WEATHER DATA

The automated weather station was installed on 20 May 2001 in an area on Nevado de Colima that is relatively accessible with a four-wheel-drive vehicle and that is actively being protected for conservation and research purposes by El Patronato del Nevado de Colima y Cuencas Adyacentes, A.C. It should be emphasized that no act of vandalism or human interference has occurred to date, despite the presence of people all year round and the high visibility of the equipment. On the contrary, a few days after the installation, local land managers built a fence all around the weather station (Fig. 2), as well as around the tree growth monitoring sites (see DENDROMETER DATA), to protect them from cattle damage, since grazing is allowed in the area. Here we present the instrumental design of the automated weather station and a summary of the observations recorded from 22 May 2001 to 24 March 2004 on atmospheric pressure, precipitation, incoming solar radiation, air and soil temperature, relative humidity, soil moisture, and wind speed and direction.

The automated weather station is located in an open, relatively flat area about 200 m below treeline on Nevado de Colima (3760 m elevation, 19°34.778′N latitude, 103°37.180′W longitude), within the Pinus hartwegii Lindl. forest, from which a 400-year-long tree-ring chronology has been developed (Biondi, 2001). The station is made of well-proven, commercially available components, configured to operate at high elevation in the North American Monsoon region. Measurements and control functions are accomplished by a Campbell Scientific, Inc., model CR10X datalogger with 2 Mb of memory. All sensors are controlled by the datalogger, which stores processed data at 30-min intervals. With the current settings, the datalogger has sufficient memory to store about 12 months of data. Specialized software and RS232 interfaces are used for communication between the datalogger and a laptop computer.

Sensors consist of a Vaisala ultrasonic wind sensor, a LI-COR silicon pyranometer, a Kipp & Zonen pyranometer, a Vaisala temperature and relative humidity probe, a Vaisala barometric pressure sensor (specially calibrated for high-elevation sites to record pressures between 500 and 1100 mb), a Texas Electronics rain gauge1, a Hydrological Services rain gauge (for high-intensity rainfall, up to 500 mm/h), a soil water content probe, and a soil temperature probe. Electrical power to the station is supplied by a 12-volt DC system that uses a 20-watt solar panel and regulator to recharge a sealed, 7 amp/h battery. The datalogger, battery, pressure sensor, and peripheral electronics are housed in a white, fiberglass-reinforced enclosure, which also contains desiccant packs to control humidity. Field maintenance followed standard guidelines (Blauvelt, 2001; Hubbard and Sivakumar, 2001). Currently, it is not possible to remotely retrieve data from the station in near-real-time mode. A GOES satellite connection comprised of high-data-rate satellite transmitter with on-board GPS, antenna, and cables was included in the May 2001 installation, but it was never possible to obtain data via satellite, apparently due to a manufacturing defect in the transmitter. Because of this, sensor failures could only be detected during inspection trips, and repairs had to be delayed until the next visit. Despite such limitation, relatively few missing data are present in the 49,824-long record of 30-min observations (Table 1).

Post processing of the data required additional corrections. Wind speed and vector magnitudes, as well as wind gust, occasionally reached impossible values (up to 1000 mi/h!), possibly because of a “glitch” in the software that handles communication between the datalogger and the sonic anemometer. Based on the range of existing measurements from other high-elevation sites (Gregory D. McCurdy, Western Regional Climate Center, Desert Research Institute, Reno, Nevada, personal communication), values of wind speed and wind vector magnitude greater than 50 mi/h, and values of wind gust greater than 100 mi/h, were considered missing. Furthermore, any relative humidity reading that exceeded 100% was considered equal to 100%. Since records provided by the two pyranometers were extremely similar in both absolute values and time series variability, as suggested by their exceptionally high correlation (r > 0.99), we report only measurements from the Kipp & Zonen one. The original Julian day format was converted to a month and day format, and all variables recorded in imperial units were transformed into metric units.

