CANOPY GREENNESS BY LEAF PHENOPHASE & SITE MICROCLIMATE
Figure R1: Leaf phenophase specific greenness and microclimate scatter plots depicting daily average NDVI plotted against temperature, relative humidity, vapor pressure deficit (VPD), and soil moisture for each of the 365 days of the year for both tropical dry forest monitoring sites, Mexico and Brazil. Phenophases are represented by color
When the microclimatic variables are paired with the canopy greenness on a daily basis and categorized in the four different phenophases several patterns emerge. First of all the annual canopy leaf dynamics can be observed moving from dormancy through to senescence in general counterclockwise cycle for all plots except for vapor pressure deficit (VPD) in figure R1. Starting with dormancy, greenness values remain low on the NDVI scale, between 0.40 and 0.45 for mexico and between 0.5 and 0.6 for Brazil. During the green-up a high rate of change in NDVI brings the greenness scale up near 0.80 for most of the growing season when leaves are mature. For both sites the canopy greenness remains above 0.75 during the mature leaf phenophase despite larger fluctuations in the met variables. Then towards the end of the growing season the NDVI can be seen to slowly decrease back down to the dormancy state. Except for some general trends emerging during the mature phenophase, there is little apparent relationship between the meteorological (met) variables and the state of greenness of the canopy. During green-up we see a large change in the met variables over an 8-12 day period before we observe an increase in canopy greenness. Looking at the light green points in figure R1 we see that temperature and VPD decreases while soil moisture and air humidity increase at the end of the dormancy and the start of the green-up phenophase with little change in NDVI, and then NDVI increases very quickly with little to no change in the met conditions. The only site that shows a synchronized increase during the green-up is in the Mexican soil moisture vs NDVI plot, where we see the NDVI increasing with a decreasing soil moisture content, the opposite of what we might predict. Indeed soil moisture values are a much more transient phenomenon than the other met variables and will increase rapidly following rainfall, then decrease as surface water evaporates, runs-off, and is absorbed deeper into the soil and surrounding plant roots. What we see is not a synchronized cause-effect relationship but rather an offset in meteorology, water availability and forest canopy response. This mistiming is explored in further detail in the next section (figure R2) and on the next page .
Exploring figure R1 further, during the growing season mature phenophase we see that canopy greenness decreases slightly with increasing temperature and VPD, and with decreasing humidity for Brazil, but does not seem to change with temperature at the Mexican site. While the Mexican TDF canopy does change with VPD and humidity in the growing season, there is much more scatter in the relationship than demonstrated in the Brazilian dry forest. These trends are plotted in greater detail along with linear models in figure R3. Now looking at leaf senescence trends there are marked differences between the sites depending on the met variable. The Mexican dry forest canopy shows a high scatter negative linear relationship with temperature while no trend is apparent between these metrics at the Brazil site. However, the Brazilian dry forest canopy does show a decrease in canopy greenness with decreasing humidity and increasing VPD during leaf senescence while no such trend exists at the Mexican site for this phenophase. In relation to soil moisture content, we see very little scatter for both sites and a coincident change in both greenness and water content for the onset of senescence, once soil moisture is below 0.08 (8%) then a rapid decrease of NDVI is observed for both sites. Overall, we can see that soil moisture regimes explain the seasonal cycle of tropical dry forest leaf expression better than the air meteorological variables. However, during canopy maturity changes in air humidity and VPD explain variations in canopy greenness better than soil moisture.
When the microclimatic variables are paired with the canopy greenness on a daily basis and categorized in the four different phenophases several patterns emerge. First of all the annual canopy leaf dynamics can be observed moving from dormancy through to senescence in general counterclockwise cycle for all plots except for vapor pressure deficit (VPD) in figure R1. Starting with dormancy, greenness values remain low on the NDVI scale, between 0.40 and 0.45 for mexico and between 0.5 and 0.6 for Brazil. During the green-up a high rate of change in NDVI brings the greenness scale up near 0.80 for most of the growing season when leaves are mature. For both sites the canopy greenness remains above 0.75 during the mature leaf phenophase despite larger fluctuations in the met variables. Then towards the end of the growing season the NDVI can be seen to slowly decrease back down to the dormancy state. Except for some general trends emerging during the mature phenophase, there is little apparent relationship between the meteorological (met) variables and the state of greenness of the canopy. During green-up we see a large change in the met variables over an 8-12 day period before we observe an increase in canopy greenness. Looking at the light green points in figure R1 we see that temperature and VPD decreases while soil moisture and air humidity increase at the end of the dormancy and the start of the green-up phenophase with little change in NDVI, and then NDVI increases very quickly with little to no change in the met conditions. The only site that shows a synchronized increase during the green-up is in the Mexican soil moisture vs NDVI plot, where we see the NDVI increasing with a decreasing soil moisture content, the opposite of what we might predict. Indeed soil moisture values are a much more transient phenomenon than the other met variables and will increase rapidly following rainfall, then decrease as surface water evaporates, runs-off, and is absorbed deeper into the soil and surrounding plant roots. What we see is not a synchronized cause-effect relationship but rather an offset in meteorology, water availability and forest canopy response. This mistiming is explored in further detail in the next section (figure R2) and on the next page .
