Total soil porosity ($Ø$), the percentage of the total soil volume that is not filled by solid (soil) particles [50], was calculated [51] from bulk density data and particle density using the following equation:

$$Ø=⁅1-{\displaystyle \frac{\rho b}{\rho p}}⁆\times 100\%$$

(3)

Soil structural characteristics were further analyzed through aggregate stability, measured using the wet sieving method. Aggregate resistance to water-damaging forces was expressed through the Mean Weight Diameter (MWD) value, where a higher MWD value indicates a more stable soil structure and resistance to erosion [52].

In addition to static characteristics, the soil’s ability to absorb water vertically, or infiltrate, was measured in the field using the double-ring infiltrometer method [53]. Steady-state infiltrability was derived as in (Horton equation) [54]

F(t) = F_c + (F_o − F_c) e^−kt

(4)

where:

• F(t) = infiltration rate (cm hour⁻¹);
• F_o = initial infiltration rate (cm hour⁻¹);
• F_c = steady rate infiltration (cm hour ⁻¹);
• k = empirical constant (hour⁻¹);
• t = time (hour).

2.5. Soil Chemical Parameters

Standard soil chemical properties were analyzed to evaluate the fertility status and nutrient availability in the forest system. Soil acidity parameters were determined by measuring soil pH (H₂O) and pH(KCl) using an electrometric method with a pH meter on a soil and water suspension or a 1 M KCl solution with a soil: solution ratio of 1:1 [55]. Furthermore, the soil organic carbon content was determined using the wet oxidation method [56], in which carbon is oxidized with potassium dichromate in concentrated sulfuric acid [57]. A texture, pH, elevation and layer-depth (between D_up and D_Low) dependent reference value (C_ref) was calculated for comparison with measured C_org results, as in [58]

C_ref = 0.9  ×  (D_Low^0.705 − D_Up^0.705)/(0.705 × (D_Low − D_Up)) × EXP(A)

(5a)

with

A = 1.333  +  0.00994  ×  Clay% + 0.00699  ×  Silt% − 0.156  × pH(KCl) +
+ 0.000427  ×  Elev + 0.834  ×  Andisol? + 0.363  ×  Wetland?

(5b)

“Andisol?” and “Wetland?” are categorical variables with values 0 or 1. As the soils in the study area are not classified as Andisols, despite having the same andic properties, the C_ref may be an underestimate of what long-term natural forests can achieve in this environment.

For soil total nitrogen parameters, analysis was carried out using the Kjeldahl method, which includes the stages of destruction, distillation, and titration to measure the accumulation of nitrogen in the form of ammonium [59]. Meanwhile, the availability of soil phosphorus (available P) was extracted using the Bray 1 method, which is particularly relevant for soils with acidic pH, and its color intensity was measured using a spectrophotometer at specific wavelengths [60]. The characteristics of the soil cation exchange complex were analyzed using the 1 N NH₄OAc extraction method at pH 7.0. Exchangeable base cations, including potassium (K exchangeable), sodium (Na exchangeable), calcium (Ca exchangeable), and magnesium (Mg exchangeable), were determined using a spectrophotometer (Flame Photometer) [61]. The same ammonium acetate extraction method was also used to determine the Cation Exchange Capacity (CEC) value, which reflects the soil’s ability to absorb and provide cations to plants. Finally, the Base Saturation was calculated through the ratio of the total amount of base cations to the CEC value.

2.6. Earthworm Characteristics

Ecologically, earthworms are grouped into three groups: epigeic, endogeic, and anecic. These three ecological groups are distinguished based on their morphology and coloration. Earthworm sampling was carried out using the monolith method standardized by the Tropical Soil Biology and Fertility Program (TSBF). Earthworm samples were taken from a 20 × 20 m subplot. Sampling was carried out using a frame size of 50 cm × 50 cm × 10 cm. Earthworms were taken at three depths for each monolith from three depths (0–10 cm, 10–20 cm, and 20–30 cm). (Figure 4C). The worms were separated from the soil samples directly in the field using the hand-sorting method (manual). The earthworms were then cleaned in water and counted [62]. The worms were then placed in bottles containing 5% formalin for preservation, and their weight was determined and measured in the laboratory.