The 30-min data downloaded from the datalogger were reduced to daily values for further analysis. This was accomplished by selecting, for each day, the maximum 30-min value for wind gust, maximum soil temperature, maximum air temperature, and maximum relative humidity; the minimum 30-min value for minimum soil temperature, minimum air temperature, and minimum relative humidity; the total of all 30-min values for solar radiation (in kJ m⁻²) and precipitation; the arithmetic average of all 30-min values for wind speed, standard deviation of the wind vector direction, soil moisture (water percent of total volume), and atmospheric pressure. The wind vector direction, which is recorded as a polar coordinate in degrees, was first transformed into two Cartesian components, which were then averaged to obtain daily values, and finally the X and Y means for each day were converted back to a polar coordinate in degrees. In addition, the daily ranges of air temperature, soil temperature, and relative humidity were computed by subtracting the daily minimum from the daily maximum.

DENDROMETER DATA

The second component of our study on tropical timberline dendroclimatology was to obtain tree growth observations at the same temporal scale as the weather records. In order to do so, in May 2001 we installed automated point and band dendrometers (Keeland and Sharitz, 1993) to measure stem increment of Pinus hartwegii at two locations within a 1-km radius from the weather station. Point dendrometers, also called dial gage dendrometers (Reineke, 1932) or recording dendrometers (Fritts and Fritts, 1955), and dendrometer bands, also called vernier tree growth bands (Hall, 1944), have often been used to detect the seasonal growth patterns of woody species (Bormann and Kozlowski, 1962; Deslauriers et al., 2003). Our setup, as detailed below, allowed continuous and accurate recording of changes in the diameter of tree trunks (7 at Site 1, 8 at Site 2) at 30-min intervals and with a resolution of a few microns (Downes et al., 1999).

Site 1 (19°34.703′N, 103°37.137′W) is characterized by a west-northwest exposure, 25% slope, and 3790 m elevation. Site 2 (19°34.872′N, 103°36.728′W) has a north-northeast aspect and an average 3780 m elevation; it is farther away from the weather station and on a much steeper (58%) slope than Site 1. At each site, sensors were controlled by an underground field computer with 10 Mb of non-volatile memory, which is enough for recording up to 12 months of data at 30-min intervals. Specialized software and RS232 interfaces are used to communicate with a laptop computer. Electric power is supplied by a 6-volt DC system and solar panel. Trees closest to the underground computer were outfitted with sensors, hence sample selection was based solely on distance from the recording unit.

Sensors installed at each site consist of 3 band dendrometers, 11 point dendrometers, 1 phytogram per tree, 1 air temperature sensor, 1 photosynthetically active radiation sensor, 4 soil temperature probes, and 7 soil moisture probes. Soil probes were distributed around the site at similar depths. Point dendrometers were usually placed about 1.7–1.8 m above the ground on the south-facing side of the tree stem after shaving most of the bark underneath them. A few point dendrometers were installed on the north-facing side or without removing the bark or at a higher level to provide comparisons.

Unfortunately, the equipment was repeatedly damaged by lightning strikes, which are common at high elevations, but may have been even more frequent at the study sites because of the amount of electrical wire running above and below ground. Additional records were lost because of a rockfall at Site 2 and an insect outbreak at Site 1; the latter is described in detail in the section TREE GROWTH. During the first 24 months since installation, two dataloggers were destroyed by lightning, and various individual sensors malfunctioned due to grounding problems. Here we present the data that were salvaged from 2001 and 2002 at the two sites, including the remarkable growth signature produced by an outbreak of roundheaded pine beetle (Dendroctonus adjunctus Blandford), which ultimately killed most trees at one of our two experimental sites.