Exploring figure R1 further, during the growing season mature phenophase we see that canopy greenness decreases slightly with increasing temperature and VPD, and with decreasing humidity for Brazil, but does not seem to change with temperature at the Mexican site. While the Mexican TDF canopy does change with VPD and humidity in the growing season, there is much more scatter in the relationship than demonstrated in the Brazilian dry forest. These trends are plotted in greater detail along with linear models in figure R3. Now looking at leaf senescence trends there are marked differences between the sites depending on the met variable. The Mexican dry forest canopy shows a high scatter negative linear relationship with temperature while no trend is apparent between these metrics at the Brazil site. However, the Brazilian dry forest canopy does show a decrease in canopy greenness with decreasing humidity and increasing VPD during leaf senescence while no such trend exists at the Mexican site for this phenophase. In relation to soil moisture content, we see very little scatter for both sites and a coincident change in both greenness and water content for the onset of senescence, once soil moisture is below 0.08 (8%) then a rapid decrease of NDVI is observed for both sites. Overall, we can see that soil moisture regimes explain the seasonal cycle of tropical dry forest leaf expression better than the air meteorological variables. However, during canopy maturity changes in air humidity and VPD explain variations in canopy greenness better than soil moisture.
CORRELATION OF CANOPY GREEN-UP PHASE SHIFT WITH MICROCLIMATE
Figure R2: Canopy Greenness Time Phase Shift. Correlations between NDVI and microclimates after moving the greenness trend by 15 consecutive 24 hours periods. Offsetting the greenness time series resulted in the greatest correlation coefficients (r) with microclimate trends after about 10 days of time series mistiming. Greater correlations were only observed for soil moisture at the Mexico site. The plot on the right indicates the maximum r value and the number of associated offset days. With little trend in offset days for the Mexican dry forest besides soil moisture, the offset days for maximum temperature, humidity and vapor pressure correlation hold no significance.
To better assess the influence of moisture availability on leaf flushing during the green-up phenophase the NDVI time series was repeatedly shifted back a day at a time and then reassessed for correlation with each met variable (figure R2). At the Brazil site we observed an increasing Pearson correlation coefficient (r) each day the time series was offset for all met variables up until the 13 day, after which the correlation began to decrease. At the Mexican site only the soil moisture and canopy greenness correlation was seen to increase with the number of days the two time series were offset, there were no apparent correlation trends with temperature, humidity or VPD. Here the soil moisture r value increased from 0.48 to 0.87 after 11 days. At the Brazil site, without any phase shift the soil moisture and humidity had an r value equal to 0.48 and 0.46, respectively, which increased to 0.82 and 0.73 after eleven and thirteen day phase shifts. Similarly, temperature and VPD had respective r values of -0.39 and -0.48 before shift, and -0.67 and -0.76 after 9 and 13 day time series shifts (figure R2). With this phase shift we can see which meteorological variables are better associated with the onset of tropical dry forest leaf expression. While both sites correlated their green-up phenophase best with soil moisture content, the Mexican site shares very weak correlations with atmospheric water content and canopy green-up, the Brazil tropical dry forest green-up appears to be much more related to atmospheric meteorological transitions. An alternative and more likely cause-effect interpretation is that air temperature and humidity are dependent on the weather fronts that bring in precipitation to the continental tropical dry forest in Brazil, while rainfall and air moisture are largely independent variables on the semi-arid Pacific coast of Mexico. In either case, we see at least a one week time delay in forest vegetation response to moisture availability during the transition from the seasonal drought to the rainy season.
PEAK GROWING SEASON CANOPY GREENNESS AND MICROCLIMATE MODELS
Figure R3: Canopy-Climate Linear Models During Leaf Maturity Phenophase. Temperature, relative humidity, vapor pressure deficit, and soil moisture plotted against the canopy greenness index NDVI. Two sites depicted are a Mexican coastal tropical dry forest as red squares and a Brazilian continental dry forest as green squares. Each point is in a daily average time series which is temporally connected by the dashed lines. Daily average values were taken from the end of the green-up until the start of the senescence when canopy NDVI was greater than 0.7.
Exploring the interdependence of canopy greenness and meteorological variables during the height of the growing season in greater detail reveals further differences between the two dry forest sites (figure R3). General linear models indicate that increased temperature decreases canopy greenness at the Brazil continental site (p=0.053) but not at the Mexican coastal site (p = 0.59). The inverse of this is true for humidity where an increasing humidity relates to an increasing NDVI at both sites (p < 0.001). The daytime air humidity ranges between 70-95% for the Mexican site and between 35-98% for the Brazil site during the growing season, and excluding the first and last days of the time series, the linear model fits the humidity-NDVI profile of the Brazilian Dry forest better than the Mexican site (RMSE of 0.020 vs 0.016, df = 150). Comparing VPD to NDVI shows an increase in greenness with VPD for both sites (p < 0.001) and again a much greater range of VPD at the continental site than in the coastal canopy; a smaller change in VPD at the Mexican dry forest was associated with a larger change in canopy greenness. The greater apparent tolerance for higher VPDs in the Brazilian forest canopy is likely an adaptation to greater fluctuations in VPD throughout the growing season with the local continental climates regime. With respect to soil moisture content, both sites had greater NDVI values at higher moisture values, and interestingly, the same response rate was seen at both sites as indicated by equivalent slopes in figure R3 with a slightly greater RMSE for the Brazilian site.