2.7. Data Analysis

The data obtained in this study were analyzed by analysis of variance (ANOVA) using the F test at the 5% probability level using the GenStat Discovery (18.1) program. If a difference between groups (or ‘apparent effect’) was unlikely due to random variation (p < 0.05), a further test was applied at the 5% LSD level. Correlation and regression tests were carried out to determine the direction of the relationship between variables and the level of association.

3. Results

3.1. Aboveground Vegetation

The results for above-ground vegetation (

Table 1. Aboveground properties of the vegetation of plots in the survey.

	Canopy Density (%)	Tree Population (ha−1)	Basal Area Total (m2 ha−1)	Understory Biomass (Mg ha−1)	Litter Necromass (Mg ha−1)	Litter Thickness (cm)
Land use on very steep slopes:
Protection forest management	76.1 c	201.7 a	65.2 c	5.39 b	2.36	2.13
Young agroforestry (YAF)	15.1 a	722.0 b	12.2 a	0.38 a	0.28	0.27
Old agroforestry (OAF)	27.7 b	568.3 b	34.2 b	0.55 a	0.54	0.48
Relative difference 1	1.38	−0.89	1.13	2.34	1.84	1.84
t-test (PFM vs. AF, 2-sided)	**	**	**	*	*	**
ANOVA result	**	*	**	*	*	**
LSD	11.48	355.1	13.3	2.88	0.79	0.51
Age effect (YAF vs. OAF):
Young AF (YAF)	17.1 a	713.9	12.8 a	0.84	0.50 a	0.54
Old AF (OAF)	40.0 b	704.3	39.8 b	0.97	1.41 b	0.60
Relative age effect 2	0.80	−0.01	1.03	0.14	0.95	0.10
ANOVA result	**	NS	**	NS	**	NS
LSD	4.4	107.1	3.32	0.26	0.40	0.11
Slope (in YAF, OAF only)
A. Flat (0%–8%)	34.5 b	843.0 b	29.7 b	1.56 c	1.40 b	0.82 c
B. Sloping (8%–15%)	28.3 ab	757.8 ab	24.3 a	0.98 b	1.28 b	0.64 b
C. Rather steep (15%–25%)	30.2 b	704.5 ab	26.4 ab	1.02 b	0.84 ab	0.54 ab
D. Steep (25%–45%)	28.3 ab	595.0 a	27.9 ab	0.50 a	0.87 ab	0.50 ab
E. Very steep (>45%)	21.4 a	645.2 a	23.2 a	0.47 a	0.41 a	0.37 a
Relative slope effect 3	−0.93	−0.79	−0.36	−2.96	−2.49	−1.79
ANOVA result	*	*	*	**	*	*
LSD	7.0	169.3	5.25	0.41	0.63	0.18

Notes: ¹ Dimensionless Relative difference = 0.5 × (2 × PFM-YAF-OAF)/AVERAGE(PFM, YAF, OAF); ² Relative apparent age effect = 2 × (OAF − YAF)/(OAF + YAF); ³ Dimensionless Relative apparent slope effect = (2 × E+ D − B − 2 × A)/Average (A…E); statistical tests: NS = non-significant (p > 0.05), * = 0.05 > p > 0.01, ** = p < 0.01; values followed by a same letter are not statistically distinguishable in an LSD (least significant difference) test.

) confirmed impressions in the field of effects of slope and stand age. On steep slopes, canopy density increased with age of the pine plantations under agroforestry management from YAF to OAF but remained 1.38 times lower than was measured at the natural forest comparator PFM (Table 1). This increase in canopy cover was accompanied by changes in stand structure and ground surface components: tree population in AF was 0.89 times higher than in PFM, while total basal area increased but was 1.13 times lower than PFM, and accumulation of understory and ground components increased, indicated by significantly increased understory biomass, litter necromass, and litter thickness, which were 2.34, 1.84, and 1.84, respectively, compared to PFM. When the effect of age was compared across slope classes, the main pattern remained the same: canopy density in OAF increased significantly compared to YAF, while tree population did not differ, total basal area increased, understory biomass did not differ, and litter components showed a strengthening of accumulation processes with significantly increased litter necromass and litter thickness. On the other hand, the influence of land slope within the AFS shows a clear negative gradient: the steeper the slope, the greater the decrease in canopy density, tree population, total basal area, understory biomass, litter necromass, and litter thickness, indicating that steeper slope conditions limit the development of canopy structure and reduce the capacity for accumulation of biomass and litter above the ground surface.