CLIMATE

Daily patterns of recorded environmental variables (Fig. 3, Table 1) reveal important features of tropical timberline climate. First, the warmest days are in May, when solar insolation increases and the sky is mostly cloud-free. This combination of factors brings soil moisture to its lowest point during the year (Fig. 3a). As soon as the monsoon precipitation begins, soil moisture increases while temperatures decrease rapidly, with daily minimum and maximum air temperatures in July being comparable to those during the winter months. Indeed, minimum air temperatures can drop below 0°C from July until March. On the other hand, daily minimum and maximum soil temperatures keep decreasing from May until January, but never fall below 0°C (Fig. 3a). The lowest maximum soil temperature (0.86°C) was recorded on 10 February 2004, and the lowest minimum soil temperature (0.75°C) occurred just the previous day (Fig. 3a). Maximum and minimum soil temperatures tend to follow quite similar patterns, as indicated by their very high correlation (r = 0.96), whereas maximum and minimum air temperatures are loosely coupled (r = 0.49). Linear correlation between soil and air temperatures is greatest for the two minima (r = 0.75) and smallest for minimum soil and maximum air temperatures (r = 0.44). The daily range of air temperature is on average about three times as large as the soil temperature range, and it can be as high as 15.7°C (Table 1). The difference in daily range between air and soil temperatures is related to nighttime conditions. In fact, maximum temperatures reach very similar values in the air and in the soil, whereas minimum temperatures in air are always significantly lower than in soil (Table 1, Fig. 3a). The seasonal cycle also differs between air temperature range, which is highest in winter and lowest in summer, and soil temperature range, which peaks in spring and bottoms in winter (Fig. 3b).2

The wet summer season, which begins in late May to early June and ends in late October to early November, is clearly identified by the precipitation measurements (Fig. 3a). The greatest amount of daily precipitation (164 mm) was recorded on 7 October 2003, with the second greatest value (108 mm) recorded just the day before, i.e., on 6 October 2003. Although isolated winter storms can occasionally bring moisture to the area, almost all precipitation falls between June and October, which causes relative humidity to be highest in those months (Fig. 3b). This, in combination with low air temperatures, brings air to saturation, hence maximum relative humidity remains at 100% for several days in a row (Fig. 3b). As minimum relative humidity increases in summer even more than the maximum one, the daily range of relative humidity bottoms out at this time of the year. The wet and dry seasons are also reflected in wind parameters: during the monsoon, winds tend to be mostly from the east (they are mostly from the west otherwise), with slightly greater variability in direction, and higher mean speed and peak gust (Fig. 3b).

The most intense weather events were recorded in January. First, on 13 January 2002, the atmospheric pressure dropped to 646 mb, almost 6 standard deviations below the long-term mean of 654.4 mb (Fig. 3a). Air temperature was extremely low, with the maximum falling below 0°C, to −0.85°C, and the minimum reaching −7.5°C (and −7.7°C the day after; Fig. 3a). Furthermore, 12 January was extremely cloudy, as shown by the very low value for incoming short-wave radiation (Fig. 3a). On 13 January the mean 30-min standard deviation of the wind vector direction was only 0.5°, much lower than usual (Fig. 3b). Winds were therefore constantly from the south (171°) and extremely strong. In fact, both wind speed and wind gust were considered missing for that day (because the sonic wind sensor showed values of 1000 mi/h), but they likely exceeded 160 km/h.

January 2004 was equally remarkable, having the lowest minimum temperatures on record (−12.7°C on 31 January, −10.6°C on 30 January, and −10.0°C on 17 January), and two sub-freezing maximum temperatures (−0.79°C on 16 January and −0.85°C on 30 January; Fig. 3a). The highest air temperature range (15.7°C) also occurred on 31 January (Fig. 3b). Barometric pressure dropped to some of its lowest values: 647 on 17 January and 648 mb on 16, 30, and 31 January; such values are more than 4 standard deviations below the long-term mean (Fig. 3a, Table 1).

The impact of hurricane Kenna is clearly visible in several weather variables on 25 October 2002 (Fig. 3a). However, many other storms had a greater impact on the study area than Kenna, such as the cyclonic system that brought extremely high rainfall on 6 and 7 October 2003 (Fig. 3a). In conclusion, although precipitation falls mostly during the summer monsoon, extreme weather events have appeared more often outside of the summer season, i.e., during January (winter storms) and October (tropical storms; Englehart and Douglas, 2001).

TREE GROWTH

Trees monitored at Sites 1 and 2 are of various sizes (Table 2). Mean stem diameter (DBH, measured at 1.3–1.5 m above ground) is 55 ± 4 cm at Site 1, and 60 ± 7 cm at Site 2; mean tree height is 19.2 ± 1.3 m at Site 1, and 18.1 ± 1.2 m at Site 2. Inter-tree distance varies from 5.6 to 40.7 m at Site 1, and from 4.2 to 41.9 m at Site 2. Based on a two-sample t-test (Davis, 1986), mean DBH, height, and inter-tree distance are not significantly different between the two sites (p-values > 0.25). Comparisons between growth patterns at the onset of the 2002 summer monsoon season are feasible at Site 2 (Fig. 4). Both daily and seasonal patterns of radial increment are shown by the individual curves. Tree trunks can either shrink or expand at hourly to daily timescales because of thermal and hydration processes unrelated to wood growth, but those changes are most pronounced when measurements include the bark (Kozlowski et al., 1991). Inside bark, some shrinkage is expected during the day when transpiration exceeds water absorption, while a small swelling may occur at night because of excess absorption with respect to transpiration. Despite the presence of daily cycles, the growing season onset is clearly defined by the time when stem size significantly increases on a monthly timescale: although patterns differ from one tree to another, most of them initiate growth by the second half of April (Fig. 4).

A complete representation of seasonal growth patterns is given by point dendrometers at Site 1 (Fig. 5). From these measurements, it is clear that diurnal cycles unrelated to wood growth are smoothed out and irrelevant when records are plotted over several months. It is also evident that wood formation begins in March–April and continues until October–November, and that a definite dormant period occurs between growing seasons. The graphical relationship between stem size and precipitation (Fig. 5) reveals that occasional winter storms can create a temporary swelling of the trunk, which gradually returns to its previous size when water evaporates. This “hydration” noise is much larger for point dendrometers that measure radial changes outside bark. For such instruments (data not shown), stem size actually decreases after the summer precipitation, and trunk swelling from winter storms is greatly amplified. Considering the patterns of weather variables previously described, one can hypothesize that annual radial growth is initiated in response to increasing spring temperatures and then continues because of summer precipitation in conjunction with above-freezing soil temperatures.

Upon returning to Nevado de Colima in mid-March of 2003, we were surprised to find four of the seven trees at Site 1 dying from a pest infestation. Roundheaded pine beetle (Dendroctonus adjunctus Blandford) has been a problem in the Pinus hartwegii forest for some time. Local land managers usually contain the spread of the insects by cutting down affected trees and burning the litter. Trees had already been logged all around Site 1, and the only pines left with signs of infestation were those inside the fence that surrounds our instrumented trees. Because of the cluster of dead and dying trees, whose brownish crowns contrast with the green ones of the surrounding live trees, it was even possible to spot Site 1 from a distance, as it stood out in the overall landscape. While a disappointment for the long-term implications of our study, this situation yielded a unique opportunity for recording in real time the radial growth response to the beetle outbreak.

Increment patterns of infested trees are remarkably different from those of healthy trees (Fig. 5). In 2001, tree 7 was not growing as rapidly as the other three pines, which seemed quite healthy. In 2002, trees 2 and 5 (the ones attacked by bark beetles) display stem sizes that are either flat or declining, whereas trees 1 and 7 (spared by the insect outbreak) are growing normally. Damage from the beetle infestation is quite distinctive (Massey et al., 1977), as yellowish resin mixed with boring dust may exude from the entry holes, making the attack points visible on the bark. We inspected Site 1 in June 2002 and again in December of that year, but there was no notice of any characteristic beetle damage at that time. However, the dendrometer data (Fig. 5) point to a suppression of growth starting in July 2002. The initial infestation for documented cases in Nevada and New Mexico takes place during the fall, late September to early November. By May of the following year, the crowns turn light green and are often completely brown by July (Massey et al., 1977). We argue that our dendrometer measurements reflect the true timing of the tree response to the beetle attack in this tropical treeline forest. The cessation and reversal of stem growth could be a result of the progressive inactivation of the vascular cambium combined with stem dehydration and shrinkage because of the reduced xylem tissue capable of actively conducting and transporting water and nutrients.