3.2. Soil Physical Conditions

Data on texture confirmed the homogeneity of basic soil properties in the area (

Table 2. Basic soil physical properties of the 0–30 cm topsoil and three contrasts in relative values.

	Sand (%)	Silt (%)	Clay (%)	Bulk Density (g cm−3)	Particle Density (g cm−3)	Soil Porosity (%)	Mean Weight Diam (mm)	Soil Infiltration (cm h−1)
Land use on very steep slopes:
Protection forest management (PFM)	59.3	24.2	16.5	0.68 a	2.54 a	73.2 b	0.899 b	3.16 b
Young agroforestry (YAF)	64.4	19.5	16.1	1.06 b	2.72 b	60.9 a	0.555 a	1.11 a
Old agroforestry (OAF)	58.1	26.6	13.9	0.84 a	2.71 b	69.3 b	0.629 a	1.09 a
Relative difference 1	−0.03	0.05	0.10	−0.31	−0.07	0.12	0.44	1.15
t-test (PFM vs. AF, 2-sided)	NS	NS	NS	*	**	NS	**	**
ANOVA result	NS	NS	NS	*	*	*	*	**
LSD	9.53	8.87	7.13	0.21	0.06	7.7	0.15	0.28
Age effect (YAF vs. OAF):
Young AF (YAF)	61.2	22.7 a	16.3	0.87	2.71 b	67.9	0.638	2.31 b
Old AF (OAF)	60.4	25.1 b	14.3	0.82	2.57 a	67.9	0.567	1.82 a
Relative age effect 2	−0.01	0.10	−0.13	−0.05	−0.05	0.00	−0.12	−0.24
ANOVA result	NS	*	*	NS	**	NS	NS	**
LSD	1.4	1.8	1.7	0.08	0.05	3.3	0.078	0.16
Slope (in YAF, OAF only)
A. Flat (0%–8%)	59.6 a	25.7 b	15.2	0.78 a	2.72 b	71.4 b	0.630	2.81 d
B. Sloping (8%–15%)	62.2 b	23.6 ab	14.3	0.80 a	2.57 a	68.5 ab	0.629	2.98 d
C. Rather steep (15%–25%)	61.3 ab	22.7 a	16.0	0.81 a	2.67 b	69.3 ab	0.558	1.92 c
D. Steep (25%–45%)	59.5 a	24.4 ab	16.1	0.88 ab	2.51 a	65.2 a	0.604	1.53 b
E. Very steep (>45%)	61.2 ab	23.1 ab	15.0	0.95 b	2.72 b	65.1 a	0.592	1.10 a
Relative slope effect 3	0.01	−0.18	0.09	0.49	−0.02	−0.23	−0.17	−2.35
ANOVA result	*	*	NS	*	**	*	NS	**
LSD	2.2	2.8	2.7	0.12	0.08	5.2	0.12	0.25

Notes: ¹ Relative difference = 0.5 × (2 × PFM-YAF-OAF)/AVERAGE(PFM, YAF, OAF); ² Relative apparent age effect = 2 × (OAF − YAF)/(OAF + YAF); ³ Relative apparent slope effect = (2 × E+ D − B − 2 × A)/Average (A…E); statistical tests: NS = non-significant (p > 0.05), * = 0.05 > p > 0.01, ** = p < 0.01; values followed by a same letter are not statistically distinguishable in an LSD (least significant difference) test.

). Variation in sand, silt, and clay content between PFM, YAF, and OAF observation sites was as expected under a no-difference null-hypothesis. With a very low between-replicate variance, small differences in silt and clay content in the AF slope and age comparison were statistically significant (Table 2). The high sand content in the measurements may be due to incomplete (using standard laboratory methods) sample dispersion of the strongly bound aggregates in a soil with andic properties. Therefore, sand here is a ‘pseudo-sand’. Variation in inherent soil properties such as texture between the PFM, YAF, and OAF observation sites was not likely to be a major factor in subsequent analysis of other soil properties, but small differences could still be compensated in the C_ref calculation.

More dynamic soil physical properties showed consistent differences across cover types and slopes (Table 2). In PFM, the soil had a significantly lower bulk density and particle density and higher porosity compared to YAF. It was similar to values for OAF for bulk density and porosity, while specific gravity still showed differences. Internal comparisons of agroforestry showed no significant differences between YAF and OAF in bulk density and porosity, while soil particle density was higher for YAF than for OAF. The effect of land slope showed a tendency that steeper slopes had higher bulk density and lower porosity, while specific gravity showed no clear pattern.

The Mean Weight Diameter as an indicator of structural aggregate stability was higher in PFM than either YAF or OAF, while aggregate stability was not significantly different between YAF and OAF, even when compared across slope classes.

Infiltration rates varied in the range of 1–3 cm h⁻¹, which is relevant under high rainfall intensities that occur in this landscape. On steep slopes, soil infiltration was significantly higher in PFM than in YAF and OAF. When comparing YAF and OAF, the age of the agroforestry did not result in significant changes in infiltration. On steeper slopes (where it matters more, as surface runoff is faster), the soil infiltration rates were lower.

3.3. Soil Chemical Conditions

On steep slopes, soil pH in PFM was not significantly different from that in YAF and OAF (

Table 3. Basic soil chemical properties of the 0–30 cm topsoil and three contrasts in relative values.

		COrg (%)	COrg/ CRef	NTot (mg/kg)	P_Bray, (mg/kg)	Exchangeable Cations (mMolc/kg)				Base Saturation (%)
	pH_ H2O					K+	Na+	Ca++	Mg++
Land use on very steep slopes:
Protection forest management (PFM)	6.58	2.99 b	1.12 b	0.343 c	17.2 b	1.73 c	0.12 a	10.1 ab	0.68	83.89 c
Young agroforestry (YAF)	6.03	0.98 a	0.32 a	0.172 b	17.1 b	0.79 a	0.23 c	10.6 b	0.42	46.24 a
Old agroforestry (OAF)	5.90	1.42 a	0.49 a	0.102 a	10.7 a	1.14 b	0.18 b	9.4 a	0.44	65.99 b
Relative difference 1	0.10	1.00	1.11	1.00	0.22	0.63	−0.47	0.01	0.49	0.42
t-test (PFM vs. AF, 2-sided)	NS	*	**	**	NS	**	**	NS	NS	**
ANOVA result	NS	*	**	**	**	**	**	*	NS	**
LSD	0.73	1.14	0.24	0.02	1.34	0.11	0.01	0.89	0.49	6.76
Age effect (YAF vs. OAF):
Young AF (YAF)	6.20 b	1.73	0.52 a	0.178 b	21.4 b	0.94 a	0.21 b	10.5 b	0.53	69.33
Old AF (OAF)	5.90 a	1.75	0.77 b	0.158 a	16.5 a	1.02 b	0.15 a	8.9 a	0.60	68.92
Relative age effect 2	−0.05	0.01	0.39	−0.12	−0.26	0.08	−0.37	−0.16	0.12	−0.01
ANOVA result	*	NS	**	**	**	**	**	**	NS	NS
LSD	0.24	0.12	0.09	0.011	0.5	0.03	0.003	0.23	0.08	7.08
Slope (in YAF, OAF only)
A. Flat (0%–8%)	6.05	2.26 d	0.85 d	0.202 c	25.4 c	1.03 c	0.15 a	9.5 b	0.74 b	68.74 bc
B. Sloping (8%–15%)	5.86	1.91 cd	0.72 cd	0.206 c	25.2 c	1.23 d	0.17 b	8.8 a	0.65 ab	76.64 cd
C. Rather steep (15%–25%)	6.23	1.83 bc	0.66 bc	0.170 b	13.9 a	0.92 b	0.18 c	8.8 a	0.53 ab	58.01 ab
D. Steep (25%–45%)	6.16	1.50 ab	0.57 b	0.124 a	16.5 b	0.76 a	0.20 d	11.4 d	0.47 a	86.13 d
E. Very steep (>45%)	5.97	1.20 a	0.41 a	0.137 a	13.9 a	0.96 b	0.21 d	10.0 c	0.43 a	56.11 a
Relative slope effect 3	0.02	−1.45	−1.62	−1.26	−1.68	−0.62	0.78	0.36	−1.45	−0.23
ANOVA result	NS	**	**	**	**	**	**	**	*	**
LSD	0.38	0.19	0.14	0.176	0.8	0.05	0.005	0.37	0.12	0.75

Notes: ¹ Relative difference = 0.5 × (2 × PFM-YAF-OAF)/AVERAGE(PFM, YAF, OAF); ² Relative apparent age effect = 2 × (OAF − YAF)/(OAF + YAF); ³ Relative apparent slope effect = (2 × E+ D − B − 2 × A)/Average (A…E); statistical tests: NS = non-significant (p > 0.05), * = 0.05 > p > 0.01, ** = p < 0.01; values followed by a same letter are not statistically distinguishable in an LSD (least significant difference) test.

). Similarly, differences in slope in AF land did not significantly affect soil pH, but YAF land was significantly more acidic than OAF land. The analysis results showed that the organic C (C-org) content in YAF land was not significantly different from that of OAF. However, when organic C was expressed as an indicator relative to reference conditions (C_Org/C_Ref ratio), a more sensitive difference was observed: the C_Org/C_Ref ratio in YAF was significantly lower than in OAF. Differences in soil pH and slight variation in texture were reflected in the C_Ref values. The C_Org/C_Ref ratio in PFM was above 1.0, which can be attributed to the andic properties, as noted. In YAF and OAF, both C_Org and the C_Org/C_Ref ratio were significantly lower than in PFM, while they decreased with increasing slope.

Total soil nitrogen in PFM was significantly higher than that in YAF and OAF (Table 3). With increasing AFS ages, total soil nitrogen decreased significantly (as fertilized annual food crops were replaced by non-fertilized and regularly cut perennial grasses). Total soil nitrogen in AFS decreased significantly with increasing slope. Available P was also significantly higher in PFM than in OAF, but not significantly different in YAF. In the AF system, available P in YAF was significantly higher than in OAF. The same pattern as Total Soil Nitrogen, available P significantly decreased with increasing slope, except on rather steep land, where it was lower than on steep land.

On very steep land, exchangeable K in PFM was significantly higher than in YAF and OAF (Table 3). In the agroforestry system, exchangeable K in YAF was significantly higher than in OAF. Conversely, exchangeable Na in PFM was significantly lower than in YAF and OAF. In the agroforestry system, exchangeable Na in YAF was significantly higher than in OAF. Exchangeable K in YAF was significantly higher than in OAF. Conversely, exchangeable Ca in PFM was significantly similar to YAF, but higher than in OAF. In the agroforestry system, exchangeable Ca in YAF was significantly higher than in OAF. Exchangeable Mg in PFM, YAF, and OAF were not significantly different. There was a similar pattern that exchangeable K and Mg decreased with increasing slope. There was a similar pattern that exchangeable Na and Ca increased with increasing slope. In the agroforestry system, Base saturation in PFM was significantly higher than in YAF and OAF. In the overall AF system, base saturation in YAF is the same as in OAF. Base saturation decreases significantly with increasing slope, except on steeper slopes.

3.4. Earthworms

On very steep slopes, the biomass and population of earthworms in PFM were significantly higher than those in YAF and OAF, while the average size of earthworms in YAF was the smallest (

Table 4. Earthworm data for the 0–30 cm topsoil layer.

	Biomass (g m−2)	Population (m−2)	Average Size (g)	Diversity (Shannon)
Land use on very steep slopes:
Protection forest management	115.2 b	110 b	1.04 a	1.50 b
Young agroforestry (YAF)	11.5 a	33 ab	0.35 b	0.87 a
Old agroforestry (OAF)	16.5 a	14 a	1.18 a	0.63 a
Relative difference 1	2.84	3.40	1.05	0.27
t-test (PFM vs. AF, 2-sided)	2.03	0.94	1.15	0.79
ANOVA result	*	*	**	**
LSD	65.1	83	0.65	0.39
Age effect (YAF vs. OAF):
Young AF (YAF)	31.9 a	110 b	0.29 a	1.04 b
Old AF (OAF)	43.1 b	41 a	1.05 b	0.67 a
Relative age effect 2	−0.30	0.69	−1.05	0.44
ANOVA result	*	**	**	**
LSD	10.8	29	0.51	0.24
Slope (in YAF, OAF only)
A. Flat (0%–8%)	71.2 c	149 b	0.48 b	1.19 b
B. Sloping (8%–15%)	43.8 b	106 b	0.41 b	0.93 ab
C. Rather steep (15%–25%)	30.9 ab	57 a	0.54 ab	0.68 a
D. Steep (25%–45%)	27.7 ab	42 a	0.66 a	0.72 a
E. Very steep (>45%)	14.0 a	24 a	0.58 ab	0.75 a
Relative slope effect 3	2.84	3.40	0.27	1.05
ANOVA result	**	**	*	*
LSD	17.0	45	0.81	0.37

Notes: ¹ Relative difference = 0.5 × (2 × PFM-YAF-OAF)/AVERAGE(PFM, YAF, OAF); ² Relative apparent age effect = 2 × (OAF − YAF)/(OAF + YAF); ³ Relative apparent slope effect = (2 × E+ D − B − 2 × A)/Average (A…E); statistical tests: * = 0.05 > p > 0.01, ** = p < 0.01; values followed by a same letter are not statistically distinguishable in an LSD (least significant difference) test.

). Agroforestry land in YAF had a high population of earthworms but low biomass, resulting in a smaller average worm size than in OAF. Earthworm biomass and population significantly decreased with increasing slope, while the average worm size increased. Earthworm diversity was also significantly higher in PFM than in YAF and OAF. Earthworm diversity in YAF was higher than in OAF. Earthworm diversity also significantly decreased with increasing slope. In the land with a stable ecosystem in PFM, epigeic and anecic worms dominated, and endogeic worms were absent (

Table 5. Morphospecies identified in the various plots.

LUS	PFM	YAF					OAF
Slope class	E	A	B	C	D	E	A	B	C	D	E
Morphospecies:
Epigeics:
Methaphire javanica	40
Amynthas sp. (2)	33	23	23				88	58
Amynthas sp.	48	23	24	32	55	47	62	95	115	137	102
Anecics:
Pheretima sp.	44	48	38				16	5
Pheretima sp. (2)	36	43	43	38	44	49
Endogeics:
Pontoscolex corethrurus		62	62	130	101	104	34	42	85	63	98

). Epigeic, anecic, and endogeic worms were all found in YAF and OAF, but in general, the number of epigeic worms increased, and the number of anecic and endogeic worms decreased in OAF compared to YAF (Table 5).

3.5. Accounting for Differences in Infiltration

The analysis results show a positive relationship between soil porosity and soil infiltration rate (Figure 5), confirming that increasing pore space in the soil is followed by an increase in the soil’s ability to conduct water into the soil profile. Variations in both properties are primarily controlled by biological factors and soil quality, namely earthworm population, earthworm biomass, and soil organic matter content (Figure 6), which play an important role in biopore formation, improving soil structure, and increasing pore continuity.

As a single variable, variation in Canop Cover, C_org and earthworm populations has similar ‘explanatory’ skill in predicting variation in infiltration rates (

Table 6. Multiple regression models for infiltration. In the last row, all variables are replaced by their value relative to the mean for the current dataset.

Explanatory Variables	Regression Models for Infiltration (I)	r
Single variables
Canopy Cover	I = 1.38 + 0.02 Canopy Cover	0.69
Soil C-Organic	I = 0.84 + 0.74 Soil C-Organic	0.68
Earthworm population	I = 1.65 + 0.01 Earthworm population	0.68
Soil Porosity	I = −4.22 + 0.10 Soil Porosity	0.61
Soil aggregate stability	I = 0.26 + 3.10 Soil Aggregate stability	0.61
Earthworm biomass	I = 1.75 + 0.01 Earthworm biomass	0.60
Multiple variables	I = −1.9689 + 0.2232 × Soil C-Org + 0.0401 × Soil Porosity + 0.0055 × Earthworm population − 0.0034 × Earthworm biomass + 0.6165 × Soil Aggregate stability + 0.0098 × Canopy Cover	0.73
	I = 2.167 × (0.979 × RelMeanWeightDiam + 0.121 × RelC-Org + 0.00059 × RelPorosity + 0.0000070 × RelEarthworm_pop − 0.000076 × RelEarthworm_Biom + 0.000298 × RelCanCover −0.908)	0.73

). For soil porosity, soil aggregate stability and earthworm biomass, the explanatory skill is somewhat limited. However, all these ‘explanatory variables’ are highly correlated, and a multiple regression involving various factors accounts for only slightly more variation than either of the best single parameters. The easiest-to-use predictor of infiltration rates is, surprisingly, the percentage canopy cover. When variables are expressed relative to their mean value for the dataset. Variation in Mean Weight Diameter (aggregate stability) has the strongest effect on infiltration rates (last row in Table 6).

4. Discussion

Before discussing the answers we obtained to the three questions raised in the introduction, we reiterate that the data presented here were ‘survey’ data, rather than derived from randomized treatment allocations. Thus, any variation and apparent differences observed can be due to a priori conditions and variations in plot history, as well as to the current land cover at the time of sampling. A recent meta-analysis study [63] of 125 studies explored how agroforestry practices influence soil quality across different climate zones (tropical, temperate, and Mediterranean). However, the ‘apparent effect’ sizes for many properties, including aggregate stability and soil water regulation where similar to recorded differences in texture. In the absence of plausible causal mechanisms that change texture at the relevant time scales, differences in texture in observational data are warnings of confounding factors. As existing land use is usually not random with respect to soil quality, texture-corrected comparisons are probably more meaningful. A similar problem of effect sizes that are similar to observed texture differences can also be noted in a meta-analysis of deforestation studies [64]. In our data, uniformity of texture and pH (Table 1 and Table 2) gives some confidence in the validity of comparisons, but plot-level land cover history is not available at the level of detail that is desirable.

The first question, on the functional equivalence of pine plantations (with current agroforestry management) and remnant natural forest with a watershed protection forest status, had a very clear answer. The obvious aboveground differences in tree canopy cover, understory vegetation and litter layer are directly related to differences in soil organic carbon, aggregate stability, porosity, earthworms (biomass, population, and diversity) and infiltration rates. A caveat is that this comparison could only be made on the steepest slope class, as easily accessible places had been converted to agriculture and subsequent pine plantations, and agroforestry management in their second cycle. Attributing the observed differences to specific events in this and the use history is challenging. However, for the measured attributes, the pine plantations on the lowest slope class still could not match the remnant forest on the steepest slopes. On the basis of the formal criteria for forest classification, it may be hard to explain why a ‘production forest’ classification is possible on the steepest slope class, but local forest and land use history probably provides explanations [11,13]. On Eudic Eutrandepts, close to and probably similar to the soils of our study sites, a comparison across six land use types [65] found that the average bulk density of remnant natural forests was 0.74 g cm⁻³ and porosity 66%, while for upland crops, pine plantations and brushland it was in the 0.94–1.04 and 49%–57% ranges, respectively. Pine plantations were closer to remnant natural forest in aggregate stability and C_org than they were to dryland farms and (previously cropped) brushland. Our results for the natural forest—pine plantation comparison are similar (Table 2). A caveat for interpreting our data for field conditions is that, along with the soil physical properties measured here, root systems, animal activity, cracks, and stones affect soil infiltration capacity and preferential flow pathways along hillslopes [66].

The second question, a comparison of slope and age as factors in functional soil properties, had clear answers with regard to slope cases as a factor within the two age categories, and weaker evidence for the age comparisons as such, with more likelihood that confounding factors existed in different land use histories. Soils on steeper slopes had fewer ‘functional topsoil’ characteristics, and this can be due to both loss of topsoil by erosion, downhill movement of litter and organic inputs to soil health, and interaction with less understory vegetation and surface litter, and less protection from the direct influences of radiation and rainfall. Steeper soils had lower infiltration rates, while there is less time available for surface ponding and subsequent infiltration, both because of slope and the lower litter necromass conditions that can provide surface roughness. While specific data on the impact of current agroforestry management practices, as regulated in the contracts between the forest management authority and local communities, is desirable, the current research cannot tease apart which activities (between past harvest-related compaction, post-logging degradation and current management) are most important for the current status of the soil. A major part of the literature compares agroforestry with monoculture food crops on sloping land, rather than with remnant forest, and comes to positive conclusions. For example, experiments with agroforestry with fruit trees, contour grass strips and maize or coffee on sloping land in northwest Vietnam showed that such practices, in comparison with monoculture maize, reduced soil (and carbon and nutrient) losses through terrace formation [67]. Agroforestry practices within the pine plantations in our study did not include support for terrace formation or their maintenance, where they had been part of the cropping history before reforestation took place.

The third question provided multiple perspectives on answers. While earthworm data (biomass, population, diversity) were significantly correlated with variation in infiltration rates, the linkage between porosity and infiltration rate was not as tight as expected, while alternative explanatory variables (soil organic carbon, aggregate stability) had equal ‘explanatory skill’, and no single factor was more powerful than the simple canopy cover metric. A data set like we currently have cannot tease apart how multiple and interactive causative pathways work: direct protection of the topsoil by litter, the response of tree root systems and soil biota (including earthworms) to the way aboveground vegetation is managed, and the balance between soil structure regeneration and compaction by human activities.

Quantitative studies of infiltration in forest plantations tend to have positive conclusions where agricultural land use is the comparator [68,69,70], although the time course of recovery may be counted in decades rather than years [37,71] and negative or contested results where natural forests are the comparator [72,73,74]. Shifts from mono-specific to mixed stands are generally positive for infiltration rates [75], but neutral results have also been documented [76]. Detailed attribution of changed infiltration to roots, earthworms or other soil engineers remains scarce [77]. The relevant literature broadly agrees that the earthworm community activity serves as a valuable indicator of soil quality and health, particularly in response to various management practices or ecosystem disturbances [78]. Earthworm populations and biomass are typically higher in deciduous forests than in evergreen forests, such as pine plantations. A recent meta-study of earthworm population density and biomass compared undisturbed ecosystems (i.e., undisturbed grassland, primary forest) as controls against agricultural land-use treatments, with data extracted from 113 publications across 44 countries [79] found that arable cropland had significantly lower earthworm density (−18%), biomass (−15%), and species richness (−27%) compared to undisturbed sites. However, earthworm density, biomass and species richness that were higher than those in undisturbed sites were observed in pastures, sites under agroforestry, crop management with fallow periods and crop–livestock integration. A comparison of the saturated hydraulic conductivity in three forest stands (pine (Pinus taeda), oak (Quercus acutissima), and bamboo (Phyllostachys edulis) forests in Jiangsu (a subtropical coastal Chinese province north of Shanghai) that total soil porosity (and associated saturated water capacity) was the simplest predictor of differences in infiltration rates, with the Pine forests the best performing, sand content negatively and soil C positively correlated. Bamboo forests had the poorest water infiltration properties [80,81]. Studies in agroforestry systems elsewhere with Quercus suber as the dominant species concluded that earthworm abundance (three species of Aporrectodea, mostly anecic) increased aggregate stability irrespective of litter quality [82]. In a microcosm experiment, soil saturated hydraulic conductivity (Ks) showed little difference between different treatments due to the horizontal and high–branched burrows of the anecic earthworm Eissenia fetida, although higher burrowing activities were found in wetter soil and aggregate stability increased with worm abundance [83].

While the impacts of changed infiltration rates in these volcanic landscapes can be site-specific and complex [84,85], a closer ‘mimicking’ by plantation forestry of remnant natural forests is desirable for these mountain landscapes. Increased rainfall variability under climate change [86]. Testing of ‘restoration’ interventions with randomized treatment allocation will be needed to support ongoing negotiations with farmers, forest and watershed stakeholders. The overall significance of the research is that our data support the urgency of increased soil conservation measures where pine-based agroforestry practices are used on steep and very steep soils.

5. Conclusions

Pine-based agroforestry on very steep soils had only 35% of the infiltration rate of reference remnant natural forest plots, despite having 83%–95% of the soil porosity. Reduced earthworm biomass (only 10%–14% of the forest values) and compaction (23%–55% increased bulk density) probably represent two steps in the cause-and-effect chains. Understory vegetation and litter layer were more than fivefold higher in the natural forest. Where recent discussions after the floods in Sumatra do not differentiate between plantations and natural forest, conclusions may well be overgeneralized.

Within pine-based agroforestry, relative decline with slope class was pronounced in earthworm indicators (biomass −3.46 and population −4.18) and infiltration rates (−2.35) while bulk density increased (0.49). Soil chemical changes (C_Org −1.45, N_Tot −1.26, P_Bray −1.68, and exchangeable Mg −1.45) indicated a loss of topsoil. Differences with age of the agroforestry systems were much smaller and included a decreasing earthworm population density but an increase in mean earthworm weight.

Across all samples, a higher C_org and higher earthworm biomass were associated with complementary positive changes in infiltration rates and soil porosity.